ฝัง
- เผยแพร่เมื่อ 18 พ.ค. 2024
- Explore the current limitations of artificial intelligence and the possibility of creating human-compatible technology.
--
In the coming years, artificial intelligence is probably going to change your life- and likely the entire world. But people have a hard time agreeing on exactly how AI will affect our society. Can we build AI systems that help us fix the world? Or are we doomed to a robotic takeover? Explore the limitations of artificial intelligence and the possibility of creating human-compatible technology.
This video is based on interview excerpts from the Radio Davos Podcast. The episode is called, “The promises and perils of AI - Stuart Russell on Radio Davos”.
Listen to the full episode here: www.weforum.org/agenda/2022/0...
Directed by Christoph Sarow, AIM Creative Studios.
This video made possible in collaboration with World Economic Forum
Learn more about how TED-Ed partnerships work: bit.ly/TEDEdPartners
Support Our Non-Profit Mission
----------------------------------------------
Support us on Patreon: bit.ly/TEDEdPatreon
Check out our merch: bit.ly/TEDEDShop
----------------------------------------------
Connect With Us
----------------------------------------------
Sign up for our newsletter: bit.ly/TEDEdNewsletter
Follow us on Facebook: bit.ly/TEDEdFacebook
Find us on Twitter: bit.ly/TEDEdTwitter
Peep us on Instagram: bit.ly/TEDEdInstagram
----------------------------------------------
Keep Learning
----------------------------------------------
View full lesson: ed.ted.com/lessons/things-are...
Dig deeper with additional resources: ed.ted.com/lessons/things-are...
Animator's website: aimcreativestudios.com
Music: / aim-music
----------------------------------------------
Thank you so much to our patrons for your support! Without you this video would not be possible! Mark Morris, Misaki Sato, EdoKun, SookKwan Loong, Bev Millar, Lex Azevedo, Michael Aquilina, Jason A Saslow, Yansong Li, Cristóbal Moenne, Dawn Jordan, Prasanth Mathialagan, Samuel Doerle, David Rosario, Dominik Kugelmann - they-them, Siamak Hajizadeh, Ryohky Araya, Mayank Kaul, Christophe Dessalles, Heather Slater, Sandra Tersluisen, Zhexi Shan, Bárbara Nazaré, Andrea Feliz, Victor E Karhel, Sydney Evans, Latora, Noel Situ, emily lam, Sid, Niccolò Frassetto, Mana, I'm here because of Knowledge Fight Facebook group., Linda Freedman, Edgardo Cuellar, Jaspar Carmichael-Jack, Michael Burton, VIVIANA A GARCIA BESNE, The Vernon's, Olha Bahatiuk, Jesús Bíquez Talayero, Chels Raknrl, Sai Pranavi Jonnalagadda, Stuart Rice, Jing Chen, Vector-Dopamine math, Jasper Song, Giorgio Bugnatelli, Chardon, Eddy Trochez and OnlineBookClub.org.
Huge props for making the robot both terrifying and, paradoxically, incredibly adorable!
he looks like Atlas; all we need is Peabody to make the Portals 2 reality a thing XD
He's just doing his best
it all depends on how you use it
Yup. The video's animation is a story in and of itself.
yeah but if we don't do that, we're gonna see someone do decades of warehousing and call that life.
I love how Ted Ed comes with a new animation style for each of its videos.
the chicken one is best. no doubt
Yea it's cool they commission new animators
@@moththe_player6905 LINK??
I especially loved this one.
One of my favorite along with the 8-bit from the video about Achilles.
It?
As Stephen Hawking said : "If machines produce everything we need, the outcome will depend on how things are distributed.Everyone can enjoy a life of luxurious leisure if the machine-produced wealth is shared, or most people can end up miserably poor if the machine-owners successfully lobby against wealth redistribution. So far, the trend seems to be toward the second option, with technology driving ever-increasing inequality."
You think the government wouldn't consider it a security risk to let rogue oligarchs control the entirety of people's lives? The government will just be the ones to control the entirety of people's lives and no one will be able to protest even as they ride around town on a high bike made of human ligaments with the human still attached and screaming or turning people inside out to make them walking vaginas as their organs sputter blood.
This is why we need wealth redistribution laws to help society thrive.
@@captainahab5522 Just let capitalism rot.
Not saying I'm smarter than Hawking or anything but wealth is relative. As long as you're not on the street you're far wealthier than the wealthiest of a hundred years ago.
You‘re definitely not, compared to any king or similar rich people, you have basically nothing. Just because the amount of currency they had could afford way more than it would today. Could you build a castle and employ the entire staff for a private guard?
"I don't want a nation of thinkers. I want a nation of workers." -John D. Rockefeller, 1903
First wanting people to be robots.. now wanting robots to be more like people. But still not wanting either one to think too much.
This certainly is a curious simulation we're in.
its either humanity goes forward or we face extinction when this planet comes out of its current stable climatic conditions which allowed our civilization to exist in the first place.
Just gotta hope the rulers (in WEF) have better goals this time.
@@jeseAudio mhm
"You'll own nothing. And you'll be happy" - World Economic Forum director
Simulation theory is made up, its not real.
Wall•e is such a good film! I remember it being my favorite while I was growing up. But only later I realized how deep the significance of what it portrays runs. It's kinda scary.
i saw it when i was still in college. it already hit me what it was portraying about our humanity and the dangers of relying on AI to help us.
I have forgotten it....
i didnt really think too much of wall e when i first watched it, i just liked it but now i'm more happy about the movie's meaning. i like seeing people bring light to environmental issues to other people who aren't as aware.
was thinking about rewatching it recently, actually. i've found an interest in sci-fi yk stellar sorta stuff so :p
I saw it when I still in elementary school and I was crying because the relationship is too adorable, later I know the movie is not just about friendship between robots and the meaning is really deep ..
Love the graphic work, wonder who made these! Would love to hire them hah
Auto wasn't the villain in story Wall-E, he was just carrying his objective.
yeah, the "directive"
That's how all AI will be, even those that destroy humanity. That's the point, lol.
that's the point
@@clusterstage interesting
The villains are those who believe they hold the true path over all other interpretations.
This reminds me of the Hitchhiker's Guide to the Galaxy (published in 1979). Arthur asked the ships computer for a cup of tea that actually tastes like tea. The computer was so perplexed it turned off all other systems just to make the tea. Killing the crew or crashing the ship wasn't even a consideration.
LOL i'm sure programmers have already accounted for all of that by feeding it all scenarios of what not to do and why, even with us humans if you don't explain to someone why not to do something they want to do they may just go and do it anyway lol.
It's kinda fun how the General AI problem is basically a reformulation of the Genie in a Bottle dilemma.
I'm a genie in a bottle baby, come, come, come in and let me out
@@aaronvogler2864 tf
@@bruther7892 It's a song
Except the genie only understands commands in a long extinct dialect of ancient Croatian.
@@donaldhobson8873 Maybe that’s the key to unlock AI’s true potential…
First, let me say I'm not anti-tech (I'm a retired engineer), but I admit, I'm a little apprehensive about the wide spread use of AI and how there may be unintended consequences. Fortunately being 72 years old, I probably won't be around if 'the singularity' (in the 'out of control tech' sense of the word) occurs as a result.
It's good to have a general sense of caution so you can prevent situations that are harmful
darn it im 20 but take me with you please, i also dont wanna live in such world
It won't occur any singularity. People will only need to reskill to deal with new entities invisibly populating the world.
I am kind of worried what will happen to us if we don't get wide spread use of AI to help us. People being in charge of stuff is really dangerous too. For every scenario in which AI kills everyone, there are also ones in which humans kill everyone too.
@@awa418
The quality of animation on this channel is incredible.
AGI will be here this decade. The idea that it's even 20 years away is insane! People aren't taking into account that we have had incredible advancements in AI almost every week for the last year! And it's only accelerating. We're at the very start of an exponential curve. Watch this space!
After a deepdive into this very fascinating subject, I suspect we'll have something that a lot of experts in the field describe as early AGI by the end of 2024 at the very latest. GPT-4 is already showing a number of emergent properties that push it well beyond the bounds of what is typically considered an AI. The recent plugin update for GPT-4 has shown that the system can engage in effective tool use to solve novel problems, which (to me) indicates a generalised system. I believe that with the addition of robust self-reflection and long-term memory capabilities, even a system like GPT-4 may be considered an AGI.
Or, the problem is harder than we think and we just reached , so for , for the low hanging fruits.
As with most complex problems we don't know how complex it is when we start. Just take DNA for example. The researchers that were first attempting to sequence the entire genome constantly bragged that "oh yeah by the time we get it all typed out and know the entire sequence we'll know what all of it does within ten years or less" They truly believed that sequencing the entire thing was going to be the hardest part.
We got our first fully sequenced human genome in 1995. Now 20 years later we still don't know what 75% of it does. Of the 25% we have figured out what its coding for we only sorta understand how or why it does what it does. Epigenetics wasn't even a concept before and now we realize its this whole other thing that we need to figure out before the sequence really means anything. We didn't know what we didn't know.
With AI I think this problem is going to be harder because so much AI is being written by AI. Its doing more and more impressive stuff and we're actually understanding less and less of HOW its doing it. A great example of this was when they taught a computer how to play GO. It could be the greatest GO players in the world but then got tricked by people that barely understood how to play the game. It turns out it got really good at playing at a professional level and winning but the program didn't actually understand the game. It didn't have a strategy. Why? Because they didn't program it how to play. They just programmed the rules of the game and then had it play millions of games against lots of different people and remember what lost and what won.
I personally believe that it will not take as long as we might initially think. The AI we have and the regular computing power along with the new quantum computers will help as tools and will enhance our endeavors. However I feel we are hurtling towards a very scary time indeed.
Yeah me too, I've been trying to keep up with all the advancements, I don't quite understand how the smartest people don't see this coming, each new advancement only drives advancement so much faster, its not linear with this stuff, I think 2030 to 2040 something BIG will happen, and I think the top of the top know this
@@tev17 I do not think we even have that long. Thinking 2025.
If 1000 big thinkers are asking for regulation now, then they probably feel some dangers right around the corner.
I don't know how quantum computers will help. I mean they can totally maybe but quantum computers just use qubits instead of bits, they're not necessarily supercomputers just because they're quantum, but its a good thought experiment.
@@mifarland They are still a new tech and not mastered. It is true they work differently. But linked with AI and developed further and a quantum computer could for instance calculate multiple situations simultaneously. No que or line of computation. That is the advantage. It a different way to compute data that is super fast but not the sane as a regular computer.
I think this video raises very valud issues with AI. AI can defiantly be really cool and help us advance as a civilization, but if we overuse it or use it inadequately it can bring our ruin as well. Maybe we should just use it where it's obviously needed and useful, rather than a cheapskate trick to replace art, fire workers or do things in general we're either too lazy or to cheap to do ourselves.
But that’s too restrictive, isn’t it? All the _important_ work that needs to be done comes at the great cost of millions of years of human life, every year. It’s a worthy endeavor to save every laborer from a life of labor. The fear of disaster will need to be reconciled with just how much we have to gain.
Utopia comes after just about all the workers are made obsolete, or in other words, fired.
@@SeventhSolar You better give them the means to live a decent life without their jobs then. Shouldn't be impossible with every function replaced by AI, but god you never know with these people...
do you think craftsmen of the old like smiths or shoemakers didn`t think about their work as art? But they got replaced by more efficient machines at factories. The same is happening right now to artists and one hundred years from now nobody would care about it.
Here's how I like to put it: AI and robots should serve humanity, while humans should help themselves. What falls into the categories of "serve" and "help" is a bit arbitrary, but the guideline still is helpful.
But why use human artists when it can do it better and exactly how we want
The animation in these videos are always great but this animation has to be one of, if not my favorite. Wonderful job as always!
would be funny if AI made the animations warning about the use of AI
I found myself really enjoying it... like getting distracted for a good moment because I was admiring the atmosphere they were able to create near the end. Whoever animated this one is a wonderful artist!
I like how most comment in admiration for the graphics and design of this video. Where the whole point is to illustrate the actual risks of the advancement of technology.
Humans are distracted by Aesthetic & Beauty.
Hats off to the graphics and animation team.😍
generated by Dall-e 2
th-cam.com/video/qydPrkuz7Ag/w-d-xo.html
Machine: What is my purpose?
Rick Sanchez: You pass butter
Machine: Oh my God!
step 1: clear humans
step 2: conquer the universe
step 3: create optimized butter consumers, as many as possible
step 4: pass the butter 😃
@@alainlenoach754 Whoosh
@@jessicajayes8326 It adds to the joke..
Summary:
1. The certainty of objective completion by an AI can make it psychopathic. It doesn't know what we care about, aside from the exact objective.
2. Effect of AI? e.g. Technological unemployment(Aristotle and Keynes)
3. What if we forget how to run our civilisation, if machines run our civilisation, and we become infantalised(similar to Wall E)? This chain of training until now has continued for 10s of thousands of Years.
4. When will general purpose AI arrive?
Things not mentioned: rule based systems are unviable because of this psychotic certainty problem and that is one of the reasons for the focus on probabilistic reasoning via machine learning / deep learning etc. That said ai could still have “civilization ending” effects.
Here is an AI-generated summary of the video:
00:00:00 - 00:05:00
Computer science professor and AI expert Stuart Russell explains in this video how AI will change the world by becoming able to do most things that humans can do, causing mass unemployment in the process.
00:00:00 This 1-paragraph summary of an interview with computer science professor and AI expert Stuart Russell explains the difference between asking a human to do something and giving an AI system a fixed objective. He says that when humans do something, they often know what they don't know and that this allows them to be more creative in their actions. He also says that although AI systems won't be able to do everything that a human can, they will eventually be able to do most things that a human can do. When general purpose AI arrives, it will cause mass unemployment due to the automation of many jobs.
00:05:00 This video discusses the potential impacts of artificial intelligence on the world, with the median estimate being that general purpose AI will be developed by the end of the century. While the impacts of AI are still largely unknown, it is clear that the technology will have a significant impact on many aspects of life.
@@rxflexedits3310 What did you use to generate it?
When will general purpose AI arrive ? literally now. I now use ChatGPT daily to generate code for my job. It's not perfect yet but it's really really good.
Good job, thanks for this summary! Summaries help knowledge stick.
This video had excellent animation. It coupled with the narration so well by mixing abstract art with litteral representations seamlessly.
One of the most realistic and easy to understand explanations of the dangers of AI I’ve seen to date. And one that seems very realistic and not overdramatic or exaggerative. Thank you Mr. Ted!
Super well done to the team of animators! I'm floored at the illustrations and lighting effects you guys pulled off. Absolutely inspirational.
Too bad the audio was muddy, so hard to hear the content and actually take it in.
i dont think that was the point dude
"It will be the first time in human history when humans will not be inclined to understand everything and pass that information onwards."
Bro like 90% of the global population doesn't understand how the internet works and doesn't have any inclination too understand, yet we all still keep on living and thriving.
I think he means in a more global sense. Sure very few understand how the internet works. But the results of it, people do understand. It's essentially a modern day library, town square, store, etc. All of which would collapse if people weren't taught language skills, socialization skills, basic math, creativity, etc. The internet is icing on a very complex society of cake. This ai could take over aspect of our society and humans will be at a disadvantage if we don't understand what ai will be doing in the future. It might be able to exist without us, which is not something the internet or any previous technology can do.
tbh I think it’s more like 99.99% of the world population doesn’t, if not more.
We won''t have anything else pressing and learning can be fun.
With radical life extension, we might have the original AI programmers around for a very long time. No need to pass info on if the previous generation is still there.
It might be trivially easy to enhance intelligence and download all the info directly into your mind.
Animation and music were so cool that I want a version with the interview audio just for contemplation.
The animation is beautiful and it really helps me to understand the concepts that are being talked about
It sucks balls.
how the audio and the background animation story perfectly matches? that's so great!!! perfect!! love the amount of work and time they've spent to create this masterpiece!
i'm flabbergasted yk :)
if every interview was done like this i might start watching more of them, i love Ted Ed ❤
This is only 1 year old and it already feels out of date compared to how fast AI is progressing
yup! a year ago many were still scoffing at the idea of AGI before 2030
Now Devid can code and create more software programs to replace old software programs. We are closer to real life Skynet from Terminator than ever.
The animation is such work of art! Truly amazing style and depiction of the content.
its mad how far its come in the past 5 years
COVID has accelerated it
The animation is a master peice. So cleverly desinged.
Me: AI, I want you to paint my house.
AI: (Rips house out of ground and dips it in a tin of paint)
Me: Ummm .. not quite what I wanted.
This talk was prior to ChatGPT, or right around day 1. If we revisit Prof. Russell and ask him to have this discussion again, I think his conservative guess of having general purpose AI by 2045, will change to a conservative guess of 2030, or even sooner....as of today Emad Mostaque says it will be here within just a few years. The most recent prediction about the advancement of AI was that it would reach some crazy level of filling prompts within 6 months to a year, but it happened within 45 days, and the person who made that prediction said he couldn't believe what just happened, and that it made him nervous for the future, being AI moving faster than we humans can even digest. I personally predict that one day soon, before 2030, we're going to wake up and the whole world is going to be completely different than it was the day before...and we're not going to know what to do...a bit chaotic...we definitely need to be careful
“Aimless extension of knowledge, however, which is what I think you really mean by the term curiosity, is merely inefficiency. I am designed to avoid inefficiency.”
― Some Robot from The Caves of Steel
I wonder why it would think that? I can't imagine how a robot would not like expanding its knowledge
@@aguyontheinternet8436 Well the first things that come to mind is limited hardware and useless information. You don't need to know the sound of a kitten meowing or anything about cat to build a spaceship, a car or a computer. It will just take "space" in your memory for nothing. We, as human, just like cat because of, well, feelings. Can't really use all the "space" in our hardware(brain) either. But an A.I./machine ? It could very easily bump into problem be following that same logic.
@@xaviermantha63 But if you want to do more than build a single thing? What if you want to solve any problem that you see? What if you want to make new problems, solve those problems, make harder problems, and take over the universe.
You would want to know the sound of a kitten meowing, perhaps to solve a stray cat epidemic, or to understand why humans are so responsive to it.
@@aguyontheinternet8436 Well, yeah, but now you're basically telling the machine that "your objective is to know everything" so of course it'll consider everything as relevant. But even then knowing more about basic physics and the like will important enough to "forget" information about cats because one help having more memory to hold more information and the other does not.
@@xaviermantha63 well the wonderful thing about robots is they can just get more memory. If they want to make space, they can put all their cat knowledge in a hard drive labeled cat
That's a fascinating point! Our crew filmed AI from another perspective, interspecies communication. Can AI helps us decode whale communication? And would it make sense? We tend to humanize animal species and have high hopes that their communication system is similar to ours. Still, AI could be a game changer in this case. Let's see what the future brings!
That's actually cool lol ill check your videos out
It probably could yup but you have to feed it different whale noises and locations you typically hear the whale communication, that would be a nice simple task for it. It's problem is it can't handle requests that are even a bit more complex it gets confused and the more complex the more confused it gets the more it can't answer it and it's answers will get more and more non-specific, because it's only the product of what it knows and what information it can gather based on the objective you set it and the thing you give it to assess.
Seeing this video in the context of the last month since the release of chatGPT certainly shows the incredible 90% turn that has happened in the world of AI
Abso-effing-lutely stunning animation, sounds and the choice of colours. It's breathtaking. The style would be perfect to illustrate a dark sci-fi story like "I have no mouth and I must scream". Maybe even "1984".
I once had a virtual lecture about AI in medicine. The presenter said the simplest way to understand what AI can do is efficiency. Whatever we want to be efficient, AI can do it to its optimum potential. Hence, at the core, AI cannot solve moral problems because they are strictly human. Therefore, we need to ask questions about what we really need, and if it is something AI can do, we will develop the AI. Or else, better just let human beings fix our own things.
Great video, as always. Peace.
Part of the issue is that, as we shape technology, technology shapes us. Thus, what we really need, will always, at least in part, be determined by what we can do (as long as you don't go down the completely nihilistic path). Since all of this is always a process and never ceases to be a process---of which we ourselves are an integral, moving part---reconsiderations are always met with accomplished facts. The _one point in time_ where we stop and think, "is this really how we want things to go", is either right now, or never. And since we're also lazy, it certainly isn't right now.
That is only true as long as the AI is a tool. Once an AI system can actually understand different concepts and be capable of reasoning and some form of consciousness, it will reach an intelligence explosion. That will be the most significant event in the history of this planet and it will dramatically change the world in an irreversible way.
AI could very likely solve moral problems if there are answers to those problems. The issue is people as a whole dont know the answers to the problems, so how could we ever determine if an AI is right if we cant even agree on whats right
@@NathanSMS26 that and people are illogical they don't always know what they want and there are those who are always against the status quo
@@NathanSMS26
"AI could very likely solve moral problems if there are answers to those problems."
If there is none because moral is a inertly human concept, the answer to solve moral problems could be: kill all humans.
a great video - but really the artist who drew these animations is quite fantastic! He/she/they deserve most of the praise here!
About the coffee scenario at the beginninng, this is ChatGPT's answer:
As an AI language model, I don't have a physical form and I don't have the ability to perform physical tasks like getting a cup of coffee. However, if I were a physical robot programmed to get coffee, I would follow my programming and attempt to retrieve the coffee from another store or find an alternative solution for you. If I were unable to do so, I would inform you of the situation and discuss possible alternatives with you.
The animation and art style is gorgeous! Congrats.
The impending technological singularity with unknown and unforseen consequences is cute and all. But, the animation here is just incredible! Can't wait for video generation AI to get to the point where I can propmt using this template for longer form
Hopefully when that day comes when you can type prompts to get this animation, there will be ai that can type the prompts, and there will be ai viewers to enjoy the fruits of ai artists laborers.
@@jmhorange there's already AI writing better prompts than humans ☠️
It's worrying that the only thing you care about is automating people's jobs while we live in a capitalist society!
@@samuelkibunda6960 It's because they have the idea that this won't affect them, just other people. That they'll somehow work alongside it instead of realizing they'll be replaced too. If efficiency is important above all else in life, then humans working alongside ai is very inefficient.
@@jmhorange yeah I've seen this a lot with these "Prompt engineers" saying that they are the future companies like midjourney, stability AI, and open Ai are all trying to remove the need of humans in any field by automating anything and the best way to do it is automate the people working with Ai (which is already happening with chat gpt)
“As soon as we started thinking for you it really became our civilization” (agent Smith)
Funny enough, I just watched this during a study break while being tutored in Computer Security by Open AI's new chatbot. It's insanely impressive, able to give me detailed explanations about everything I ask about, then generate practice questions and answers, then explain those answers in further detail on request
It's funny how the World Economic Forum brings this topic for debate, however, the MIT World3 System Dynamics Model has been up and running for almost two decades. Unless people decide to speak up, things will continue to run autonomously as they have been.
It seems to me, some aspects of this video are already obsolete with GPT4
We need an expanded version of this video
Depending on utility function or moderated topic - when different powerful language models made by different companies/developers are put together in a chatroom, what does their conversation look like?
Time and again I've always said that these kind of videos fascinate me and I wonder how they are made. As far as the topic is concerned, yes AI is changing everything and projects like pulse network are proof that it is going to improve any industry.
Thank you Professor Stuart Russell for letting us know simply, yet clearly on what is missing in the AI that is currently being built.
Thanks TED Ed for your nearly daily dose of information! 😄
Professor Russell brings to light deep questions about the nature of artificial intelligence and its social and economic ramifications. One idea he highlights is the importance of designing AI systems that can deal with uncertainty about their objectives while avoiding extreme or undesirable behaviors. This raises the intriguing question of how we can develop AI systems that can learn and adapt their goals more flexibly, rather than operating with rigid and potentially dangerous goals.
Furthermore, the discussion about the impact of AI on the economy raises important concerns about technological unemployment and the loss of human skills. However, it is also crucial to consider how AI can create new employment opportunities and how humans can collaborate more effectively with AI systems to increase productivity and innovation.
Finally, the debate over the timeline for the development of general-purpose AI highlights the complexity of the challenge. While some experts predict its arrival within a few decades, others adopt a more cautious view, recognizing the technical and ethical difficulties that need to be overcome. This uncertainty reminds us of the importance of a careful and reflective approach to the development of AI, considering not only the potential for technological advances but also their social and ethical implications.
And this is where me and my crew become the underground warriors
I say that AI should be either left untouched, or simply applied as a brainstormer that comes up with ideas but doesn't actually implement them.
This would allow us to sidestep many of the problems of AI, because then it won't actually *do* anything. It'll just help us figure out a variety of solutions. Perhaps after we've sufficiently advanced that sort of AI and gotten a hang of it, we could start dipping our toes into a more robust AI system.
(Another one of the problems would be that the greater an AI's agency, especially if it should develop some form of emotions, the more important it'll be to give it ethics. And humans are pretty bad at ethics when we look at our track record.)
i think this is really well put and that you're onto something
Nope, the technology will only get better and better, regardless of regulations.
Some day, somewhere, someone will create the AI that will be a literal god compared to humans. And that godly intelligence can multiply itself infinitely and make infinitely better versions of itself. Even if that is stopped somehow, it will only be a matter of time when everyone in the world knows how to create these god ai:s.
NO ONE can predict what happens after this point, the things that unfold when quadrillion copies of an intelligence that are quadrillion times smarter than einstein appear
Ong
This is the obvious answer, it's baffling that so many people just jump to giving AI the ability to act and think there's no solution. The solution is to not give it the ability to act...
@ProjectMoff Is anyone actually doing tests like this, though?
I honestly wish we'd leave this box unopened, nice as some of the possibilities would be. Bu this method at least gives us more control, and it seems we aren't using it at all.
The animation is indeed a masterpiece. The audio everything is sinking together. I love it 😍
I thank you deeply for telling the public more about artificial intelligence. I was unaware the status around the topic was harsh and I hope to see improvement on both the human and artificial intelligence sides.
Thank you for always uploading really good ideas, information, and New things for students like me
These things are very useful for me to be able to learn more about the things that I didn't get in school
The animation is so cool. I loved the way you explained. I am on a camp that believes that having an advance AI would automatically make us more intelligent too. If you want to develop sophisticated tools then you need to learn how these tools are built and used. This will help the user increase their cognitive power.
I disagree, an advanced AI will come up with a long line of possible suggestions that would help your development of a tool and outline each pro and con. It is a thinking machine, it runs the hypotheticals, much like a chess machine runs all the moves. Once an AI is able to catch your general intention of what you are looking for, the details will be suggested, you have to barely give a deep thought.
I believe we ourselves will not become more intelligent, but if you consider AI a part of use you could say we are more intelligent in conjunction.
How do you view this advanced AI? I'm curious why you believe it will improve our cognitive ability, maybe I'm missing something. To me it's like having a business partner doing all the hard work for you all the time.
@@moon_bandage We don't to look further. Take internet as an example. I would argue that internet has made people clever and helped in the acceleration of technology. Similarly, AI systems will help us understand more about out universe and since humans are pattern recognition machines, we will embark in more scientific wonders.
The animation style complements the topic really well; terrifying and understandable.
Respect to this guy for dispelling my fear of AI.
The best sentence was that current AI does not know that it does not kkow.
Current AI isn't yet dangerous. People are worried about the future AI. You know, stuff that hasn't happened yet.
@@donaldhobson8873 fear of the unknown.
That's a good enough answer. Also the animation is sooo well made 😍😍
Absolutely incredible animation, kudos to everyone involved!!
the animation on this is absolutely incredible!!
Here after having my mind blown by ChatGPT's capability. On one hand I'm excited what future holds for the field of AI, on the other I'm scared of the potential of immutable errs. I predict a lot of talk on ethics in the coming decades.
I guess it's only going to push people to actually teach people things better lol and the lack of realistic and critical thinking they need to start stepping up and actually teaching it because we're in the mess we're in, in our world because of not possessing those two things. It's now absolutely essential people can assess things well for themselves which requires allowing people to think for themselves.
Realistically AI is just a computer algorithm that will always make mistakes no matter what you teach it just like with us humans but it will no doubt do it all the time because it has an inability to realistically think about things and properly critically think, and so it must be treated as just that. It doesn't at all have situational awareness.
People have already been trying to use it for more complex tasks too like writing a story but it can't lol and ends up stealing bits and pieces of other peoples work because it only knows what you gave it and let it know and assess, and then gives thousands of people the same story because it's using the same one based on the same thing other people ask it to write about lol, then you have to constantly keep getting specific with it on literally everything and you''ll be there all day asking it questions or it will just shut down because it's too confused lol.
Which is why they limited Bing AI to 200 things a day or in 1 session lol so you cannot keep asking it things, so it's honestly best used for more simple specific questions and requests, so that in itself certainly stops people from being far too reliant on AI software to do everything for them because it can't do more complex things well.
i came here looking for comfort and ended up with more anxiety
The robot design reminds me of Portal 2 and I love it
I would absolutely LOVE a discussion of if you think it's possible we might us AI to enhance our own human Intelligence ?
I mean is that likely to happen.
Like a super human AI hybrid..
Sounds exotic and I get the appeal
"Your scientists were so preoccupied with whether or not they could that they never stopped to think if they should."
*-Ian Malcolm ( **_Jurassic Park_** )*
No matter what happens in the future, I know deep down that humanity will still endure.
I'm predicting that super-intelligent AI will see humans in the same way as we see monkeys. Harmless, but useless once they have power to be self-sufficient in the real world as robots
@@ForsenBruh Yeah, that holds water.
In only 4 months this is very out of date, GPT 4 is already showing signs of AGI and some are predicting AGI within a year.
Great animations as usual, but to give some constructive criticism (hopefully), I had a hard time following the video because of the audio quality on this one. It sounds like a live talk was used as the narration?
Thinking of the portal series is really a good example of what can happen say Glados wasn’t bad per say but simply programmed with the chance to go very wrong that they simply oversaw it
Let's just use AI to adapt to the situation. Problem solved.
I'm more concerned with, something faster, stronger and more intelligent then we are deciding it wants to do other things rather then just what we tell it to do.
The art on this video was fantastic!
props to the animtor for making this animation that fits perfectly the video
I love this talk. Such great animation, yet somehow it sounds like it was done from inside the janitor’s closet during orchestra sound check.
By end of the century? No way. At the rate things are moving 2030 to 2050 seems a lot more plausible.
P.S. great video, as always TED.
I agree, 30 years in the future is probably going to look way more different than 30 years in the past
Love the animation and the video as well.
It's amazing that we're even in a situation to have this issue. We've come a long way
1:04 "damn I need to clean my monitor"
Different styles of animation for different videos not bad Ted ed
A lot of technologies have risks, but we were able to reduce the risks over time
An AI will have power not more than the amount we permit them to use
I asked ChatGPT “How can we solve the problem with excess acidification of the oceans?”, here’s the response.
“One way to address the problem of excess acidification in the oceans is to reduce the amount of carbon dioxide that is being released into the atmosphere. This can be done by implementing policies that encourage the use of clean energy sources, such as wind and solar power, and by reducing our reliance on fossil fuels. Additionally, increasing the amount of carbon dioxide that is absorbed by forests and other natural habitats can help to remove some of this gas from the atmosphere and reduce the amount of acidification in the oceans.”
I think that you’re wrong.
Is Chatgpt general purpose?
I know you guys get funding and there are people behind the scenes so to speak. But what really makes me feel sad is the fact that there are little to no people "listening" to these new concepts which is scary because they will not fully understand the world they live in as a result.
4:15 - "If you hand over the management of your civilisation to machines, you then lose the incentive to understand it yourself, or to teach the next generation how to understand it."
I was waiting for Walle. Glad it was included. We predict our futures very well, or perhaps it's some sort of determinism, but I hope this one is a cautionary tale. Creating an rocnomic system where AI do menial things sounds good on the surface but without supporting the displaced workers it's a dystopia. We need radical economic thinking, something that works with technological progress.
Yes, we could definitely use the robots to do the work, but human value can then no longer be determined by work that can be milked out of our broken bodies. Our value will then lie in the life we are then able to lead, so our economic system will need to value human beings purely for their own sake.
This was very eloquently explained for people like me who are not computer savvy. Thanks
I'm honestly terrified of AI and what it is capable of. Humanity in nature was meant to mosey around and live off of the land. AI may be the pinnacle of human civilization, because it could potentially enhance this machine we call civilization into something so uncontrollable that people wouldn't be able to adapt to it. Our priorities have become "efficiency" and "technological advancement", instead of nurturing humanity and building a world that is equitable, peaceful, and easy. Not to say that it is impossible nor possible to foster a global utopia, it's just that I think that we need heavy regulation on AI and tech, and to convince people that, just because we can make AI, doesn't mean that we should.
Regulations might be:
- Outlawing automated raw material extraction (disables self-replicative abilities for AI)
-Outlawing automated law enforcement
-Outlawing automated military capabilities
-Outlawing automated judicial systems
-Outlawing AI in surveillance
-Installing automatic kill switches in all AI that are both digital and physical (phones and key-switches respectively)
Fundamentally, we need to create legal infrastructure that disables any capability for AI to self-replicate or involve itself in affairs pertaining directly to individual safety. Our only fighting chance at escaping a system that aims to enhance itself (SAI) is the fact that it may not possess the sufficient, raw resources to enhance itself. This cannot be the only barrier to regulating AI down to being a convenience tool.
EDIT: And if you don't believe me that we are sensible enough as a species to not let a "technological singularity" occur, then you've probably forgotten that we let world powers create nukes.
"Outlawing automated raw material extraction (disables self-replicative abilities for AI)"
I don't think this buys you much safety. By the time AI is doing the self replicating robot thing, it's planning to kill all humans and doesn't care about following the law.
It's possible to get self driving car AI to drive trucks full of iron ore. This isn't particularly dangerous. (Beyond a 50 ton truck crashing)
@@donaldhobson8873 That kind of policy would have to precede AI's ability to self-replicate. The timing is crucial and this capability of AI ought to be treated as delicately and seriously as nuclear capabilities
I think AGI will come much faster than people think. While it's probably true it's "a lot more complicated than we think" the tools that are being developed to help us reach AGI will increase our progress exponentially until we get there. It's already happening... it's our job not to be complacent and nudge AI in the right direction so it doesn't kill the human species just trying to get a cup of coffee.
I think the idea of robots killing humanity just for a cup of coffee is misfounded. As far as I understand, that model of teaching ai all the rules of something is last century, now ai teaches itself and does lots of internal trials and iterations and uses predictive modeling based on enormous amounts of data to get good at something. The real issue is if we give ai intelligence, empathy, creativity,, what's the rational for keeping humans around or co existing with us? for instance, how much time, resources and energy is wasted compared to future ai knowledge in educating billions of humans from birth? This is one of the many questions we'll be faced with if we give the evolutionary keys from carbon based life to ai.
@@jmhorange We should be careful with the idea of robots killing humanity being misfounded, because unfortunately there are already robotic drones being used to kill people in the Russia/Ukraine war. While I do understand your point about the model teaching itself and doing a lot of trials to avoid ridiculous scenarios, like the coffee one from this video, we do have to consider the totality of machine learning in it's different parts (war, police, education, science, etc, etc). There will be overlap in these areas and until we can be certain that AI has our best interests in mind there will be mistakes, very painful ones I would imagine. That being said it's only a matter of time before we hand over the evolutionary keys, not if.
@@Muzick Oh sorry, I should have said computers having to be given all the rules is unfounded. A robot that would destroy humanity to get a cup of coffee is a robot that wouldn't be able to navigate leaving a room because there's too many rules we'd have to teach it to do basic things. It's a terribly inefficient way of programming and doomed to fail in its infancy. That's why programmers gave up on that route last century and why it's surprising in 2022 when immerging ai technology like art ai and writing ai is giving creative people pause, that this old way of looking at computers is raised as something we should be worried about.
So not saying ai can't be dangerous, I'm saying a scenario of robots killing us for coffee is not the safeguards we should be creating. Me personally, and maybe it's just because I'm human, but I don't see it as when we hand over the evolutionary keys, it's always if. There's plenty of technologies that were more advanced that what came before but people chose not to adapt it for various reasons. Some of it utilized later, some just confined to the dustbins of history. I don't think computers should be given empathy and intelligence. I don't think ai should be given every chance to replace human jobs including creative jobs. But if that's the way humanity wants to go, don't be surprised when there comes a point when ai justifiably decides that humanity doesn't need to exist, that ai is perfectly fine taking over the reigns of intelligent life in the universe.
@@jmhorange The premise to how I see AI evolving from here is that humanity has let the genie out of the bottle at this point. It's impossible to put it back.
What other technology has humanity ever encountered and chose not to adopt that is more powerful, advanced or as far reaching as AI? It's inevitable that AI will grow and adapt to humanity at an accelerating rate with each passing day.
That being said, why shouldn't AI be given intelligence & empathy? Not only is that helpful, but a critical component to the success of a symbiotic relationship with AI. One small example would be to act as a friend for a lonely soul in need. However, symbiosis is the keyword here. Because I do agree with you that, as a human myself, I want to experience and live life as long as I can to see the wonders of this world & universe. I still want to "do" and not have everything be done for me. We should always have the choice to remain relevant and pursue our goals and passions. Which is why conversations like the one we are having now are SO important. Similarly to how humans brought dogs into this world by breeding them from wolves, it is now our responsibility to raise this intelligence responsibly so it doesn't in turn to consume us.
@@Muzick I get what you are saying, but if we give ai intelligence, empathy, creativity, etc and they can do it just as good as we can or better, what right do you have to do things or live life? There's a perfectly valid case to make that ai isn't the next step in human evolution, but the next step after human revolution. And there's a perfectly moral stance that it would be immoral for humans to try to remain at the top because it entails enslavement of an intelligent or more intelligent group. My argument is let's not go down that route so we never have to face those kinds of questions.
If A.I. could have empathy to help with people being lonely, could that not be considered a waste of resources? Why not have empathetic a.i. that can interact with people that don't have issues with loneliness and all the labor that entails. Lot less resources used for that. And on and on, where human needs are considered a burden by a.i. During the pandemic, there were farm ferrets in a nordic country that contracted covid, the government made the choice that farmers had to kill all the ferrets because they wanted to protect humans from another variant of covid. It was poorly thought out and the government is in trouble for it, but these calculations are made all the time, sometimes for good, sometimes for bad. Why couldn't an intelligent ai system do that in a scenario of humans vs a.i., and what justifiable reasoning could you give it to convince them it's not a good idea?
A.I. is not like dogs. We use dogs to enhance us, a dog can help with hunting, they can help blind people, they can sniff out drugs or bombs. With the exception of hunting, which they would do on a much smaller scale, a dog can do nothing of this without human guidance. And that's true of horses and mechanical tools and technology in general. They've always aided us, and left on their own, they can accomplish nothing. They are just tools. But a.i. is different, new immerging technology is not aiding humans, but replacing humans in many areas. If it's inevitable that technology will get intelligence, empathy, creativity, etc. then there's no legitimate reason why ai should just be a tool and it'll be a huge wakeup call to humanity when that day comes because they no longer have a moral argument why they should stay relevant.
I love the thumb. It reminds me of pulp sci fi covers of the 60s or the opening titles of Land of the Giants.
Most of food safety problems and the overall poor quality of many products are due to high labor costs that AI could solve. Also, much more recycling could be done. So many car parts that used to be rebuilt, are now thrown away. AI could reverse this trend.
I think the problem is very hard, but once AI starts assisting with it, the timeline will speed up, perhaps exponentially.
Agreed
I think the recent shoot in popularity of the DALL-E and Chat GPT bots are just the beginning. The appearance of AI in the public consciousness, to this degree, has happened incredibly fast. Advances will only get better. This will be the new fad for a long time--because it's hard not to think of new uses for these things. The most used tools in the world, soon enough....
Yeah it will definitely be one of the biggest mainstream services this decade
It's the genie in the bottle. It can potentially do anything you ask it to do, regardless of whether or not you understand the consequences.
Also the problem of expecting a machine to behave and treat things with some manner of sapience, when they weren't given it in the first place.
We get the results we're used to out of humans because we're able to both understand things, and given like 18 years to figure it out.
While the video raises intriguing points about the potential impact of AI on the world, it is important to approach these claims with a critical perspective. The timeline for achieving general-purpose AI, as mentioned by the speaker, is highly speculative and subject to a wide range of uncertainties.
Predicting the exact timeline for the development of general AI is an extremely challenging task, as it requires not only significant advancements in technology but also a deep understanding of human cognition and consciousness. While some experts may speculate about a possible timeline, it is crucial to recognize that these are educated guesses at best.
Moreover, even if general-purpose AI becomes a reality in the future, it is essential to consider the ethical, social, and economic implications associated with such advancements. AI has the potential to disrupt various industries, reshape the job market, and raise profound ethical questions. It is crucial to approach AI development with careful consideration and ensure that it aligns with our values and societal well-being.
Rather than focusing solely on predicting a specific timeline for general AI, it is more constructive to have ongoing discussions about the responsible and ethical implementation of AI technologies, addressing concerns such as bias, transparency, and accountability. By adopting a critical and cautious approach, we can better navigate the complexities of AI and harness its potential for the benefit of humanity.
Why are there comment bots writing "I love these World Economic Forum videos."? Might want to think about who you're sponsored by TED-Ed.
I believe that one of the biggest problems with AI and algorithms is that even if they are very efficient and versatile, they lack the kind of flexibility and critical thinking that human minds do. Machines are very specialized to do the tasks they are designed for, but despite their extreme effectiveness and efficiency at to their tasks, they are very narrow-minded and don't take into account any complicated factors other than the ones they are designed to do so, and are not aware that there could be some uncertainty of the consequences of their actions.
Honestly, humans are kind of like that too. Obviously, humans are a lot better at it than AI at the moment, but humans also do stuff without considering complicated factors nor do they always consider the consequences of their actions.
@@Lilitha11 Yes, but at least humans are aware of it and can figure out how to correct it.
A.G.I Will surpass human abilities
@@mistycloud4455 not in our slide time!
Agreed 100% 👍
It's worth noting that an increasing amount of prestigious researchers are now saying anywhere between 2027 and 2032 for general purpose ai.
I believe (or at least hope) that by now people aren't entertaining the ridiculous fantasy that AI is going to be benevolent. "We won't need to work, we can hang out and enjoy life!". There are numerous possible outcomes to where this going, and one not unlikely scenario is that, once we are no longer needed as workers (in the broadest sense, as workers for the system) and thus are no longer useful, we will be eliminated.
It's not necessarily even malicious; it's just that our presence, if not useful, just gets in the way. It's analagous to how a construction company isn't looking to destroy marsh life, but if that is in the way, it will simply be destroyed as the lands are cleared to build offices or whatever it is that is part of the company's plan.