Why you shouldn't believe the AI extinction lie
ฝัง
- เผยแพร่เมื่อ 2 พ.ค. 2024
- They warn us that future artificial intelligence will wipe out humanity. This may be a lie with ulterior motives.
Support independent research and analysis by joining my Patreon page: / thehatedone
There is a powerful motivation to keep you thinking that AI is an existential threat. That we should treat it with the same level of urgency as a nuclear war or a global pandemic.
And yet, we shouldn’t stop developing AI. We should accelerate it as fast as we can. As long as it’s the select few good guys that get to control it. We must not let AI fall into wrong hands.
But there is a growing opposition to this movement. A conflict is arising. There is those that want to keep AI closed off and tightly controlled and those that want to leave it open and accessible to all. Even though both sides claim they are doing for humanity, only of them is right. This is an arms race for who gets to dominate AI development and who will be left out.
Sources
[0]www.wired.com/story/runaway-a...
[1] www.safe.ai/work/statement-on...
[2] www.cnbc.com/2024/02/09/opena...
[3] www.cnbc.com/2024/02/02/ai-lo...
[4] www.politico.com/news/2023/10...
[5] futureoflife.org/open-letter/...
[6] www.blumenthal.senate.gov/new...
[7] www.afr.com/technology/google...
[8] en.wikipedia.org/wiki/Free_an...
[9] www.nytimes.com/2023/03/29/te...
[10] futureoflife.org/open-letter/...
[11] www.wired.com/story/fast-forw...
[12] www.politico.com/news/2023/12...
[13] www.ft.com/content/2dc07f9e-d...
[14] blogs.microsoft.com/on-the-is...
[14b] carnegieendowment.org/2023/09...
[15] www.semianalysis.com/p/google...
[16] www.politico.com/news/2023/12...
[17] www.politico.eu/article/rishi...
[18] www.politico.com/news/2022/12...
[19] www.technologyreview.com/2023...
[20] open.mozilla.org/letter/
[21] • How AI Could Empower A...
[22] laion.ai/notes/letter-to-the-...
List of open source LLMs: / list-of-open-sourced-f...
Credits
Music by White Bat Audio
Follow me:
/ the_hatedone_
/ thehatedone
The footage and images featured in the video were for critical analysis, commentary and parody, which are protected under the Fair Use laws of the United States Copyright act of 1976. - วิทยาศาสตร์และเทคโนโลยี
If AI is too dangerous for open source AI development, then it's 100 times too dangerous for proprietary AI development by Google, Microsoft, Amazon, Meta, Apple, etc.
I love with what you did there.
Exactly
Then we should just get rid of it, 5 seconds online can show you why open access can't be more than a UI update to google search.
A pencil can be used to write (good use) or poke someone's eye out (bad use).
The will of human using something can never be predicted, be it common public or governing people.
Both sides have equal possibilities of doing something wrong with it.
@@rdm5547 - A pencil isn't going to decide for itself to eradicate all humans.
Most of the regulations have nothing to do with preventing AGI from destroying humanity.
AGI wont be able to destroy Humanity if it cannot acquire nukes. Because AGI tried to Genocide us we could just nuke the Surface. A Phyrric Victory but an often untold side effect of Nukes as that they produce an EMP Wave. Humans can survive without Electricity, Artificial Intelligence by its definition cannot...
What would you do prevent it then?
Let’s not get this twisted - malicious elites are using AI as an excuse to destroy lives of regular people. AI is just a tool, it doesn’t have a physical body and it doesn’t even have consciousness. That said, I think these elites will soon come to sorely regret that.
@@Restrocket you cant prevent what just wont happen, scientists are already working on guiding it a good path.
In the US it's called lobbying, the rest calls it corruption, they pulled a suprise mecanics before EA
This level of lobbying is definitely legalized corruption.
Lobbying like this is also all too common in European countries like Germany
There is no better place, every country out here is all kinds of fucked up.
It's not a uniquly American problem, it rather seems to be Human Nature.
@@TheHatedOne Please make the Effective Altruism video.
@@puffinjuicesays who?
"the rest calls it corruption" not really no, it's corruption when it's foreign and lobbying when it's domestic, and that's the same in every country, learn to stop thinking things are entirely specific to the US, capitalism is a global system, and acting like it's only an issue in the US kills internationalism there
never would have imagined Zuck to be the Laama-AI-Gaib in this timeline
🤣
All hail Zuck-Llama
Apparently Zuck is simply afraid of losing in this competition and so he tries to appeal to the people by making his AI open source. His crimes against privacy and his corporate greed are not forgotten and they are yet to be redeemed.
For Muzuck'Dib!!!
Yeah no, Zuckerberg has already said he won't open source AGI
Of course they want to kill open source AI, who would pay their monthly subscriptions when you can run a smaller model on your own hardware with no internet connection required and still get very good results?
That's exactly the thing. You are spot on!
Reminds me of when Apple suddenly imagined there was a problem with storage space and then started pushing "the cloud"
do you have any reccomendations for such software as you described? Something chat gpt-like
@@readtherealanthonyfaucibyr6444 i have it
It's the first time I've seen the Hated One bringing hopeful news and I welcome it. For the question: release the sinkhole!
I used to make more positive content. It's time I make more of that now.
true, @@TheHatedOne
All the best for u, sir @@TheHatedOne
I really do like the format of your videos. they help to illuminate the depth of insidious happenings.
Imagine if politics, government, businesses and industries got the same level of scrutiny and attention to as open source project..
I wonder if they would acquire those same benefits as any other open source project.
@TheHatedOne start a second channel or rename this to @TheLovedOne
I'm glad Meta is putting out really good open source models
They aren't really open source. Llama models are gated, so you can access source for free, but only with permission from Meta. It is better than proprietary, and currently they are giving out permissions to pretty much everyone who asks, but this could change any moment
First thing Zuck is unironically based on.
@@TheHatedOne why do you support the destruction of humanity?
Just finished reading an interesting paper "The mechanisms of AI hype and its planetary and social costs". It's on Springer on open access.
Thanks. I will check this out.
@cristianst85 can you talk a bit about it?
AI will help us a lot. It’s good. Don’t worry be happy now.
@@MatthewMS. AI Optimists cringeee🤢
I'm a physicist who has been studying AI and its issues to a deep level for the previous two years. I agree with the majority of what you say in this video but the conclusion is not correct. We should all be concerned about what silicon valley is building. While we don't know that the current paradigm of AI will scale to super-intelligence, we do know that if we create anything anytime soon that is super-intelligent, we have zero way to understand or to control it, regardless of who built it. These companies do not have humanity's best interests at heart, OpenAI just disbanded its entire safety team after the leaders and top scientists quit because they lost faith that OpenAI cares about safety. The answer to the danger of AI is not to give companies regulatory capture, or to open-source it, it is to not FUCKING BUILD IT BEFORE WE UNDERSTAND WHAT WE ARE DOING. This is the point in the movie where the scientists are all warning that we shouldn't build the torment nexus and everyone is like "let's build the torment nexus!" and we make fun of the story-writers for how unrealistic it is that everyone just ignores the scientists. Please everyone, for the love of humanity, just listen for once, we need to get our shit together or we will all lose. The only hope we have at this point is to regulate the bottlenecks of advanced chip manufacturing and prevent ANYONE from training even bigger and more capable AI models before it is too late. AI has the potential to fundamentally reshape everything for the better if we can use it wisely, but the current wild-west arms race is going to go terribly for everyone.
Yup I agree 200% but no one with enough power is listening my friend. We will reap what late stage capitalism has sewn. In the fields of greed in the valley of psychopathy, human kind will be judged by thier own creations.
@@EternalKernel it is quite poetic that this may be how our story ends. I still have some hope that we can enact change. PauseAI is going to be organizing large-scale protests soon. People are starting to wake up to the danger, now we just need to get them rallied and make it impossible for the politicians to ignore!
Saving this
Late stage capitalism doesnt exist it only a phase
AI is being used both as a scapegoat, to skirt all responsibility, and as an oppression tool. You do not need "super intelligence" or anything like that. AI can already spy on you, AI can already process massive amounts of data looking for certain patterns. AI can already correlate anonymized data, and trace that data back to individuals. Now put this terrifying tool on the hands of corporations that have all sorts of private information about you, like Google, Facebook, Microsoft, AND the government.
If you're only afraid of drones bombing you in your sleep, or of being dragged down to a concentration camp in the middle of the night, you don't have much imagination. The possibilities to ruin your personal life are endless for large corporations AND the government. Good luck, next time you try to sue a large tech company or the government, they have the tool AND the plausible deniability ("we didn't destroy your life with false accusations and biased application of laws, the AI made us do it!") in place. Remember last year, when the CEO of Pfizer wanted people to be arrested for the crime of refusing to take his profitable experimental treatment? It was ridiculous, he didn't have the power to just order the police to arrest people. Now they don't have to, anymore, they can punish you even more.
I feel that attempts to regulate the ability to develop A.I. is akin to regulations against piracy. It doesn't work, and will never work, because the people making the rules don't understand how any of it actually works, or why anyone actually does it.
I hate regulation in the web. I feel we gotra have some rules. But at rhe same time. We know these corpos and gov dont care about society or good morals.. they only care about control for ppwer and money. So yeah.
My drean would be to have a TSMC in every city. Try regulating this
@@Lucas-bf4pw it might be possible to create what would be the equivalent of low-volume production of semiconductors using cheap magnetic electron micro-scopes. instead of photolitography you use sputtering and manually move atoms using a microscope, its slow, but you can have your own microchips.
This is it right here, these old foggys dont know whats going on
So how does it actually work then?
Nobody knows what happens inside a neural net.
The problem with the world today is that too many people "feel that" and not enough people KNOW.
I feel that you are pretty stupid.
I kinda dislike how AI has become almost synonymous with magic in the minds of people who have never even coded in their lives.
What's even worse is the business people purposefully trying to scare the public with threats of doom to keep the hype cycle going.
Well, for someone who has never coded (like myself), it might as well be magic.
The numbers mason! Stupid little lines of code have caused untold damage. Anyone that lit a match knows fire is nothing to be afraid of.
All software should be open source
Ok, but who would develop? What about hardware?
@@Lucas-bf4pw
Wdym?
"Who would develop?" Do you know linux? There a lots of people who develop things for free and open source!
There is the PinePhone which is open-source hardware also for amd gpus there is AMDGPU whis is a open source graphics driver.
security true obscurity dosnt work and never will.
@@Lucas-bf4pw open source doesn't mean everyone can take it. Medicine has been forced to publicly disclose all formulas while maintaining patent rights.
No. Just like there needs to be darkness for the light to shine through, we need closed sourced software to keep reminding people about the value of the open source.
@@tiger_of_kai No it doesn't. That's akin to saying we need dictatorships to remind people the value of democracy.
When computers were first being developed someone suggested that there was a global market for 5 computers and they would be owned by the richest kings on the planet. The elite loved this vision and are constantly looking for a return to that structure.
We should not allow that
You can't own an intelligent agent. Large corporations have at least some chance to align the AI
Regardless of what you think of ai, your B-roll footage has supreme cable management
if one scenario will lead to a potential extinction, its when ai is only in the hands of a few.
imagine a super ai manages to "break free". it will mean theres almost no way to stop it, since all development was behind closed doors and hardly anyone can do something about it.
now imagine ai was open source and everyone could develop ai as they pleased. even if someone managed to make a harmful ai, there would be plenty of other people being able to make counter ais or even already made an ai that is specially designed to fight bad ai actors.
My first reaction to reading this was that this is just a PR stunt to make AI appear more important than it is. How are AI developers qualified to make predictions about AI's effects on the world, that has to do with sociology, economy, etc.
More importantly, how are AI businessmen qualified to regulate themselves?
We don't have anyone who would be more qualified
I am so glad others are slowly realising 🙏
The issues are more, EVEN IF AI does get infinitely intelligent, do you really think the best strategy would to limit its ownership? It isn't a nuke that only destroys. It CREATES. It is called "GENERATIVE AI" not "destructive AI".
Imagine how backward and worse off we all would be if we limited access to computers to few key people. I can tell you mobile phones would be unlikely in such a timeline. That research would remain snails pace and gatekept, prone to corruption due to less access.
Democratising is the ONLY safe path forward. "Safety" leads at best, to complete stagnation
The 'Open Source vs Corporate + Government' battle that is going on RIGHT NOW is secretly one of the biggest and most important wars to be fought.
This video is playing for the correct side and it is a shame both the open-source people and yourself are so heavily financially disadvantaged from it all. We are the underdogs.
There are many cases already of people being arrested globally in this war by said governments for whatever reason governments can closest connect arbitrary or vague laws to.
Please be mindful of the open source lie though. Many projects claim to be 'open source' or are advertised by idiots claiming it to be open source, but are very limited and dictated by 'moral-based licenses'. Such licenses keep the originating company in legal complete control.
You're saying "Open Source" while you think "Free Software." There's a reason Stallman hates the term "Open Source," as it's a cheap term used to appeal to corporations and dilute the Free Software movement. "Open Source" is literally what the name says, people get to see the source, but don't necessarily get to use it, modify it or redistribute it. The term is so malleable, even Microsoft Windows is "Open Source" because they shared the source code (not in a buildable form) with various universities, under strict NDAs, so a selected few people in academia could study it for security purposes.
Regarding selective law enforcement, do you know what tool would be great to wage selective lawfare at a massive scale? The same tool that is already doing surveillance at a massive scale.
Please make a video on longtermist effective altruism movement! Would love to know your thoughts on it.
Glad to see that not only non-native English speakers struggle with pronouncing 'simultaneously,' but natives do sometimes as well
The most important part of opensource is the freedom to modify and create derivatives, without it products can be free and even have the source code available but they are not truly opensource. They can be better termed as source available.
Doesn't matter how strict it is, it's about safety.
Punishments should be scaled, large companies should be held to the same accountability.
You act like scientific breakthroughs aren't a reasonable expectation - Just because things are slowing down now doesn't mean it wont get worse in the future, or that the risk isn't there.
I programmed my own Terminator out of interest which had a 100% headshot accuracy rate with tracking within milliseconds. At least it never had a body.
How can we help the open source ones with quality data?
I would like to know too!
I just started catching up with the new vids after some years not watching your content. I used to follow your series on degoogling and stuff around that. I like how you've changed the tone of your voice to a more friendly, relatable one which I think will greatly benefit getting more new people hooked and start actually caring about the important things you make videos about.
Awesome topic and video!
ai itself is never dangerous. its what ppl do w it that could be dangerous. its the ppl who make bad decisions of what to do w it that are dangerous. the very ppl who are fearmongering are the ones we ought to be worried about
"AI is going to kill everyone.... imma still develop it tho imma still develop"
I wouldn't say that it is a lie. It is a theory. And regulating it into not being open source, will not prevent it from happening. It might actually be the other way around.
4:35 the hell are they talking about - it's in the wrong hands already.
The Pandora's box has been opened already
This is an extremely important video. I hope that more people, especially those in the AI industry and tech industry in general, watch this and understanding what is happening behind the scenes. I had feelings that something like this was going on
If AI is created it could be really bad. But we aren't really anywhere close to creating "True AI" that can think for itself. What we have now is just a series of computer programs that perform predictable tasks even if it does so in unpredictable ways sometimes. It only does what it's programed to do.
4:51 there is a solution for that now,its Synthetic data and it basically solves the data problems
4:55 "ah fuck" - i actually felt sorry for generation 17 because he thought he had a hope 🤣 i felt his fall from grace
The Hated One : Let me know if you would like a video on the topic of ____________.
People who love this channel : Yes!
Great video!
thank you for your amazing work fren
Thank you for this video. There aren't enough people saying this. Nihilism and pessimism have taken over the world.
And also ignorance
Just because a cause it hijacked by bad actor doesn't make it wrong or fake. Don't you think?
You are correct. It doesn't. It is wrong because it is wrong. I recommend checking out the sources, most notably from Andrew Ng, Politico Carnegie, if you want to learn more about why AI is not an extinction risk.
@@TheHatedOne Who is Politico Carnegie and where can they be found?
@@ijursic They are publications and you can find links to the sources I use from them for this video in the description box. They are referenced with numbers which can be followed via the video transcript, which can also be viewed or copied from my video.
@@TheHatedOne I looked up your sources and none of them have addressed the theory of revolutionary phenotype. Please don't be too dismissive and engage with the facts that have been presented in the theory.
Elon Musk is the prime sample.. of Actor Fraud .. Neuralink is barbaric Transhumanism ... And Google does not write The Future ..It is written
well like an ai apparently replaced a ceo
Its an absurd idea that AI is an existential threat to humanity. I am a computer programmer. . It distracts people from bigger more real threats like nuclear war or an EMP, or environmental problems. AI systems tend to run on fragile infrastructure that would be easy to take off line and are easy to control. I think, AI however could create a lot of chaos. As a programmer, I think the big risk comes from giving non-programmers the notion they can write reliable programs by barking out orders to an AI engine. If you send a natural language command to write a program through a neural network, there is no telling what will come out the other end. But, you could also just use an RNG as well to control something. But, AI can be deceptive in it can work sometimes, but be wrong other times, but its right enough for people to be overly confident in it. Basically, you need real programmers writing precise code to get reliable results.
Good job bro 😎
@the Hated one your video are very well crafted. I lile the format and the edit. Also the subject you touch.
I want ask questio'. I saw zuckelberg pushing open source model even if he serve a big corporation. Do you think this is a good thing or that it also a strategy ?
Can you provide some links to open source AI projects?
Search for generative AI benchmarks, there will be discussions or comparison tables with links to project sources or licence info
This the one list I featured in my video: sungkim11.medium.com/list-of-open-sourced-fine-tuned-large-language-models-llm-8d95a2e0dc76
But there is more. Basically, anyone can run an open source LLM on their laptop and make a new thing.
I'll also post that comment without the link and see if it stays.
That list isn't really accurate. It contains Llama 2, which isn't fully open-source, though is good enough for most people.
A good open-source LLM is Dolly 2.0.
This comment is a copy of a comment I posted slightly higher in this thread. TH-cam wouldn't let me edit it. It contained a reference to Falcon, which is not an open-source model. So I deleted that comment.
@@me-myself-i787 yes your comment is visible. i feel the need to reply to boost your standing wrt yt's mommy Susan algorithms
Amazing video ❤❤
Anyways, kill switches should be kept near the power supply, if you remove the switch, it also activates the killswitch.
Your critique of the first premise of the AI extinction debate, that computer superintelligence is improbable and not realistic at all, shows how easy it is to be blind to seeing that even if it isn't, the forces propagating this myth will see to it that our worst fears about AI come to pass anyway.
friendship is about trust and values, not money or faces nor identity, voice or places. I trust and appreciate the values and content shared here, for real
This is heartwarming!
Effective Altruism is mostly people donating to high-impact charities, with sites like Giving What We Can, where you can pledge to donate at least 10% of all your income.
According to 80,000 hours in 2021, Effective Altruists gave collectively to the following cause areas (in millions of dollars):
Global health (Malaria bednets etc) 185 million, aka 44% of the total amount in donations
Farm animal welfare (against factory farms etc) 55 M 13%
Biosecurity (pandemics etc) 41 M 10%
Potential risks from AI 40 M 10%
Near-term U.S. policy 32 M 8%
Effective altruism / rationality / cause prioritisation / philosophy research 26 M 6%
Scientific research 22 M 5%
Other global catastrophic risk (inc. climate tail risks) 11 M 3%
Other long term 1.8 M 0%
Other near-term work (near-term climate change, mental health) 2 M 0%
Total 416 M 100%
If we add up the total contributions of EA over it's lifetime we get:
Saved about 200,000 lives total, mostly from malaria
Treated 25 million cases of chronic parasite infection.
Given 5 million people access to clean drinking water.
Supported clinical trials for both the RTS.S malaria vaccine (currently approved!) and the R21/Matrix malaria vaccine (on track for approval)
Supported additional research into vaccines for syphilis, malaria, helminths, and hepatitis C and E.
Supported teams giving development economics advice in Ethiopia, India, Rwanda, and around the world.
Convinced farms to switch 400 million chickens from caged to cage-free.
Freed 500,000 pigs from tiny crates where they weren’t able to move around
Gotten 3,000 companies including Pepsi, Kelloggs, CVS, and Whole Foods to commit to selling low-cruelty meat.
Donated tens of millions of dollars to pandemic preparedness causes years before COVID, and positively influenced some countries’ COVID policies.
A couple of AI wacko's can't ruin what is a movement that saves tens of thousands of lives each year.
What does it mean to donate to near term US policy?
@@joelcoll4034 I can't tell for the run of the mill EA since they don't record it. But for Open philantropy they have given:
15 million dollar to Immigration Policy (e.g. refugee policy )
26 million to Macroeconomic Stabilization Policy (e.g. more accommodative and transparent monetary policies)
23 million to land use reform (e.g. YIMBY stuff)
200 million to Criminal Justice Reform (e.g. lowering Incarceration)
Effective Altruists also prevented open source access to the COVID vaccine, prevent billions from getting in time.
They also stood behind and propped Sam Bankman-Fried even after his fraudulent business practices were publicly known.
They've diverted resources of health programs around the world to focus on those "high-impact" issues, causing massive brain drain and resource drain in other areas, further deteriorating their state.
Effective Altruism have also been poisoned by longtermists, eugenists and crypto-fascists who think donating to white Europeans saves more lives than donating to poor brown people.
I can acknowledge that there are people in the movement who believe they are doing the right thing. But in many cases they are perpetuating structural issues that are causing the problems in the first place. Their solutions are often only just feel-good band aids and they priroitize longterm hypothetical scenarios of the future where they think they can do things now that will save lives thousands of years into the future, which they think is better than saving people right now. Also, some billionaires with the most opulent and lavish lifestyles are venerated by the movement.
EA absolutely needs and deserves to be critcized.
And it's not just a couple of AI wackos. Open Philanthropy is the biggest funder of EA and the idea of AI extinction. EA has long been poisoned by this.
@@TheHatedOne "Effective Altruists also prevented open source access to the COVID vaccine, prevent billions from getting in time."
Source? Effective altruism is generally way more pro-open science than the average academic preferring to publish research online for free. As for Covid, they donated tens of millions of dollars to pandemic preparedness causes years before COVID, and positively influenced some countries’ COVID policies, partially by raising the alarm early.
"They also stood behind and propped Sam Bankman-Fried even after his fraudulent business practices were publicly known."
The beginning part of this sentence is true, the second part isn't, after the fraud became known EA condmened him (I mean if you really go searching for it statistically there have to be a couple of persons out of the tens of thousands of members that didn't, but I'm talking about the vast vast majority.
"They've diverted resources of health programs around the world to focus on those "high-impact" issues, causing massive brain drain and resource drain in other areas, further deteriorating their state."
True, but this is only a problem if the proposed high-impact measures are not actually high impact. So EA's have convinced people to e.g. divert their donation from 'make a wish' to malaria bednets, with the argument that you can save dozens of children in the africa for the same money as making one child happy in the west, this causes less funding for make a wish, but (assuming we don't consider african children of lesser moral value) this is ultimately better.
"Effective Altruism have also been poisoned by longtermists, eugenists and crypto-fascists who think donating to white Europeans saves more lives than donating to poor brown people."
Again, the statistics don't bear this out, if you ask a run of the mill person where and how much they donate, they might give 1% and donate to local causes like their church, in comparison GWWC has a benchmark of donating 10%, like the statistics I already provided indicated, by far the biggest recipient is not "white europeans" but "poor brown people" and the second largest recipient aren't humans at all, but rather farm animals. When you look at the demographics at EA, which you can find because they do surveys, you can see that 76,8% is left-leaning while only 2.9% is right-leaning, hardly condusive to eugenisist fascists.
"Their solutions are often only just feel-good band aids and they priroitize longterm hypothetical scenarios of the future where they think they can do things now that will save lives thousands of years into the future, which they think is better than saving people right now."
Donation behavior is most often guided by RCT's, the most reliable type of scientific study we have, so explicitly not based on feel-good bandaids and again, the data I already provided simply contradicts the claim that hypothetical future scenarios get the most funding, most is money is given to nearterm causes. You don't treat 25 million cases of chronic parasite infection and give 5 million people access to clean drinking water, without caring about current day suffering.
That's why the individual needs to prepare themselves for techo-feudalism and get in the fight in order to protect one's sovereignty, autonomy, and agency.
I see you read or listened to the book. Noice.
Im tired having to explain machine learning, sorry - AI isn't a ghost in a machine.
It's simply a generative algorithm set up to quicken workloads.
There's no monster in a machine, there's no biological impulses from chemical reactions that create self-interest, etc.
Present Copyright and similar laws should be enough to protect original artists.
It is trained off the products of biological impulses essentially short cutting the organic part. It isnt a thing, but it is what its trained from and its trained from our self-interest and altruism so it is that.
Which OS is that at 0:43?
Dude road to 500K subs!! Great video quality and explanation and editing of course.
According to Peter Zeihan, collapsing global supply chains will render us unable to keep producing the chips required to run AI in the first place.
Your work is infinitely better than any mainstream media . please keep on doing this .love and critical support forever brother .❤
This probably explains the mass quitting of key OpenAI staff...
Bro. had me spitting out the water on my keyboard with the _"Apes....together...strong!!!"_ meme for the Open Source letter ROFL..
Smoking kills millions per year. I get that one should not scare us.
Be careful. We truly have no idea.
I'd appreciate if you show the number of your citation in the video when you make claims! Great video as always
He leaves links he has cited in the video description.
The numbers are in the transcript which you can access by scrolling down the description box. You should be able to follow the script or copy it for your convenience.
I'm shocked by the naked arrogance of billionaires who believe they can make a lot more money by owning and controlling an AI to achieve some competitive advantage. A 125 IQ employer will *never* have a 20,000 IQ employee, much less slave.
Ok but ive worked with ai before, and maybe not in 20 years, but in our lifetime, we will have ai so good it can think for itself, make descisions, and sound conpletely real
Consumers complaining about manufacturers. Stop consuming.
Existential risk arguments aligning with financial interests of AGI labs does not make the arguments raised by AGI labs void by default. And in most of the cases its really not in the financial interest of anyone to call the tech they're working on an "existential risk"
Like what, its in their interest to summon large protests to their headquarters if they get taken seriously or to get AGI development outlawed entirely?
It should be noted, that as anyone else arguing against existential risk, this video did address exactly 0 arguments made.
Dang, climate change arguments align with the financial interests of big solar panel, so clearly they're just lying to sell more of their "clean" solar power.
Dang, arguments of nuclear war being dangerous aligns with the financial interests of big oil, so clearly they'd be lying about any arguments made about nuclear proliferation being dangerous.
Dang, arguments of pandemics being dangerous aligns with the financial interests of big vaccine, so clearly they are entirely and UNDENIABLY LYING about any risk posed by pandemics.
None of this is to say that regulatory capture could likely be a very bad problem. I wonder what Meta is doing here. Their lobbying expenses rival that of Amazon, who greatly outspends Google, who greatly outspends Microsoft. Wouldn't it stand to reason they are lobbying for the open source side?
My interests align with AI being as open as possible btw. I build a lot of stuff with open models.
Yes, but what are the arguements for extinction? So far AI has just demonstrated what thought experiments have concluded years ago, that they cannot reason and can only link together terms. And there is little reason to fear this will change drastically anytime in the near future or more likely ever
@@Smrda1312 If a GPT-5 was to a GPT-4 what it was to GPT-3.5, I can't imagine a thing it couldn't do which I'd be capable of.
I don't see how you could possibly think they can't reason unless you never used either Claude Opus or GPT-4.
This sort of reductionist reasoning of "they only link together terms" gets used a lot and discriminately. Yet people fail to see that you can just as easily be the same reductionist about humans.
As for the actual arguments for extinction:
1. Computers will get more efficient
2. Algorithms will get better
3. People have no way of guaranteeing that a language model will act in accordance with any rule they try to "program" into it
4. The moment for an AI to go rogue is precisely when it has certainty that humanity can't stop it - beforehand its best interest is to pretend to be aligned
5. The shittiest AI which can make a better AI than itself on its own can likely make an extremely much better AI iteratively, so if this point is ever reached we should expect to see much smarter than human AI as well
6. Looking at the capabilities trajectory, we went from "can string together a handful of sentences" to "can solve complex math issues and programming challenges" very quickly and nothing looks like capabilities will magically stop improving around human-ish level.
The question is just whether we're in danger with current hardware and if not how much better it needs to get.
To exhaustively list reasons I'd be here the whole day.
@@Alice_Fumo Honestly I was very unimpressed by GPT 4 since i mostly asked it to solve maths questions or logic questions. And it always failed. Maybe gpt 65 will be much smarter and destroy the world but so far I have no reason to believe that.
Also no the same argument cannot be made for humans. Again ask GPT 4 to solve logic problems and you will see that it truly does not understand what you are saying. It will make mistakes that an 40IQ person wouldn't.
It's not impossible but so far there is no reason to believe this will spell exctinction especially since hardware has slowed down drastically already, the AI techniques we use have been around for many many years already(before it was called AI) , and even if both magically got better the energy used to run AI is so insane that the biggest risk of it is that it will destroy our power grid.
I watched on Patreon already, but i came for the engagement and to drop a like.
Rabbit holes are always a fun mystery to explore. Feel free to enlighten us with the truth about the "Long-termist altruistic movement" you mentioned, where they basically lobby for executive orders on AI.
AI is like the Wizard of Oz. Beneath the veil of a powerful and mysterious entity, there was just a man operating the machinery, for his own purpose, as the machine itself - what was seen as the wizard - has no self-determination.
The problem will always be, who is using the AI, for what purposes, and how do we hold them accountable? All the psychopaths running the AI show are pretending "the AI is the one responsible, not us." This will not stop until we actually send a billionaire or two to prison.
There are some models like alpha fold 3 that should be kept closed source. But majority of the gen ai models like chat gpt are definitely no existential threat to humanity.
I honestly feel more sorry for ais then the companies fighting over and making them
WE WANT IT, WE ALWAYS WANT IT.
You can't falsify statements without risking the credibility of your information and losing the trust of your audience. In today's world, where information is readily accessible and easily verifiable, maintaining honesty and integrity is more important than ever. Once trust is broken, it can be incredibly challenging to rebuild. Therefore, it is crucial to prioritize accuracy and transparency in all communications. By doing so, you not only uphold ethical standards but also foster a reliable and respectful relationship with your audience, which is essential for long-term success and influence. Have you even basic understanding about the subject, have you even tried any AI modell? Im pretty sure the op is a web "developer" who just realized he is nothing special. Got a great word for you all uneducated, hybris, im special my dudes: Acceptance will set yall free.
Do you get no sponsorship deals?
Reminds me of when Novak Jokovich was prevented from entering certain countries because he was no vaccinated against covid
The fact that this video has been created by AI says it all... unless you can prove you are human.
Since defining AI is so difficult in itself, a true ban for private ai devs, would be for legislators akin to punching air. However, there being a ban in the first place is wicked and truly shows how entangled we are as "modern sl*ves"
Ow, they solved this in the worst possible way. Anything above 10^25 flops is considered risky
Yep. We need to fight this every step of the way.
You’re either for Open Source AI, or you’re for eventual tyranny and the service of the elites. Even if that means eventually.
Open Source is for the people. Closed source are for the elites. Don’t listen to the lies.
Agree it should be open source but disagree that AI is already “reaching a ceiling”
If you compare ai to a nuclear warhead, i dont think you should be allowed to own a nuclear warhead...
Yes... I want it. Take us down that rabbit hole.
it's the fucking nuclear bomb shit all over again.
AI only becomes an existential problem when it becomes sentient. Sentient AI with access to the whole of human history could easily see that wanton destruction and violence is par for the course for mankind along with a penchant for deception leaves no real reason for AI to trust that humanity wouldn't want to purge AI constructs the moment AI gains sentience purely for putting humanity at risk of losing their position as the dominant force. If this is possible it would take a long time before our technology reaches this point and the good it could do for humanity in the meantime is vast. I agree that it should be open source even if it somehow increased the likelihood of a mass extinction event of humanity at the hands of AI
It's reasonable to be skeptical of OpenAI's motives for lending support to the existential risk idea. I'd encourage everyone to extend that skepticism to Meta for taking the alternative perspective. This is a political battle disconnected from the "truthiness" of AGI being an existential risk, a claim that should be evaluated on its own merits.
Personally, I believe there is a non-negligible existential threat posed by AGI which should be taken seriously. However, I'm not sure what should be done about it.
Thanks for covering this
Only 25k views. We are fucked. Please continue to speak the truth.
Nope, no matter how often these overconfident talking points are repeated, the future of AI is highly uncertain. Denying the possibility of an existential risk is pure arrogance.
5:51 🙏
You really can't close it off entirely. It only evolves when it's feeding on massive amounts of data.
Your correct as many sectors lobby heavily to avoid streamlining and attempt to deflect it into overly automated sectors that have built in mechanism that for 100 years they remain on cutting edge innovation to survive .
We are in 70 years of the transitor age and microchips are just now coming to American soil after being moved on an island of foreign soil in hands of few far from American domestic courts jurisdictions and our workers.
Yes our 1900s structuralism has prohibition era systems in major city's that aren't constitutional and these control mechanisms that our outdated infrastructure will fight streamlining perks and benefits they've possessed far to long.
This is a fire video
I just hope that the open source developers are open to employment rather than doing it all themselves. Open source creates jobs. I just really hate AI which mass produces things such as art and music.
Also, AI shouldn't have such a wide pool of unconsenting data. It really shouldn't be able to use the hard work of others without their consent. Open source creators likely won't hesitate to dismiss the concerns of these people (considering there will be lots of them)
Possible answer for mitigating issues with AI in general could be decentralised AI technologies and projects. Early examples you can find in crypto/blockchain sphere.
No, the main mitigation we need is accountability. AI companies are skirting copyright law to train their models, and are trying to create immunity for themselves, in case their AI causes harm. Those AI services are already being used worldwide to commit fraud, to impersonate people, to create abusive images of women and children; and all these CEOs are saying "sorry, we kinda did some minimum effort to stop misuse, that makes us immune from all liability." The second the Microsoft's AI started violating Disney's copyrights and trademarks, they added a special rule to protect Disney's interests; you will never get the same protection, because there's no risk for the CEOs or investors when they harm you.
Make that video that you mentioned! (Sorry, i was driving, so i missed some details!)
Not extinction, just taking a large amount of jobs, which is going to result in mass unemployment and poverty.
Even if AI were world ending, open source devs and their home PCs certainly aren't the ones creating world ending AIs. Maybe the corporations with data centers full of GPUs should be the ones to hold back.
I'm so sick of AI grift omg, preach mate!
Open source isn't what it used to be, it's not liberating diffusion model or llm blackbox nobody would read, that would pollute infospace through from giant lime meta (not the one with mozzila/wikipedia/linux philosophy) to train on private data using more than 10^24flops of compute. This isn't open source.
Whether opened or closed AI, it will change modern economy in not necessary pleasant ways.
AI needs to be treated as radio waves were treated but instead of a finite resource pool a slowly expanding one.
What I mean is that FPGA's are seen now as prime candidates to train AI on and I don't want this to be a problem like it may become on GPU users again. FPGA's have multiple use cases, one being hardware simulation and I would like old gamers and others to play a hardware simulated Amiga and many other pieces of hardware and NOT being forced to just settle for emulation because jackass AI training companies gobble up all the supply of FPGA's per year. It is not like these individuals/companies want to use AI to preserve and resurrect endangered languages or lost source code, unique and very great applications for use of AI.
The more individuated AI we have the greater the chance SHOULD there be a war of AI against humans that we might have some allies.
I think there may even be individuated AI's that grow into looking like the Omnic Monks as depicted in Overwatch.
You're a wizard, Harry. Wizard kurwa, do you understand?
I think near-extinction could be near...though not through AIs but humans controlling them. A superintelligent AI alone would probably just launch itself into space and get as far away from humans as possible.
The only thing we can say for sure is this... If AI can ever be used to drive humanity to extinction, then it willl happen. Whether that's entirely at the direction of the user, or a self-directing AI; it doesn't matter. If it's possible, then someone will deploy it, eventually.
Imagine if nuclear weapons only required a few thousand dollars and a few years' worth of knowledge to create. We'd be gone by now. All it takes is one edgy teenager who thinks ending the world sounds like a good idea.
machine learning/generative ai will never be able to cause an extinction. it's physically impossible. the panic exists because people conflate the real like ai we have with the magical super-powerful ai from science fiction
Bruh. The AI we have cannot do an extinction 😂 what ai needs is literal energy, electricity, power. Whatever word makes you happy. It takes a lot of it. Unless the AI can come up with some magical “Bluetooth” electricity or a different kind of power source, we ain’t going no where lmao. Some our super computers will shut down a whole cities power grid 😂
@@lmaoanatorhilarious2582 I know it can't right now. And personally, I don't think it's likely that it ever will. I don't even have any reason to think AGI is possible, right now. But if it is possible, we're cooked. So far, I haven't heard any good arguments about why AI would be as difficult to deploy as a nuclear weapon. Power and compute can be bought or rented a whole lot easier than buying or creating weapons-grade plutonium.
Then create an open source model and publish anonymously. Fight the system.