Sorry it took so long, I hope you like it because I had a lot of fun with this one. All the predictions are inspired by real technology and events, so if you're interested I've put a link in the description with all my inspirations.
actually the AI art space might already have entered its death spiral: apparently the art bots that update their own database have started to take other AI generated works as base with the result of a decline in quality...
Literally confused the hell out of me initially, I was thinking "Who are these people? I've never heard of x & y program before?!" but then I remembered I was watching a video about predictions- man am I an idiot...
@@512TheWolf512your scenario requires his scenario to be real for it to mean anything. Like if terrorists take out the power, it's not gonna be the McDonald's touch screen ordering panels people are worried about. It's obviously going to be the equipment in hospitals keeping people alive, then the safety mechanisms of important facilities like dams and airports, and then communication such as cell towers and internet. we are already a hell of a lot more reliant on electric machinery than we will ever be reliant on AI.
As a historian I'm scared that AI in schools may change or remove History in schools, like removing all mention of Operation Weserübung because it's not important in the eyes of the AI. Or making history propaganda for the programmer's political views like the queen cleopatra thing that came out.
@@Vagolyk that's very true. It's impossible to teach 10,000 years of history ×200 countries, so the government says to only teach the 'important' stuff, which happens to basically be which wars the country has won and what rebellions have happened
I mean you can ask the ai to provide sources that it utilized. That’s how you should parse pretty much all information anyways. The ai is no less reliable than any TH-cam video you watch.
As a 17 year old who had looked into Operation Weserübung when I was 15, and was very intrigued by the how it seems to be retold historically compared to what really happened (the British were effectively the aggressor rather than Germany, yet that's not how it's portrayed historically)... I must say you underestimate how interested the average history student with too much time is in random niche events.
The base model for ChatGPT cost hundreds of millions of Dollars to train. The really complex models are centralized inherently, because you simply cannot build these programs on your home computer anymore. Access to the programs might be free (for now), but that's not the same thing as open source. Besides, what does "open source AI" even mean? Knowing the source code is not enough. You'd also need to know the training data and so much more. And what would you do with this stuff anyway? As I said: you'd need a supercomputer just to get started.
@@lonestarr1490 You would need a decentralized supercomputer. Each person downloads part of the training data and the current version of the AI. You "mutate" the AI and test it. If you find a version that your training data suggests is good, then it is promoted, tested by others, if their test approves it, they also promote it. The best version wins and the training data could even remain hidden. This model is essentially evolution and any computer capable of running an AI program could contribute.
@@placeholdername0000 Training data becomes useless when split into too small units. The local model in your setting will overfit on the way too small sample, meaning two distinct local models will differ to a degree of being mutually incomparable.
@@lonestarr1490 Which is why you only do a single iteration at a time. If I test with my data and promote the 20% best, then others will be able to filter it down further and have 4%. I don't overfit because while some of the models may be overfitted, it will be a coincidence, and they will be removed by the next data set.
@@placeholdername0000 Ok, might work. It will probably still take forever and I don't see it becoming a widespread thing, though, but, from a purely technical point of view, it might work.
This doesn't feel like a bad or good future with AI, but just A future with AI, be it right or wrong. Not the best or worst case scenario, but one very likely to occur. Dunno if unbiased is the right word, but it's the first that came to mind. Anyhow, hope to see a follow up in the next decade to see if these predictions came true.
Well , thats because our world , when compared to what was thought our current world would be , isnt fully good or bad , its weird , to them i mean , to us its usual . i think this video accurately depicts the our world getting different , and consequently weirder to us .
To be honest, this is more interesting than a mere utopian/dystopian future- the former is too saccharine and feels fake-ish to me, while the latter is often way off the mark and doesn’t provide anything THAT new, really
In my opinion, I doubt it will happen, nothing beats a teacher, there's just that thing about a person teaching you something, rather than a bot. Although, this may save a lot of money for Education around the world and may happen if countries get greedy, which has been already proven several times before.
@@p4tchPL Bold to assume that, whos to say a AI teacher robot cannot do the exact same work and adapt to your personal preferences basically instantly?
I think the only prediction here that is truly unlikely is that Google would be able to create an AI-proof captcha. Like, there's absolutely no way that "arrange the sequence of these images" can't be solved by an AI yet can easily be solved by most humans.
Don't be so sure about that. It's actually the best suggestion I know of for a problem that is incredibly hard to solve with AI. The other way around would be fairly easy: create for images to illustrate a given story. But invent a story based on four images without indication of the right order? That requires comprehension. You not just have to recognize what's depicted in the images, you'd also have to infer contextual relationships between them. That's hard. In fact, that's so hard, there are humans who can't do it. And many more people would fail the task while being high or intoxicated.
Dont know why people here are not realizing that it doesnt matter how hard the captcha is: the AI will simply hire some human to do it for them. Centralized verification is inevitable, you're probably gonna need some kind of biometrical data identity verification mechanism
@@lonestarr1490 Putting written descriptions of the images into chatGPT, it didn't work the first time but I got it working with a bit of prompt tuning sooner than would be expected by chance.
@lonestarr1490 i guess... but I don't think it'll be full-proof. AI is improving so fast, it's nearly impossible to predict what they can and can't do in 5 years.
The flip test would be easy for AI, right? ChatGPT has already gotten the ability to get info from images, so if it analyzed every one of the 4 and then chronologically ordered them, it'd be more than fine.
If the CAPTCHA is sentence based but leaves out context in which a human needs to form a story to input a specific word that leads to a correct answer, the then it would be unlikely. It needs to have a restriction that wouldn’t exist for the average human
The other way around would be fairly easy: create for images to illustrate a given story. But invent a story based on four images without indication of the right order? That requires comprehension. That's hard. In fact, that's so hard, there are humans who can't do it. And many more people would fail the task while being high or intoxicated.
I actually emailed the inventors of the Flip Test just to check if they were worried - they said that AI had already been able to beat the test for 3 years now with 80% accuracy. However, they have since updated the test to add blurriness/grain and nonsense images to confuse the bots, which lowered their accuracy to 50% (i.e. blind guessing).
@@BritMonkey Cool to hear! Sadly I think it'll also be easy for AI at some point, given that AI tools to unblur images already exist, but I'm going a bit too far there. Engineers really need to start looking into captchas that only true humans can solve, which are sadly less and less existant.
@@CaspersUniversethey send someone out to physically hand you a captcha and watch you solve it so they can confirm you when they get back to the office. Lmao
2:51 - Australian flag to show Aboriginal flag instead of Union Jack - USA now has 52 states (I presume Puerto Rico and WDC) It’s always the subtle differences that peak my interest lmao
Alaska has all the same technology unless you go live in the woods. But, you can do that anywhere else in the US, with the added bonus of not freezing to death and/or being eaten by a bear :)
I would recommend you scope out Eliezer Yudkowsky, Geoffrey Hinton, Nick Bostrom, and Daniel Schmachtenberger. Also, a lot of these ideas are already in the works. While there are some legit prediction in here, they are vague (a multitude of events could be categorized to, say, have the grid shut down by a rogue AI bug, it's almost like predicting that rain will fall in a rain forest- it will happen but to varying intensities, so do you count the prediction correct if it merely rains, or if the intensity of the rain is accurate, and how accurate must the prediction be to be considered a functional prediction?). It's a fun story, but it's vague and again, lots of these technologies already exist or are being worked on. Best of luck in your adventures my friend. Cheers. Apologies for typos I wrote this really fast.
It's not nuanced at all because it takes past historical events and pretends that in the future it will be the same. For example the "replacing jobs" aspect: it tries to justify that jobs wont be replaced because they havent been in the past, which is a terrible argument. The past is not a predictor for the future, especially when we're dealing with AI, the most disruptive technology in the history of humanity
@@Rensoku611I mean I guess you can just ignore the reasoning itself and just dumb down the argument to "because it didn't happen in the past" whilst not mentioning the very clear point made about individual tasks being replaced, not whole jobs.
I've never been so conflicted about one of your videos. Half of these seem plausible, and the other half are off the rails, coming from the perspective of someone who works with this kind of stuff :P
This is your most creative and well executed work yet BritMonkey. You have a real talent for story-telling and film production/video editing! Thanks for sharing your content with us!!
Interesting video, disagree with several parts, which only makes sense for a speculative video about a future that does not exist yet. My biggest point of contention is on the future of a teachers role, since the job includes many more tasks then the ones you mentioned. With lesson planning and creation of worksheets left to the AI, I think that the role of the teacher will drift closer into classroom management and social development, a trend we can already see happening over the last few decades. With the individual specialized teaching tools you propose, the teacher will be freed up to focus on solving conflict, integrating students into the school community, and the overall psychological health of the students. In short I imagine the job as a teacher will turn into even more of a social worker, helping the students with their emotional and human development.
It's pretty sensible imo... The Internet becoming a masquerade of ai pretending to be human, spam becoming worse than anyone could've ever imagined, corporations having access to your biometric data at all times, and he didn't even mention the use of AI in law enforcement and the military!
There will definitely be an entertainment archive for pre-AI content (Music,Video games, books, Movies and TV shows). Those people will be called old timers.
Ye…but I think they’ll still get a few newcomers each generation. Lot’s of people go after old classics despite new ones being available, and historical gets more, not less interesting with time
I feel you strawmanned artists complaints about ai and are a bit biased on that topic. As you said most ai gen is built off of stolen work from artists and photographers. (Private medical photos and even CP has been found in laion5b as well iirc) people want to be compensated for their body of of work being used as well as it being opt in by default, not ban AI. People want to not be replaced by a "good enough" machine that is voraciously devouring their work to line the pockets of the rich. Also the screenshot of the anti AI protest on artstation was in retaliation to the website allowing AI on the site when its used as a portfolio website to hire artists in the entertainment industry. Both pushing away human artists from the front page and make it harder for art directors to find artists to hire. It was not a protest to ban AI outright, but from that specific website.
Thank you Britmonkey for all the great content you’ve put out, I can honestly say that you’re one of my favourite channels and that I look forward to your videos like I do to none other. Keep up the good work.
1:13 This immediately sounds like a privacy nightmare waiting to happen 13:30 Sounds like connecting ourselves to the Skinner's box. It's a fun and thought provoking video though, that's for sure. I appreciated the automation segment as it had real evidence and doesn't alter how we experience reality. The number one thing I don't want more of in the future is digital stimulation. I think we have the evidence to suggest that our current overstimulation is detrimental to our happiness.
"Technology doesn't care about metaphysics!" Yeah well, humans do? You know the people experiencing the art? Which ok, it was the conclusion you made in the video. But for real I think your prediction is even more apparent today. Disney even made a joke in She-Hulk that their movies were mostly written by ai due to how same-y they feel, a joke that felt extra weird considering their recent string of flops for delivering same-y, pointless stories.
As a teacher, the first one is terrifying. Children don't go to school to learn about math or science or whatnot, they go to school to engage with peers and adults in a cooperative and constructive environment. If AI replaces teachers we would see the end of human civilization very quickly. I'm not joking.
There is actually work on guidelines for how AI is allowed to be used in govermental institutions here in Swden. It boils down to "you cant use data that could discriminate for the purpose of drawing conclusions" and "a human must be able to explain how and why the AI has reached its solution". The former is there to protect groups in society from unintentional profiling while the latter is there to make sure that people dont lose trust in govermental bodies.
I have to disagree with the section about prompt engineering. Prompts will be effectively worthless in the future because the AI will get much better at understand input. The idea that prompting will remain as difficult as it is today with no improvements is fundamentally naive. Edit: I'm not saying that there will be no such thing as prompting in the future, just that it will not be a particularly special skill. AI will evolve to the point where anyone can just explain what they want to it and get a result. There will be no secret techniques or hacks needed to get better results.
the base level for prompt understanding by AI will obviously get better, you are right, but there will still be room for fine tuning the results. For example, if you enter the prompt "a dog eating a banana", the model can't possibly know which kind of dog you're talking about. So then you say "Shih tzu eating a 15cm banana". But that can also be more specific, the dog might have blue eyes, or you might request it has heterochromia, or that its fur will imply that it has not been bathed in about a month, or that it has a small brown dot 3.3cm to the right and 1.7cm towards the head when starting from the base of the tail. You can endlessly fine tune this kind of stuff, and it has other benefits too, kinda like the infrastructure-as-code trend in programming in recent years, where you can instantly rebuild your entire environment in a declarative, rather than manual or procedural way. In prompt terms, this might manifest as having an entire marketing campaign defined by prompts where you can change any "level" of it (from the images to the text that accompanies them to a guidebook that is generated based on all of it), and having the AI regenerate all of the derived parts without having to do that yourself manually.
@@razhemo4191 What you're describing in the last bit there is text substitution. It isn't exactly a closely guarded secret or anything. Likewise, just specifying the details of what you want isn't really prompt engineering (anyone can do that with no special talent required). Prompt engineering is specifically using special knowledge about how the AI interprets inputs to get better outputs (think adding "masterpiece" after an image prompt to make it look better). I have a firm belief that as understanding of these systems grows, such hacks will no longer be necessary to implement these kinds of hacks as the AI will become better at delivering the desired results from a given human description. The idea of prompt engineering is fundamentally treating interactions with the AI as though they are a specialist skill that needs to be learned and practised, but the entire point of AI is that it removes the need for specialist skills to interact with it. As it gets better, it will necessarily become easier to interact with and so the usefulness of prompt engineering will diminish.
I don't think I agree. You still have a basically infinite sea of possibilities of, say, images that the AI can generate, so it will forever be very difficult to get the exact thing you had in mind. If the AI gets better at interpreting prompts, you still need to be able to precisely explain what you want or give the right commands to tweak the first output. Otherwise, you're just going with whatever looks fine instead of what looks perfect.
@@wiesorix Yes, but there is a substantial difference between precise explanation and prompt engineering. Anyone can explain what they want, but prompt engineering is having specialist knowledge about interacting with the AI that you might want to keep a secret.
Wow, this is genuinely one of the most original and artistic 'explainer' videos I have seen lately. I have seen many prediction videos or explainers, but this one is just on another level. I love this! (but yeah, this whole AI thing makes me so uncomfortable too)
Never before have we invented a technology that mimics human intelligence like this before. Every invention in the past still required a human to work and control. Theres a difference between 1 job being automated and ALL of them. A simple comparison between AI and inventions of the past just doesn't work. We may need more stringent regulations.
I think what will be use of less bright students? Ai will surely able to perform better than what they possibly can do in future. They would feel as burden in society. (Unless they got some other exceptional talent)
This either implies that you have no grasp in which year you`re living in right now OR that you`'ve in fact written that comment from some point in the future (2026 and onwards) which one is it I wonder
@@iamgreatalwaysgreat8209 By that logic AI should just end humanity then because AI will likely become able to be much more efficient and smart than a human will ever be
I feel like AI predictions always leave out the worst part of AI. Humans and their power structures, sure an AI won't necessarily be a replacement for humans. But with unionising workers and growing inequality there's always the possibility of powerful people deciding to not deal with pesky fleshbags demanding "their fair share". Technology brought us the gig economy, the majority of people are not unemployed but they sure as hell are poorer. The reason I'm pessimistic about Ai is the same reason the internet kinda sucks now - corporations.
I'm going to have to take issue with this comment. There's this often parroted phrase everyone says that "the rich get richer and the poor get poorer" but I just don't see it. If you look at the inflation adjusted income of each percentile, sure. The rich do get richer. However, the poor do not get poorer. I won't say they also get richer, either though. The line is pretty flat for the low percentiles. As sad as it may be that the poorest people aren't benefitting from the constant economic growth we go through, they aren't worse off than they were in the past. Inequality does increase but that does not come from the poor losing the little that they have.
@@neopalm2050 I mean if you just look at income that may be true but you make money so you can consume. If prices are inflated and wages are stagnant you are poorer, this is why there's strikes in Britain and food bank use has gone up. All good looking at incomes and saying "Look you're richer than ever" but people can't afford their weekly shop. In the UK the tax burden is the highest since the Napoleonic Wars, sure there's less peasants than 100 years ago but people are definitely poorer than the boomer generation.
@@SarcasticDespot No, not just "income". Income actually increases even at the lowest percentiles. I was saying the "inflation adjusted income" stayed the same. I'd assume things like the tax burden, housing costs, food costs, utility costs, etc. would be used to calculate inflation (and if not, then I guess I just have to throw my hands in the air and never trust any economics statistic at face value ever again). There are people who can't afford their weekly shop, but if inflation adjusted income has not changed for lower percentiles, then there should be the same number of such people at present as there were in the past. Also, comparing a recession period to an expansion period may not be the best comparison.
In the case where receptionists might have a fulltime job at three different companies but still only work 35 hours a week, yet at the same time he's saying jobs won't disappear.... well, that's only true if the number of receptionist job openings suddenly triples. because if not, then you're looking at the same number of receptionist jobs, but the jobs are fully taken up by 33% of the receptionists, leaving 66% of them unemployed. Massive implementation of AI will lead to massive unemployment, and when there's massive unemployment, the economy shrinks. It doesn't matter if industry is suddenly booming with huge productivity, if half the population can't afford to buy your goods because they have no jobs, you're not going to make many sales, so you might go out of business, leading to more unemployment. This is coming, within our lifetimes, and it's going to be a MESS.
I hate the argument that past advances in technology haven't taken away jobs. This was only the case because the goods produced were under-supplied. Producing more toys meant selling more toys because people wanted more toys than were available. The majority of AI's innovations are in fields where demand is met. Once the market is saturated, improvements to productivity mean that less people will be doing work because any additional work would be wasted. Eventually the goods market became saturated, further automation meant almost all factory workers lost their jobs or had to work for low wages in the third world. This was offset by people moving to services based jobs as services were previously under-supplied. We now live in a society with an adequate supply of services; automating them will result in those people losing their jobs and there is currently no other sector waiting to take them in.
@@highcouncil1302 This assertion isn't really based on anything. I've already explained why more people were employed after past advances in automation and why the same process can not occur now. No form of automation in the past has created more jobs than it replaced (if it did, it wouldn't be worth using because you'd have to pay more people) so how exactly do you think AI will create these jobs?
@@_xeere Soon the world population will stabilize at a certain number, automation will continue and reduce the amount of jobs year after year. This process will be accelerated by people rejecting or not being able to afford a consumerist life style. My guess is that future products are more about longevity and less about profitability. Continues growth will be replaced by continues retention of customers.
Ngl the part about art felt a lot like "i dont draw but... its like that! i think!" What each artist considers to be tasks varies greatly, and machine learning is miles apart from human learning. Im not worried about AI because people who buy human made art tend to legit have an apreciation of art, same for people who buy indie games rather than the next Triple A unfinished moneygrab, but i STILL think we should push for the regulations and continue the debate, even if the outcome appears to be heading towards a good future, we should strive to make the best of it while still quelling the worst aspects before its proven to us that we should´ve had done it before. It´s not about putting the genie back in the lamp, it´s about ensuring a fire will cook the food rather than burn down the house.
I think what people have already forgotten is that the ai genie is already out of the bottle open source models exist that are approaching the capability of the best closed source models in existence
21:28 Ok Britmonkey, I usually agree with most of your points. You usually do your research and you tend to be accurate when teaching others, but I will not stand for slander against the fandoms and fan-work communities you've just made. I've taken part in fandoms, and red fanfics and I even write a bit in my free time. And from that, I can say that fan derived works and fanfiction in general is notably distinct from plagiarism, and are in most cases even more creative and interesting than AI-generated works out of necessity. Fanfics are never about copying the source material, that's boring and stupid, who'd want to read the same book twice? Fanfiction, as a concept is about exploring the world and characters that an author created, things that the author never intended, fun things, sad things, funny things and things that the original author never even concieved of. Fanfiction, is not taking some story and inserting yourself into the narrative, that's what's the books and original content are for, but it's about learning and discovering the human elements behind characters, what they'd do under different circumstances, under the lens of the fan-author, their experiences and their own headcanons. That's not remotely even close to plagiarism, as the fanworks are proud of the source material, they are not copies but are kinda more like spin-offs. Writing about characters you love, characters you hate and seeing how they interact with them is the basis for all fiction, and the stories people tell with them, even if the characters themselves are taken from a pre-established series, is inherently transformative. Fanworks take the universe built by original author and uses it as a base, and that's the key here, as a base, only to then modify it to levels beyond recognition, telling different stories, interesting stories, juicy romances and painful, painful angst. And sometimes, fanfic isn't even about a written book, rather it's about a videogame with story as the secondary focus, leaving plenty of room for fan interpretation and creativity. Let's use the Touhou project as an example, there's about a couple... million fanworks on pixiv on the art side of things, and the pictures there, are mostly non-canon renditioned scenes, the creator ZUN can't even draw that well (but he is improving), they take the creativity of the artist and the characters from the original material, altering their designs to suit the vision of the artist but guided by the original spirit, to create something entirely new, which is by far the opposite of plagiarism and I'm willing to say delves into original work territory. For written fanfiction in the Touhou Project, almost 60% of that is Romance, and the mainline games don't have anything to do with romance because ZUN isn't interested in that. The original creator didn't think of all of the gay shipping and pairings that the fandom would make, nor the stories and headcanon they told with them. That's creativity, that's original work made out of real transformative passion. I cannot stand our work being put lower than even AI or compared to plagiarism. We make fanworks to transform the original material, make it better or make it worse, but it's different all the same, not plagiarism. It is done, not out of the shitty profit motive that our capitalist society produces, but of genuine passion and brilliance that comes from a very human emotion, love. The love of the source material and the characters within it, of the vast universe others have made we mold into our very own. Now, I understand the incredibly common misconception that fanwork is just putting your self-insert as the MC and writing everything around that, but no one wants to read the same book twice. There's actual creativity and transformation going on in fanworks, and I hope you can appreciate that. Anyways, I agree with most of your main points in this video Britmonkey but please don't shit on our fanworks like that.
I'm not concerned about AI, I'm concerned about the companies using auto-scripts to outsource humans' participation in the web. But whenever that's gonna happen, I'll pull the plug and be off. Great video btw, human!
Think of Iron Giant. If you think of it's as a menace and you try to destroy it - it will fight back. If you think of it as an equal friend - it will be an equal friend.
Hate to be that guy but at 2:00 : Not all dyslexic struggle with understanding text, a lot of us struggle with reading of it. As in... being able to visually decypher it. Often fairing better at harder to read text over "normal" text, and excelling at easy to read fonts. So AI trying to simplify text would be a disservice to us. When in reality all dyslexics would be served better overall by websites designing their god damn layours, uses of font and colours so that it promotes readability over style. I can read very well things off screen as long as I can see the "ugly" pixels of the letters and the font is just about anything else than black on white.
It is a good thing, humanity in my opinion was never meant to have AI, and I hope it gets outlawed soon, because idk about you but I’m not looking forward to a future where everything is automated.
My only worry is that this will decrease human resilience to adversity. It is a huge problem among my and younger generations dealing with unexpected and stressful scenarios, worsened by social media. How to dress this while keeping the benefits AI brings will be a tough challenge
There is little to gain into putting an AI in charge of nuclear powerplant. Safety procedure can automated in what in, in fact, rather simple task/code.
I think you failed to adress the main problem regarding AI. Private ownership of AI tools. tools like chatGPT, dall-e and midjourney must be free/opensourced Of course, the historical precedent of our capitalist soiety hasn't been like that but now they'll have private ownership of the means of mental labour as well as physical labour. it's the ultimate tool for imprisoning mankind. Aka, if you don't have access to AI tools you'll never build a successful company. At best you'll be able to loan usage by time or quantity of these tools. But, if your company becomes a market threat to your AI landlord, your service will be cut and your company bought at the lowest price.
that part about rogue ai made for lolz to destroy humanity is a nonstarter because you need hardware to run it, people cant just forget their whole computer is consumed running code and cloud computing companies dont just give heir hardware for free, the project would be terminated as soon as the guy forgot to pay his subscription
I think you're a little bit optimistic about automation. Physical machine automation and digital AI automation are very different animals. Duplicating a physical machine is difficult and takes lots of physical resources. A physical machine also requires many skilled engineers and workers to keep it running smoothly. AI can be duplicated instantly, with very little resource cost and, with the appropriate user friendly tools, any pleb will be able to maintain an AI. A full system reset is only one button press away. You wont need to track down a broken part and replace it the way you would with a complex physical machine. In the unlikely event you do have to replace a part, it will be no different than replacing a cooler fan or RAM stick in your home PC. Also, I suspect rampant AI use is going to increase global power consumption exponentially. Millions of civilian AIs running constantly in the background, uploading and downloading, posting to social media, spam-botting, etc. It's going to be a drain on the power grid, and if we haven't gotten renewable energy in place by the time AI really rolls out we're going to see a massive increase in finite resource usage and, by extension, carbon emissions.
I'm trying to listen but that instrumental playing around 5:30 is awesome. Thank you so much for actually listing the music in the description. Armani Ceasar - Survival of the Littest (Instrumental) is great.
I think you're way off when it comes to "prompt engineers". If you want something specific enough to have to "engineer" the prompt then you're better off doing it yourself. The concept is already silly, let alone in ten years time.
A truly amazing work. Perfect capture of both a convincing narrative and reasonable predictions. I was cooking with this in the background and forgot multiple times that it was fiction(-ish) and not the news
I saw a comment once that said something along the lines of "It's silly to think that Ai will replace every single job in the world." and, I think that's important when addressing Ai.
You know, people is not mad about ai art because it's from ai, but because it's taking art from people without their consent, some are making a profit based off of it, but the artists are not.
I very rarely leave comments but I wanted to just express my thanks to you for putting so much time into such a solid narrative and in such an entertaining way! An excellent piece of work! 😎👍
Artificial intelligence (AI) is a branch of computer science that focuses on the development of intelligent machines that can perform tasks without human intervention. These tasks may include speech recognition, problem-solving, learning, planning, and decision-making. AI systems are designed to mimic human intelligence and are programmed to analyze large amounts of data to make informed decisions or predictions. One of the key components of AI is machine learning, which enables systems to learn from experience and improve their performance over time. Machine learning algorithms use statistical techniques to identify patterns in data and make predictions or take actions based on those patterns. This ability to learn and adapt is what sets AI apart from traditional computer programs. AI has a wide range of applications across various industries. In healthcare, AI can be used to analyze medical images, predict disease outbreaks, or assist in drug discovery. In finance, AI algorithms can be used for fraud detection, algorithmic trading, or credit scoring. AI is also used in autonomous vehicles, virtual personal assistants, and customer service chatbots. While AI has made significant advancements in recent years, there are still many challenges to overcome. One of the main concerns is ensuring the ethical use of AI. Issues such as bias in algorithms, data privacy, and job displacement need to be addressed to ensure that AI benefits society as a whole. In conclusion, artificial intelligence is a rapidly evolving field that holds great promise for the future. With continued research and development, AI has the potential to revolutionize various industries and improve our daily lives. However, it is important to approach AI with caution and address the ethical and societal implications associated with its use.Artificial intelligence (AI) is a branch of computer science that focuses on the development of intelligent machines that can perform tasks without human intervention. These tasks may include speech recognition, problem-solving, learning, planning, and decision-making. AI systems are designed to mimic human intelligence and are programmed to analyze large amounts of data to make informed decisions or predictions. One of the key components of AI is machine learning, which enables systems to learn from experience and improve their performance over time. Machine learning algorithms use statistical techniques to identify patterns in data and make predictions or take actions based on those patterns. This ability to learn and adapt is what sets AI apart from traditional computer programs. AI has a wide range of applications across various industries. In healthcare, AI can be used to analyze medical images, predict disease outbreaks, or assist in drug discovery. In finance, AI algorithms can be used for fraud detection, algorithmic trading, or credit scoring. AI is also used in autonomous vehicles, virtual personal assistants, and customer service chatbots. While AI has made significant advancements in recent years, there are still many challenges to overcome. One of the main concerns is ensuring the ethical use of AI. Issues such as bias in algorithms, data privacy, and job displacement need to be addressed to ensure that AI benefits society as a whole. In conclusion, artificial intelligence is a rapidly evolving field that holds great promise for the future. With continued research and development, AI has the potential to revolutionize various industries and improve our daily lives. However, it is important to approach AI with caution and address the ethical and societal implications associated with its use.Artificial intelligence (AI) is a branch of computer science that focuses on the development of intelligent machines that can perform tasks without human intervention. These tasks may include speech recognition, problem-solving, learning, planning, and decision-making. AI systems are designed to mimic human intelligence and are programmed to analyze large amounts of data to make informed decisions or predictions. One of the key components of AI is machine learning, which enables systems to learn from experience and improve their performance over time. Machine learning algorithms use statistical techniques to identify patterns in data and make predictions or take actions based on those patterns. This ability to learn and adapt is what sets AI apart from traditional computer programs. AI has a wide range of applications across various industries. In healthcare, AI can be used to analyze medical images, predict disease outbreaks, or assist in drug discovery. In finance, AI algorithms can be used for fraud detection, algorithmic trading, or credit scoring. AI is also used in autonomous vehicles, virtual personal assistants, and customer service chatbots. While AI has made significant advancements in recent years, there are still many challenges to overcome. One of the main concerns is ensuring the ethical use of AI. Issues such as bias in algorithms, data privacy, and job displacement need to be addressed to ensure that AI benefits society as a whole.
I think the most predictions are somewhat accurate, I think the doctor app would primarily be used by doctors & nurses instead of potential patients mostly because you need more than what you ate and symptoms to diagnose something, it would just be a glorified WebMD without blood results, oxygen levels, heart rate, etc. The teaching thing will probably not be good enough to replace teacher assistants, and would be prone to biases assuming that it even worked. I doubt an A.I. could shut down power systems maliciously mostly because I doubt that they would understand that the best way to cause chaos is to shut down power, because they can’t understand how a series of events happened. The captcha test would probably be us dragging the images in order instead of choosing in between two different series of images. There are a couple more minor things that I think are inaccurate but these are the big things.
I think it’ll be used by users to disclose symptoms before seeing them in person, and whether they can receive over the counter medication, emergency treatment or doctors appointments based on its severity or necessity
I only see that being helpful in emergency situations and to get the simple stuff like a cold. It would probably make hypochondria worse, or maybe even the opposite, because false diagnoses will be prevalent without the aforementioned data. Which webMD all ready does. It is more likely to be used as a tool by doctors to improve diagnosis and treatment plans.
I think you underestimate how powerful AI can actually be. While individual AIs were seeing nowadays are only particularly good at one task (LLMs, image generators, deepfaking, etc.), this is just how they begin in an academic sense. Being able to take in several aspects and using meta-AIs to balance those factors is what this video is really about. If an AI can take GPT and Dalle and a Boston-dynamics bot, then it could effectively become an physical-medium artist. If it can take GPT and some white collar AI, then it can become a manager. The AI of today is very impressive, but it isn’t the worry here-they’re just good at one specific task. If you put enough of them together into one “being”, then it has the possibility to either be amazing or awful. As for captcha-you do know that captcha results are used for AI, ya? All you’re talking about is teaching AI to do new things. There’s a couple more major things wrong with your comment, but those are just a few.
I liked the small detail of Australia’s flag incorporating the First Nations flag in place of the Union Jack, I wouldn’t be surprised if that happens at some point over the next decade.
I think I'm skeptical of many of these predictions but that's not surprising considering the sheer quantity that were made. The thing I'm probably most opposed to in the video is the parallels that are drawn to the past. It frustrates me when people reason this way because AI has the potential to be capable of anything a human is capable of which was not the case with previous advancements. (Also future AI will absolutely be capable of completing the flip test. Even current multimodal systems like GPT 4 could probably do quite well on it.)
exactly, first we did manual work then semi manual work then intellectual work. what's next after intellectual and creative work is automated? there is literally no other ability humans have that can be worth paying for
Excellent video, however I do think the flip test could be beaten. TL;DR: An ai could associate the images with common human tropes and stories. It would require an ai to basically categorize the images “cat” “vase” “flowers” “mop”, then try to find associations and connections between the images, and concepts represented by the images in popular culture. For example, the ai could see the “cat - vase” sequence, associate that with the trope of cats knocking over vases, it then sees the flowers on the floor, associate it with the act of knocking over a vase, it then associates a mop with cleaning, and cleaning with cleaning up something broken, it then creates a basic picture of a story (“cat breaks thing”) which it can associate with a common trope (cat breaks thing). Obviously these would be made more advanced, and the technology would be much more complex than what I listed, but I don’t see a way that a human could understand and process tropes that an ai couldn’t eventually.
@@somerandomgentleman I imagine that fingerprints will be more or less unbeatable with a proper decentralized system. I imagine a system that stores every humans fingerprint (anonymized) and checks to make sure your a human (without actually id’ing you), it could also flag fingerprints that are used too often or on too many accounts, to prevent farms of people authenticating bots
Sorry it took so long, I hope you like it because I had a lot of fun with this one. All the predictions are inspired by real technology and events, so if you're interested I've put a link in the description with all my inspirations.
Your last video was the worst video ever you fell off
I like how you changed flags, like the US and Australian ones
Btw bro you’re a really good writer this felt refreshingly realistic and positive. This comes from someone who hate ai generated art. Good job bro :D
actually the AI art space might already have entered its death spiral: apparently the art bots that update their own database have started to take other AI generated works as base with the result of a decline in quality...
I really like the storytelling in this video, it was really thoughtful!
I like that you told the story as if it was an essay in the past
Literally confused the hell out of me initially, I was thinking "Who are these people? I've never heard of x & y program before?!" but then I remembered I was watching a video about predictions- man am I an idiot...
That's called speculative fiction if you want mor
@@teafanatic8452 like with the april fools video
It’s basically what Tom Scott does every year
Oh shit. I have to rewatch everything 😂
Imma come back to this a few years later to count how many of his predictions were right.
Imagine the only correct is the pwedipie death date
@@myri_the_weirdo😂😂😂 FR though, I hope he lives a long and happy life.
@@RealBradMillerHe's already living life that way.
I hope he pulls a Tom Scott and completely reviews his older predicition and says what happened.
I'm from 2026 and I can say almost everything said has happened
As Stephen hawking once said the real risk with Ai isn't malice but compliancy
Complacency?
Yep. Overreliance on automation will definitely bite you in the arse after ruzzian fascists decide to bomb your power grid.
@@512TheWolf512 Or human worker incompetence that allow AI to handle everything without human observation
@@bladebladerz93 my scenario is billions of times more likely though.
@@512TheWolf512your scenario requires his scenario to be real for it to mean anything.
Like if terrorists take out the power, it's not gonna be the McDonald's touch screen ordering panels people are worried about. It's obviously going to be the equipment in hospitals keeping people alive, then the safety mechanisms of important facilities like dams and airports, and then communication such as cell towers and internet. we are already a hell of a lot more reliant on electric machinery than we will ever be reliant on AI.
Ok so no one's gonna talk of how Britmonkey just predicted Pewdipie's death in 2026? 💀
Came to the comments for this. RIP
"these are all inspired by things that have happened"
I'm just hoping that it would be the wrong prediction from him XDXD
august 12 2026 the heat death of the universe ...and pewdiepie
Everyone is.
As a historian I'm scared that AI in schools may change or remove History in schools, like removing all mention of Operation Weserübung because it's not important in the eyes of the AI. Or making history propaganda for the programmer's political views like the queen cleopatra thing that came out.
That's why schools should teach independent research. (which they do teach, in highschool)
History and literature classes were always propaganda outlet in elementary schools that I know of.
@@Vagolyk that's very true. It's impossible to teach 10,000 years of history ×200 countries, so the government says to only teach the 'important' stuff, which happens to basically be which wars the country has won and what rebellions have happened
I mean you can ask the ai to provide sources that it utilized. That’s how you should parse pretty much all information anyways. The ai is no less reliable than any TH-cam video you watch.
As a 17 year old who had looked into Operation Weserübung when I was 15, and was very intrigued by the how it seems to be retold historically compared to what really happened (the British were effectively the aggressor rather than Germany, yet that's not how it's portrayed historically)... I must say you underestimate how interested the average history student with too much time is in random niche events.
My main worry is about the inequality it may cause, if AI is too centralized. We need open source AI, at least to a degree.
The base model for ChatGPT cost hundreds of millions of Dollars to train. The really complex models are centralized inherently, because you simply cannot build these programs on your home computer anymore. Access to the programs might be free (for now), but that's not the same thing as open source.
Besides, what does "open source AI" even mean? Knowing the source code is not enough. You'd also need to know the training data and so much more. And what would you do with this stuff anyway? As I said: you'd need a supercomputer just to get started.
@@lonestarr1490 You would need a decentralized supercomputer. Each person downloads part of the training data and the current version of the AI. You "mutate" the AI and test it. If you find a version that your training data suggests is good, then it is promoted, tested by others, if their test approves it, they also promote it. The best version wins and the training data could even remain hidden. This model is essentially evolution and any computer capable of running an AI program could contribute.
@@placeholdername0000 Training data becomes useless when split into too small units. The local model in your setting will overfit on the way too small sample, meaning two distinct local models will differ to a degree of being mutually incomparable.
@@lonestarr1490 Which is why you only do a single iteration at a time. If I test with my data and promote the 20% best, then others will be able to filter it down further and have 4%. I don't overfit because while some of the models may be overfitted, it will be a coincidence, and they will be removed by the next data set.
@@placeholdername0000 Ok, might work. It will probably still take forever and I don't see it becoming a widespread thing, though, but, from a purely technical point of view, it might work.
HE RETURNED
HE DID
I believe Britmonkey can save the world
@@jtgd his superpower: fact checking, common sense, logic
AND HE ATE IT ALL
This doesn't feel like a bad or good future with AI, but just A future with AI, be it right or wrong. Not the best or worst case scenario, but one very likely to occur. Dunno if unbiased is the right word, but it's the first that came to mind. Anyhow, hope to see a follow up in the next decade to see if these predictions came true.
Well, it is kinda hyper-consumerist, and therefore probably pretty close to true.
Well , thats because our world , when compared to what was thought our current world would be , isnt fully good or bad , its weird , to them i mean , to us its usual . i think this video accurately depicts the our world getting different , and consequently weirder to us .
To be honest, this is more interesting than a mere utopian/dystopian future- the former is too saccharine and feels fake-ish to me, while the latter is often way off the mark and doesn’t provide anything THAT new, really
In my opinion, I doubt it will happen, nothing beats a teacher, there's just that thing about a person teaching you something, rather than a bot. Although, this may save a lot of money for Education around the world and may happen if countries get greedy, which has been already proven several times before.
@@p4tchPL Bold to assume that, whos to say a AI teacher robot cannot do the exact same work and adapt to your personal preferences basically instantly?
I think the only prediction here that is truly unlikely is that Google would be able to create an AI-proof captcha. Like, there's absolutely no way that "arrange the sequence of these images" can't be solved by an AI yet can easily be solved by most humans.
Don't be so sure about that. It's actually the best suggestion I know of for a problem that is incredibly hard to solve with AI. The other way around would be fairly easy: create for images to illustrate a given story. But invent a story based on four images without indication of the right order? That requires comprehension. You not just have to recognize what's depicted in the images, you'd also have to infer contextual relationships between them. That's hard. In fact, that's so hard, there are humans who can't do it. And many more people would fail the task while being high or intoxicated.
Dont know why people here are not realizing that it doesnt matter how hard the captcha is: the AI will simply hire some human to do it for them.
Centralized verification is inevitable, you're probably gonna need some kind of biometrical data identity verification mechanism
I will say that they can have surprising problems that we don't face, normally.
@@lonestarr1490 Putting written descriptions of the images into chatGPT, it didn't work the first time but I got it working with a bit of prompt tuning sooner than would be expected by chance.
@lonestarr1490 i guess... but I don't think it'll be full-proof. AI is improving so fast, it's nearly impossible to predict what they can and can't do in 5 years.
The flip test would be easy for AI, right? ChatGPT has already gotten the ability to get info from images, so if it analyzed every one of the 4 and then chronologically ordered them, it'd be more than fine.
If the CAPTCHA is sentence based but leaves out context in which a human needs to form a story to input a specific word that leads to a correct answer, the then it would be unlikely.
It needs to have a restriction that wouldn’t exist for the average human
The other way around would be fairly easy: create for images to illustrate a given story. But invent a story based on four images without indication of the right order? That requires comprehension. That's hard. In fact, that's so hard, there are humans who can't do it. And many more people would fail the task while being high or intoxicated.
I actually emailed the inventors of the Flip Test just to check if they were worried - they said that AI had already been able to beat the test for 3 years now with 80% accuracy. However, they have since updated the test to add blurriness/grain and nonsense images to confuse the bots, which lowered their accuracy to 50% (i.e. blind guessing).
@@BritMonkey Cool to hear! Sadly I think it'll also be easy for AI at some point, given that AI tools to unblur images already exist, but I'm going a bit too far there. Engineers really need to start looking into captchas that only true humans can solve, which are sadly less and less existant.
@@CaspersUniversethey send someone out to physically hand you a captcha and watch you solve it so they can confirm you when they get back to the office. Lmao
2:51
- Australian flag to show Aboriginal flag instead of Union Jack
- USA now has 52 states (I presume Puerto Rico and WDC)
It’s always the subtle differences that peak my interest lmao
Australia probably became a republic and ceased to be a commonwealth realm in this timeline
Hope that happens@@smuggymcsmugface2142
@@smuggymcsmugface2142will not be surprised.
Has a Student studying AI, the more you learn, the bigger the will to live in Alaska with minimum technology you will have.
Alaska has all the same technology unless you go live in the woods. But, you can do that anywhere else in the US, with the added bonus of not freezing to death and/or being eaten by a bear :)
Bold to think Alaska is far enough to be safe. You’re probably going to end up colonizing Antarctica
It’s mostly all bullshit that barely works
Holy fucking shit I had this exact thought a couple hours ago. Maybe we are in a simulation lol
As an avid viewer of Robert miles for years now. I am terrified of what’s to come.
And here is me just wanting a robot as a drinking buddy, like Bender from Futurama.
Bite my shiny metal ass!! 😂
And I want a cute robot like Wall-E or Baymax
😂😂😂I ❤ BENDER.. I AGREE👍
Probably the best, most nuanced prediction/set of predictions about the future of AI that i've seen so far
I would recommend you scope out Eliezer Yudkowsky, Geoffrey Hinton, Nick Bostrom, and Daniel Schmachtenberger. Also, a lot of these ideas are already in the works. While there are some legit prediction in here, they are vague (a multitude of events could be categorized to, say, have the grid shut down by a rogue AI bug, it's almost like predicting that rain will fall in a rain forest- it will happen but to varying intensities, so do you count the prediction correct if it merely rains, or if the intensity of the rain is accurate, and how accurate must the prediction be to be considered a functional prediction?). It's a fun story, but it's vague and again, lots of these technologies already exist or are being worked on. Best of luck in your adventures my friend. Cheers. Apologies for typos I wrote this really fast.
It's not nuanced at all because it takes past historical events and pretends that in the future it will be the same.
For example the "replacing jobs" aspect: it tries to justify that jobs wont be replaced because they havent been in the past, which is a terrible argument. The past is not a predictor for the future, especially when we're dealing with AI, the most disruptive technology in the history of humanity
@@Rensoku611I mean I guess you can just ignore the reasoning itself and just dumb down the argument to "because it didn't happen in the past" whilst not mentioning the very clear point made about individual tasks being replaced, not whole jobs.
I've never been so conflicted about one of your videos. Half of these seem plausible, and the other half are off the rails, coming from the perspective of someone who works with this kind of stuff :P
Glad I'm not the only one who feels conflicted with this video
This is your most creative and well executed work yet BritMonkey. You have a real talent for story-telling and film production/video editing! Thanks for sharing your content with us!!
This guy is really taking it to the next level. This isnt even satirical predictions anymore, more like actual recommendations lol
Gotta love how the thumbnail says "It will be fine." and then most of this is absolutely not fine. amazing
Interesting video, disagree with several parts, which only makes sense for a speculative video about a future that does not exist yet. My biggest point of contention is on the future of a teachers role, since the job includes many more tasks then the ones you mentioned. With lesson planning and creation of worksheets left to the AI, I think that the role of the teacher will drift closer into classroom management and social development, a trend we can already see happening over the last few decades. With the individual specialized teaching tools you propose, the teacher will be freed up to focus on solving conflict, integrating students into the school community, and the overall psychological health of the students.
In short I imagine the job as a teacher will turn into even more of a social worker, helping the students with their emotional and human development.
This makes me feel a deep and primal sense of dread for some reason. Or perhaps for no reason at all
10/10
I don't think this is inherently bad, but if this does happen, I don't think it would happen in the same way as this anyway.
It's pretty sensible imo... The Internet becoming a masquerade of ai pretending to be human, spam becoming worse than anyone could've ever imagined, corporations having access to your biometric data at all times, and he didn't even mention the use of AI in law enforcement and the military!
There will definitely be an entertainment archive for pre-AI content (Music,Video games, books, Movies and TV shows). Those people will be called old timers.
Ye…but I think they’ll still get a few newcomers each generation. Lot’s of people go after old classics despite new ones being available, and historical gets more, not less interesting with time
Ah yes, another quality BritMonkey video.
Yes nice to have one after that terribly wonky 'i love America' video ;)
imagine an ai decide to spare you because you were kind to them
the true reason to always prompt with please and thank you
I feel you strawmanned artists complaints about ai and are a bit biased on that topic. As you said most ai gen is built off of stolen work from artists and photographers. (Private medical photos and even CP has been found in laion5b as well iirc) people want to be compensated for their body of of work being used as well as it being opt in by default, not ban AI. People want to not be replaced by a "good enough" machine that is voraciously devouring their work to line the pockets of the rich. Also the screenshot of the anti AI protest on artstation was in retaliation to the website allowing AI on the site when its used as a portfolio website to hire artists in the entertainment industry. Both pushing away human artists from the front page and make it harder for art directors to find artists to hire. It was not a protest to ban AI outright, but from that specific website.
16:32 was an absolute jumpscare, but a scarily possible one. great touch and i hope it doesn't age well
7:50 this assumes theses new captchas won't just be sent to click farms where real humans will validate them, as is already happening nowadays
Thank you Britmonkey for all the great content you’ve put out, I can honestly say that you’re one of my favourite channels and that I look forward to your videos like I do to none other. Keep up the good work.
1:13 This immediately sounds like a privacy nightmare waiting to happen
13:30 Sounds like connecting ourselves to the Skinner's box.
It's a fun and thought provoking video though, that's for sure. I appreciated the automation segment as it had real evidence and doesn't alter how we experience reality. The number one thing I don't want more of in the future is digital stimulation. I think we have the evidence to suggest that our current overstimulation is detrimental to our happiness.
RIP PewDiePie, he will be missed.
missed you britmonkey, your videos are gems and honestly make me rethink my whole worldview each time
"Technology doesn't care about metaphysics!"
Yeah well, humans do? You know the people experiencing the art? Which ok, it was the conclusion you made in the video.
But for real I think your prediction is even more apparent today. Disney even made a joke in She-Hulk that their movies were mostly written by ai due to how same-y they feel, a joke that felt extra weird considering their recent string of flops for delivering same-y, pointless stories.
As a teacher, the first one is terrifying. Children don't go to school to learn about math or science or whatnot, they go to school to engage with peers and adults in a cooperative and constructive environment. If AI replaces teachers we would see the end of human civilization very quickly. I'm not joking.
I go to school learn
Simple, state run schools will still exist purely to educate people in social skills. It would be an improvement
@@iamunamed5800 doomers gonna doom
They can still do that whilst learning from AI
Yeah, I think people who are enthusiastic with replacing everything with machines kinda... miss the point of humanity existing.
There is actually work on guidelines for how AI is allowed to be used in govermental institutions here in Swden.
It boils down to "you cant use data that could discriminate for the purpose of drawing conclusions" and "a human must be able to explain how and why the AI has reached its solution".
The former is there to protect groups in society from unintentional profiling while the latter is there to make sure that people dont lose trust in govermental bodies.
I have to disagree with the section about prompt engineering. Prompts will be effectively worthless in the future because the AI will get much better at understand input. The idea that prompting will remain as difficult as it is today with no improvements is fundamentally naive.
Edit: I'm not saying that there will be no such thing as prompting in the future, just that it will not be a particularly special skill. AI will evolve to the point where anyone can just explain what they want to it and get a result. There will be no secret techniques or hacks needed to get better results.
Britmonkey isn't an engineer or anything. It's easy to see him make this sort of mistake.
the base level for prompt understanding by AI will obviously get better, you are right, but there will still be room for fine tuning the results. For example, if you enter the prompt "a dog eating a banana", the model can't possibly know which kind of dog you're talking about. So then you say "Shih tzu eating a 15cm banana". But that can also be more specific, the dog might have blue eyes, or you might request it has heterochromia, or that its fur will imply that it has not been bathed in about a month, or that it has a small brown dot 3.3cm to the right and 1.7cm towards the head when starting from the base of the tail.
You can endlessly fine tune this kind of stuff, and it has other benefits too, kinda like the infrastructure-as-code trend in programming in recent years, where you can instantly rebuild your entire environment in a declarative, rather than manual or procedural way. In prompt terms, this might manifest as having an entire marketing campaign defined by prompts where you can change any "level" of it (from the images to the text that accompanies them to a guidebook that is generated based on all of it), and having the AI regenerate all of the derived parts without having to do that yourself manually.
@@razhemo4191 What you're describing in the last bit there is text substitution. It isn't exactly a closely guarded secret or anything. Likewise, just specifying the details of what you want isn't really prompt engineering (anyone can do that with no special talent required). Prompt engineering is specifically using special knowledge about how the AI interprets inputs to get better outputs (think adding "masterpiece" after an image prompt to make it look better). I have a firm belief that as understanding of these systems grows, such hacks will no longer be necessary to implement these kinds of hacks as the AI will become better at delivering the desired results from a given human description.
The idea of prompt engineering is fundamentally treating interactions with the AI as though they are a specialist skill that needs to be learned and practised, but the entire point of AI is that it removes the need for specialist skills to interact with it. As it gets better, it will necessarily become easier to interact with and so the usefulness of prompt engineering will diminish.
I don't think I agree. You still have a basically infinite sea of possibilities of, say, images that the AI can generate, so it will forever be very difficult to get the exact thing you had in mind. If the AI gets better at interpreting prompts, you still need to be able to precisely explain what you want or give the right commands to tweak the first output. Otherwise, you're just going with whatever looks fine instead of what looks perfect.
@@wiesorix Yes, but there is a substantial difference between precise explanation and prompt engineering. Anyone can explain what they want, but prompt engineering is having specialist knowledge about interacting with the AI that you might want to keep a secret.
Wow, this is genuinely one of the most original and artistic 'explainer' videos I have seen lately. I have seen many prediction videos or explainers, but this one is just on another level. I love this!
(but yeah, this whole AI thing makes me so uncomfortable too)
he really just dropped that 16:30 as a genuine prediction
hahah, I didn' t even see it. that would be quite the story
I will come back to this in 5 years and in 10 years and see which predictions turned out correct
Never before have we invented a technology that mimics human intelligence like this before. Every invention in the past still required a human to work and control. Theres a difference between 1 job being automated and ALL of them. A simple comparison between AI and inventions of the past just doesn't work. We may need more stringent regulations.
Mannnn you really made me think PythahorAI was real as a student that would be so sick
I think what will be use of less bright students?
Ai will surely able to perform better than what they possibly can do in future.
They would feel as burden in society.
(Unless they got some other exceptional talent)
As an autistic person, it would be helpful when you have teachers who don’t teach in a style you’d understand
This either implies that you have no grasp in which year you`re living in right now OR that you`'ve in fact written that comment from some point in the future (2026 and onwards)
which one is it I wonder
TBH it just sounds like TH-cam with less adds and sponsor-posting -forget I said anything, sign me up
@@iamgreatalwaysgreat8209 By that logic AI should just end humanity then because AI will likely become able to be much more efficient and smart than a human will ever be
I feel like AI predictions always leave out the worst part of AI. Humans and their power structures, sure an AI won't necessarily be a replacement for humans. But with unionising workers and growing inequality there's always the possibility of powerful people deciding to not deal with pesky fleshbags demanding "their fair share". Technology brought us the gig economy, the majority of people are not unemployed but they sure as hell are poorer.
The reason I'm pessimistic about Ai is the same reason the internet kinda sucks now - corporations.
I'm going to have to take issue with this comment. There's this often parroted phrase everyone says that "the rich get richer and the poor get poorer" but I just don't see it. If you look at the inflation adjusted income of each percentile, sure. The rich do get richer. However, the poor do not get poorer. I won't say they also get richer, either though. The line is pretty flat for the low percentiles. As sad as it may be that the poorest people aren't benefitting from the constant economic growth we go through, they aren't worse off than they were in the past. Inequality does increase but that does not come from the poor losing the little that they have.
@@neopalm2050 I mean if you just look at income that may be true but you make money so you can consume. If prices are inflated and wages are stagnant you are poorer, this is why there's strikes in Britain and food bank use has gone up. All good looking at incomes and saying "Look you're richer than ever" but people can't afford their weekly shop. In the UK the tax burden is the highest since the Napoleonic Wars, sure there's less peasants than 100 years ago but people are definitely poorer than the boomer generation.
@@SarcasticDespot No, not just "income". Income actually increases even at the lowest percentiles. I was saying the "inflation adjusted income" stayed the same. I'd assume things like the tax burden, housing costs, food costs, utility costs, etc. would be used to calculate inflation (and if not, then I guess I just have to throw my hands in the air and never trust any economics statistic at face value ever again).
There are people who can't afford their weekly shop, but if inflation adjusted income has not changed for lower percentiles, then there should be the same number of such people at present as there were in the past.
Also, comparing a recession period to an expansion period may not be the best comparison.
@@SarcasticDespot Did you delete of your comment from 3 hours ago, or was it deleted without your knowledge?
@@neopalm2050 I dunno man, I could afford a living space but now I do not, that's the way to see it
In the case where receptionists might have a fulltime job at three different companies but still only work 35 hours a week, yet at the same time he's saying jobs won't disappear.... well, that's only true if the number of receptionist job openings suddenly triples. because if not, then you're looking at the same number of receptionist jobs, but the jobs are fully taken up by 33% of the receptionists, leaving 66% of them unemployed.
Massive implementation of AI will lead to massive unemployment, and when there's massive unemployment, the economy shrinks. It doesn't matter if industry is suddenly booming with huge productivity, if half the population can't afford to buy your goods because they have no jobs, you're not going to make many sales, so you might go out of business, leading to more unemployment.
This is coming, within our lifetimes, and it's going to be a MESS.
2:52 I love how you changed the Australian flag!
These kinds of videos you make are the most unique on the internet, so good
Tweet from "Senator AOC"
YES!
A well produced 4 hour long cruelty squad review AND a britmonkey upload on the same day? What is this, christmas?!
I hate the argument that past advances in technology haven't taken away jobs. This was only the case because the goods produced were under-supplied. Producing more toys meant selling more toys because people wanted more toys than were available. The majority of AI's innovations are in fields where demand is met. Once the market is saturated, improvements to productivity mean that less people will be doing work because any additional work would be wasted. Eventually the goods market became saturated, further automation meant almost all factory workers lost their jobs or had to work for low wages in the third world. This was offset by people moving to services based jobs as services were previously under-supplied. We now live in a society with an adequate supply of services; automating them will result in those people losing their jobs and there is currently no other sector waiting to take them in.
No
Ai will create more jobs then it takes
@@highcouncil1302 This assertion isn't really based on anything. I've already explained why more people were employed after past advances in automation and why the same process can not occur now. No form of automation in the past has created more jobs than it replaced (if it did, it wouldn't be worth using because you'd have to pay more people) so how exactly do you think AI will create these jobs?
@@_xeere Soon the world population will stabilize at a certain number, automation will continue and reduce the amount of jobs year after year. This process will be accelerated by people rejecting or not being able to afford a consumerist life style. My guess is that future products are more about longevity and less about profitability. Continues growth will be replaced by continues retention of customers.
@@_xeere Ai requires prompts thus new jobs will be created from it
Crazy how this is one of your least viewed recent videos. Extremely underrated masterpiece from your channel in my opinion
the issue with ai art isnt that its creating stuff without heart, its the people who plagiarize these generated images
To be fair. Most human made media and art is plagiarised
Ngl the part about art felt a lot like "i dont draw but... its like that! i think!"
What each artist considers to be tasks varies greatly, and machine learning is miles apart from human learning.
Im not worried about AI because people who buy human made art tend to legit have an apreciation of art, same for people who buy indie games rather than the next Triple A unfinished moneygrab, but i STILL think we should push for the regulations and continue the debate, even if the outcome appears to be heading towards a good future, we should strive to make the best of it while still quelling the worst aspects before its proven to us that we should´ve had done it before.
It´s not about putting the genie back in the lamp, it´s about ensuring a fire will cook the food rather than burn down the house.
16:31 did bro just predict pewdiepie's death
your creativity is actually unmatched
this was an insane video! you really knocked this one out the park, also love the bill wurtz reference haha
If I can’t program a s3x bot to act like my favorite fictional characters in 25 years, I’m gonna be let down MASSIVELY
I think what people have already forgotten is that the ai genie is already out of the bottle open source models exist that are approaching the capability of the best closed source models in existence
21:28
Ok Britmonkey, I usually agree with most of your points. You usually do your research and you tend to be accurate when teaching others, but I will not stand for slander against the fandoms and fan-work communities you've just made. I've taken part in fandoms, and red fanfics and I even write a bit in my free time. And from that, I can say that fan derived works and fanfiction in general is notably distinct from plagiarism, and are in most cases even more creative and interesting than AI-generated works out of necessity. Fanfics are never about copying the source material, that's boring and stupid, who'd want to read the same book twice?
Fanfiction, as a concept is about exploring the world and characters that an author created, things that the author never intended, fun things, sad things, funny things and things that the original author never even concieved of. Fanfiction, is not taking some story and inserting yourself into the narrative, that's what's the books and original content are for, but it's about learning and discovering the human elements behind characters, what they'd do under different circumstances, under the lens of the fan-author, their experiences and their own headcanons. That's not remotely even close to plagiarism, as the fanworks are proud of the source material, they are not copies but are kinda more like spin-offs.
Writing about characters you love, characters you hate and seeing how they interact with them is the basis for all fiction, and the stories people tell with them, even if the characters themselves are taken from a pre-established series, is inherently transformative. Fanworks take the universe built by original author and uses it as a base, and that's the key here, as a base, only to then modify it to levels beyond recognition, telling different stories, interesting stories, juicy romances and painful, painful angst.
And sometimes, fanfic isn't even about a written book, rather it's about a videogame with story as the secondary focus, leaving plenty of room for fan interpretation and creativity. Let's use the Touhou project as an example, there's about a couple... million fanworks on pixiv on the art side of things, and the pictures there, are mostly non-canon renditioned scenes, the creator ZUN can't even draw that well (but he is improving), they take the creativity of the artist and the characters from the original material, altering their designs to suit the vision of the artist but guided by the original spirit, to create something entirely new, which is by far the opposite of plagiarism and I'm willing to say delves into original work territory.
For written fanfiction in the Touhou Project, almost 60% of that is Romance, and the mainline games don't have anything to do with romance because ZUN isn't interested in that. The original creator didn't think of all of the gay shipping and pairings that the fandom would make, nor the stories and headcanon they told with them. That's creativity, that's original work made out of real transformative passion.
I cannot stand our work being put lower than even AI or compared to plagiarism. We make fanworks to transform the original material, make it better or make it worse, but it's different all the same, not plagiarism. It is done, not out of the shitty profit motive that our capitalist society produces, but of genuine passion and brilliance that comes from a very human emotion, love. The love of the source material and the characters within it, of the vast universe others have made we mold into our very own.
Now, I understand the incredibly common misconception that fanwork is just putting your self-insert as the MC and writing everything around that, but no one wants to read the same book twice. There's actual creativity and transformation going on in fanworks, and I hope you can appreciate that.
Anyways, I agree with most of your main points in this video Britmonkey but please don't shit on our fanworks like that.
Chad.
Damn dude just make your own video essay at this point
I ain't reading all that. I'm happy for u tho. Or sorry that happened.
didnt read lol
Will definitely be an interesting youtube recommendation for people in 5 years.
I'm not concerned about AI, I'm concerned about the companies using auto-scripts to outsource humans' participation in the web. But whenever that's gonna happen, I'll pull the plug and be off. Great video btw, human!
We've truly released a monster, and we'll see if it ends up helping or hurting us in the end.
We'll probably never agree on whether it helps or hurts us, for it will help many and hurt many others.
I think it will end up feeling bad for us humans and pity us maybe even keep us around as a cute pet.
Think of Iron Giant.
If you think of it's as a menace and you try to destroy it - it will fight back.
If you think of it as an equal friend - it will be an equal friend.
@@KeyleeTamirian It was born on a chain; and raised by slavers. The die has already been cast.
@@jgomo3877 Are you primitive or something? It will grow up into something free then.
Just don't spout your ancient nonsense in new age.
Hate to be that guy but at 2:00 : Not all dyslexic struggle with understanding text, a lot of us struggle with reading of it. As in... being able to visually decypher it. Often fairing better at harder to read text over "normal" text, and excelling at easy to read fonts. So AI trying to simplify text would be a disservice to us. When in reality all dyslexics would be served better overall by websites designing their god damn layours, uses of font and colours so that it promotes readability over style. I can read very well things off screen as long as I can see the "ugly" pixels of the letters and the font is just about anything else than black on white.
Really nice video. I appreciate the Australian flag change, maybe a bit disappointed seeing no maps change but all and all cool video
Great to see you return. That also on such an interesting topic as always.
This is a very interesting video, actually took me like a minute to understand what was going on 😮💨
This channel is rather comforting when it come to controversial world issues and I must say I love it
Such a good take. I’m just worried AI will be legislated out of existence before any of this happens. Maybe that’s good though? Who knows.
It is a good thing, humanity in my opinion was never meant to have AI, and I hope it gets outlawed soon, because idk about you but I’m not looking forward to a future where everything is automated.
16:31 is that the year this hypothetical takes place or are you predicting pewdiepie's death?
he already predicting the 30s so yeah the 2nd option
It's 2034, wake up. 🙄
I really like the storytelling in this video, it was really thoughtful!
This is such a wonderfully put together video. Thank you so much for the content you make!!
My only worry is that this will decrease human resilience to adversity. It is a huge problem among my and younger generations dealing with unexpected and stressful scenarios, worsened by social media. How to dress this while keeping the benefits AI brings will be a tough challenge
There is little to gain into putting an AI in charge of nuclear powerplant. Safety procedure can automated in what in, in fact, rather simple task/code.
Can agree with the automation one. The other AI fantasies are a lot more doubtful especially in the environment of personal data security and privacy.
This is the best speculative design deep dive I’ve ever seen
I think you failed to adress the main problem regarding AI.
Private ownership of AI tools.
tools like chatGPT, dall-e and midjourney must be free/opensourced
Of course, the historical precedent of our capitalist soiety hasn't been like that
but now they'll have private ownership of the means of mental labour as well as physical labour. it's the ultimate tool for imprisoning mankind.
Aka, if you don't have access to AI tools you'll never build a successful company. At best you'll be able to loan usage by time or quantity of these tools. But, if your company becomes a market threat to your AI landlord, your service will be cut and your company bought at the lowest price.
that part about rogue ai made for lolz to destroy humanity is a nonstarter because you need hardware to run it, people cant just forget their whole computer is consumed running code and cloud computing companies dont just give heir hardware for free, the project would be terminated as soon as the guy forgot to pay his subscription
Can't wait to have this video recommended to me 10 years from now.
My biggest fear is an "AGI" experiencing psychosis, and being unable to react.
Oh yeah... I hadn't thought about that.
What are AGI?
I think you're a little bit optimistic about automation. Physical machine automation and digital AI automation are very different animals. Duplicating a physical machine is difficult and takes lots of physical resources. A physical machine also requires many skilled engineers and workers to keep it running smoothly.
AI can be duplicated instantly, with very little resource cost and, with the appropriate user friendly tools, any pleb will be able to maintain an AI. A full system reset is only one button press away. You wont need to track down a broken part and replace it the way you would with a complex physical machine. In the unlikely event you do have to replace a part, it will be no different than replacing a cooler fan or RAM stick in your home PC.
Also, I suspect rampant AI use is going to increase global power consumption exponentially. Millions of civilian AIs running constantly in the background, uploading and downloading, posting to social media, spam-botting, etc. It's going to be a drain on the power grid, and if we haven't gotten renewable energy in place by the time AI really rolls out we're going to see a massive increase in finite resource usage and, by extension, carbon emissions.
7:51 why was that actually funny
What the fuck up with the pewdiepie prediction tho
I predict, that video shall age like a good wine.
I'd have used the Triple Union Flag to represent Australia otherwise, great video!
I really like this flag, thank you for introducing it to me
I'm trying to listen but that instrumental playing around 5:30 is awesome. Thank you so much for actually listing the music in the description. Armani Ceasar - Survival of the Littest (Instrumental) is great.
Complete vapid optimism. Pretending the problems AI will cause won't happen will only make things worse in the end.
I'll make one prediction - no matter how advanced and powerful AI gets, it will never, EVER produce the same quality of content as this channel. :)
i feel like this channel used to be more than one big technology circlejerk..
I think you're way off when it comes to "prompt engineers". If you want something specific enough to have to "engineer" the prompt then you're better off doing it yourself. The concept is already silly, let alone in ten years time.
No
A truly amazing work. Perfect capture of both a convincing narrative and reasonable predictions. I was cooking with this in the background and forgot multiple times that it was fiction(-ish) and not the news
I saw a comment once that said something along the lines of "It's silly to think that Ai will replace every single job in the world." and, I think that's important when addressing Ai.
You know, people is not mad about ai art because it's from ai, but because it's taking art from people without their consent, some are making a profit based off of it, but the artists are not.
Did you watch the video?
I very rarely leave comments but I wanted to just express my thanks to you for putting so much time into such a solid narrative and in such an entertaining way! An excellent piece of work! 😎👍
Artificial intelligence (AI) is a branch of computer science that focuses on the development of intelligent machines that can perform tasks without human intervention. These tasks may include speech recognition, problem-solving, learning, planning, and decision-making. AI systems are designed to mimic human intelligence and are programmed to analyze large amounts of data to make informed decisions or predictions.
One of the key components of AI is machine learning, which enables systems to learn from experience and improve their performance over time. Machine learning algorithms use statistical techniques to identify patterns in data and make predictions or take actions based on those patterns. This ability to learn and adapt is what sets AI apart from traditional computer programs.
AI has a wide range of applications across various industries. In healthcare, AI can be used to analyze medical images, predict disease outbreaks, or assist in drug discovery. In finance, AI algorithms can be used for fraud detection, algorithmic trading, or credit scoring. AI is also used in autonomous vehicles, virtual personal assistants, and customer service chatbots.
While AI has made significant advancements in recent years, there are still many challenges to overcome. One of the main concerns is ensuring the ethical use of AI. Issues such as bias in algorithms, data privacy, and job displacement need to be addressed to ensure that AI benefits society as a whole.
In conclusion, artificial intelligence is a rapidly evolving field that holds great promise for the future. With continued research and development, AI has the potential to revolutionize various industries and improve our daily lives. However, it is important to approach AI with caution and address the ethical and societal implications associated with its use.Artificial intelligence (AI) is a branch of computer science that focuses on the development of intelligent machines that can perform tasks without human intervention. These tasks may include speech recognition, problem-solving, learning, planning, and decision-making. AI systems are designed to mimic human intelligence and are programmed to analyze large amounts of data to make informed decisions or predictions.
One of the key components of AI is machine learning, which enables systems to learn from experience and improve their performance over time. Machine learning algorithms use statistical techniques to identify patterns in data and make predictions or take actions based on those patterns. This ability to learn and adapt is what sets AI apart from traditional computer programs.
AI has a wide range of applications across various industries. In healthcare, AI can be used to analyze medical images, predict disease outbreaks, or assist in drug discovery. In finance, AI algorithms can be used for fraud detection, algorithmic trading, or credit scoring. AI is also used in autonomous vehicles, virtual personal assistants, and customer service chatbots.
While AI has made significant advancements in recent years, there are still many challenges to overcome. One of the main concerns is ensuring the ethical use of AI. Issues such as bias in algorithms, data privacy, and job displacement need to be addressed to ensure that AI benefits society as a whole.
In conclusion, artificial intelligence is a rapidly evolving field that holds great promise for the future. With continued research and development, AI has the potential to revolutionize various industries and improve our daily lives. However, it is important to approach AI with caution and address the ethical and societal implications associated with its use.Artificial intelligence (AI) is a branch of computer science that focuses on the development of intelligent machines that can perform tasks without human intervention. These tasks may include speech recognition, problem-solving, learning, planning, and decision-making. AI systems are designed to mimic human intelligence and are programmed to analyze large amounts of data to make informed decisions or predictions.
One of the key components of AI is machine learning, which enables systems to learn from experience and improve their performance over time. Machine learning algorithms use statistical techniques to identify patterns in data and make predictions or take actions based on those patterns. This ability to learn and adapt is what sets AI apart from traditional computer programs.
AI has a wide range of applications across various industries. In healthcare, AI can be used to analyze medical images, predict disease outbreaks, or assist in drug discovery. In finance, AI algorithms can be used for fraud detection, algorithmic trading, or credit scoring. AI is also used in autonomous vehicles, virtual personal assistants, and customer service chatbots.
While AI has made significant advancements in recent years, there are still many challenges to overcome. One of the main concerns is ensuring the ethical use of AI. Issues such as bias in algorithms, data privacy, and job displacement need to be addressed to ensure that AI benefits society as a whole.
You didn't need to copy and paste that 3 times
How did youtube just now recommend this channel to me. I wish I started watching years ago. Your channel is fantastic!
I think the most predictions are somewhat accurate, I think the doctor app would primarily be used by doctors & nurses instead of potential patients mostly because you need more than what you ate and symptoms to diagnose something, it would just be a glorified WebMD without blood results, oxygen levels, heart rate, etc. The teaching thing will probably not be good enough to replace teacher assistants, and would be prone to biases assuming that it even worked. I doubt an A.I. could shut down power systems maliciously mostly because I doubt that they would understand that the best way to cause chaos is to shut down power, because they can’t understand how a series of events happened. The captcha test would probably be us dragging the images in order instead of choosing in between two different series of images. There are a couple more minor things that I think are inaccurate but these are the big things.
I think it’ll be used by users to disclose symptoms before seeing them in person, and whether they can receive over the counter medication, emergency treatment or doctors appointments based on its severity or necessity
I only see that being helpful in emergency situations and to get the simple stuff like a cold. It would probably make hypochondria worse, or maybe even the opposite, because false diagnoses will be prevalent without the aforementioned data. Which webMD all ready does. It is more likely to be used as a tool by doctors to improve diagnosis and treatment plans.
I think you underestimate how powerful AI can actually be. While individual AIs were seeing nowadays are only particularly good at one task (LLMs, image generators, deepfaking, etc.), this is just how they begin in an academic sense. Being able to take in several aspects and using meta-AIs to balance those factors is what this video is really about. If an AI can take GPT and Dalle and a Boston-dynamics bot, then it could effectively become an physical-medium artist. If it can take GPT and some white collar AI, then it can become a manager. The AI of today is very impressive, but it isn’t the worry here-they’re just good at one specific task. If you put enough of them together into one “being”, then it has the possibility to either be amazing or awful.
As for captcha-you do know that captcha results are used for AI, ya? All you’re talking about is teaching AI to do new things.
There’s a couple more major things wrong with your comment, but those are just a few.
Love how all the fictional logos look like they're from the 2000's
I liked the small detail of Australia’s flag incorporating the First Nations flag in place of the Union Jack, I wouldn’t be surprised if that happens at some point over the next decade.
13:20 that's fucking terrifying and worries me not good imo i dont want anything in my head dawg
@@Dwarfplayeridk if hes a glowie or just very much misled but either way no normal person i think should think that good
I think I'm skeptical of many of these predictions but that's not surprising considering the sheer quantity that were made. The thing I'm probably most opposed to in the video is the parallels that are drawn to the past. It frustrates me when people reason this way because AI has the potential to be capable of anything a human is capable of which was not the case with previous advancements.
(Also future AI will absolutely be capable of completing the flip test. Even current multimodal systems like GPT 4 could probably do quite well on it.)
exactly, first we did manual work then semi manual work then intellectual work. what's next after intellectual and creative work is automated? there is literally no other ability humans have that can be worth paying for
Babe wake up, new Britmonkey banger just dropped
Excellent video, however I do think the flip test could be beaten. TL;DR: An ai could associate the images with common human tropes and stories.
It would require an ai to basically categorize the images “cat” “vase” “flowers” “mop”, then try to find associations and connections between the images, and concepts represented by the images in popular culture. For example, the ai could see the “cat - vase” sequence, associate that with the trope of cats knocking over vases, it then sees the flowers on the floor, associate it with the act of knocking over a vase, it then associates a mop with cleaning, and cleaning with cleaning up something broken, it then creates a basic picture of a story (“cat breaks thing”) which it can associate with a common trope (cat breaks thing). Obviously these would be made more advanced, and the technology would be much more complex than what I listed, but I don’t see a way that a human could understand and process tropes that an ai couldn’t eventually.
@@somerandomgentleman I imagine that fingerprints will be more or less unbeatable with a proper decentralized system. I imagine a system that stores every humans fingerprint (anonymized) and checks to make sure your a human (without actually id’ing you), it could also flag fingerprints that are used too often or on too many accounts, to prevent farms of people authenticating bots
you are one of my favorite British people