They Think AI Has Gone Too Far! Time To Rethink?

So up until now, most of my videos have been around cool and fun gadgets that you can play with in the AI space to make art and videos and 3D objects and to get it to do monotonous tasks for you like build Google spreadsheets and things like that. But because I want this channel to sort of give everybody the TLDR of what's going on in AI right now, I can't avoid this topic anymore. And that's the topic of the negatives of AI. The downsides, the scale, scary stuff, the riskiness of AI. That is the hottest topic in the world of AI right now. And if I'm being true to my word of giving you the TLDR in the AI space, this is what's going on in the moment. This is the big talk of the AI space. This week a ton of big names in the AI space all signed an open letter to pause the advancements of large language models that are more powerful than GPT4 for at least six months. The paper states that recent months have seen AI labs locked into an out of control race to develop and deploy ever more powerful digital minds that no one, not even their creators, can understand, predict or reliably control. The paper asks should we let machines flood our information channels with propaganda and untruth? Should we automate away all the jobs, including the fulfilling ones? Should we develop non human minds that might eventually outnumber, outsmart all, obsolete and replace us? Should we risk loss of control of our civilization? They go on to state powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable. They call on all AI labs to immediately pause for at least six months the training of AI systems more powerful than GPT4. They then say if such a pause cannot be enacted quickly, governments should step in and institute a moratorium. These protocols should ensure that systems adhering to them are safe beyond a reasonable doubt. They also also say AI developers must work with policymakers to dramatically accelerate development of robust AI governance systems, including new and capable regulatory authorities dedicated to AI oversight and tracking of highly capable AI systems and large pools of computational capability, provenance and watermarking systems to help distinguish real from synthetic and to track model leaks. A robust auditing and certification ecosystem liability for AI caused harm. Robust public funding for technical AI safety research and well resourced institutions for coping with dramatic economic and political disruptions. Now this was published on the Future of Life Institute website and this open letter was signed by some very prominent names including Elon Musk, Steve Wozniak, Yuval Noah Harari Andrew Yang, the co founder of Pinterest, the co founder of Ripple Ahmad Mostak, the CEO of Stability AI, who has been one of the biggest proponents for for advancing AI, as well as a handful of scientists from companies like Google and Meta. Notably there is nobody that signed this from OpenAI. However, there are a handful of people that signed it directly from Microsoft. Also this week, Gary Marcus, a scientist and author of the book Rebooting AI, wrote a pretty well thought out article about the risks of AI. Now from his article he's not arguing that we're about to see Skynet and that AI is going to take over the world and try to destroy humans. He he's arguing that the risks of AI are things that we're already starting to see. He actually says I still don't think large language models have much to do with superintelligence or artificial general intelligence. He's not worried what happens if computers rapidly and radically self improve themselves, which he says I don't see as an immediate possibility. But he says you don't have to be super intelligent to create serious problems. He then goes on to talk about mai risk or mediocre AI that is unreliable. He talks about how with the help of large language models, there have been a lot of failures, phishing and online fraud issues, including this report here where somebody used a voice cloning tool to clone the voice of somebody who then called that person's parents and scammed them out of money, pretending to be the person that they changed their voice to. It says here a Houston, Texas couple was scammed out of $5,000 after thieves used artificial intelligence to clone their son's voice and tell the parents he was in trouble. The scammer disguised as their son told them that he had hit a pregnant woman with, was going to be jailed and was badly injured in the crash. This is just one example of the type of issue that he's referring to. As AI gets more prolific, he says we need to stop worrying just about Skynet and robots taking over the world and think a lot more about what criminals, including terrorists, might do with large language models and what, if anything, we might do to stop them. In this tweet here, Gary says phishing and online fraud can be created faster, much more authentically, and at significantly increased scale. Large language models can be abused to mislead potential victims into placing their trust in the hands of criminal actors. Large language models may facilitate terrorist activities. He also says perhaps coupled with mass AI generated propaganda. Large language model enhanced terrorism could in turn lead to nuclear war or the deliberate spread of pathogens. Worse than COVID 19. Many people could die. Civilizations could be utterly disrupted. Also in the news this week, the US Government is gearing up for an AI antitrust fight. They're worried about the big companies like Google and Microsoft and OpenAI all having too much power with these large language models in their hands. The FTC chair says, as you have machine learning that depends on huge amounts of data and also depends on huge amounts of storage, we need to be very vigilant to make sure that this is not just another site for the big companies becoming bigger and really squelching rivals. Also this month, Google put out a statement on their website talking about a shared agenda for responsible AI progress. Now, Google seemingly has been behind the race in AI right now, rolling out stuff slower than Microsoft and OpenAI and some of these other companies, but mostly due to their own concerns over ethical reasons and the harm that it could produce if they release these models. But now they find themselves thrown into this race where other companies like OpenAI and Microsoft are rushing products to the market, likely before they're really ready. And Google is getting a lot of flack for essentially trying to keep up with companies releasing products that probably weren't totally ready for public consumption yet. In this article, Google calls for shared industry standards and sound government policies that promote progress while reducing risks of abuse. Now, with all of these calls for pumping the brakes on AI from big corporations, government and billionaires, what is the general consensus among users of these tools? Well, prominent voice in the tech and AI scene, Robert Scoble asks this question. Let's say USA pauses its AI development. Will China know? Can we afford to fall six months behind? Here's a good counterargument from Ahmad, the CEO of Stability AI. He said, oh, I forgot to mention one thing. Nobody is launching runs bigger than GPT4 for six to nine months anyway. Why? Because it needs new H100 TPUV5 clusters anyway to get scale above, which are six to nine months away after being installed, burned in, optimized, et cetera. Good time for thinking. So he's essentially saying there's gonna be a pause for six to nine months anyway because the compute power isn't really there to go much further beyond GPT4 anyway. Brendan O', Donoghue, who's a research scientist at Deepmin, says, I don't think a pause makes sense. I would like to see a lot more investment in alignment and safety, research and engineering, though we need more practical AI alignment work and less waffling about Roko's, Basilisk and other similar fantasies. I personally took to Twitter and asked what people think and as you can see it was quite a hot button topic here. Silky Han from Heist Online agrees that it's time to slow down all your tech says Are other countries going to stop governments? I doubt it. We've already let the genie out of the bottle, scotty says. I think the only ones that are Beyond G are OpenAI currently, so it really is just a petition for them to pause while everyone else is free to continue. Now personally I agreed and said Sam Altman is notably not someone who signed the agreement. But as Ahmad also pointed out earlier today, OpenAI did put out a statement saying we think it's important that efforts like ours submit to independent audits before releasing new systems. At some point it may be important to get independent review before starting to train future systems and for the most advanced efforts to agree to limit the rate of growth of compute used for creating new models. We think public standards about when an AGI effort should stop a training run, decide a model is safe to release or pull a model from production use are important. Finally, we think it's important that major world governments have insight about training runs above a certain scale. So it actually does seem that the people over at OpenAI tend to agree with this idea that maybe we should slow down a little bit and that there should be some more government oversight into how all of this AI stuff is handled. Esteban here says, at best, it's naive and childish to think that an open letter will stop state nations and companies from pursuing more powerful models. At worst, it's disingenuous and a way to stop the competition from getting a stronger lead. Ivan Skellin here says, I think the motivations are as mixed as the pool of people signed, ranging from genuine concerns on one side to desire for slowing down the front runners in order to bargain catch up with them. I tend to think there's probably some truth to that now. Much of the replies on this Twitter thread are essentially saying the genie's out of the bottle, cat's out of the bag, there's no slowing this down. Some companies may try to stop the progress, but that's only going to give other companies time to catch up. And even if there is government involvement, the bad actors that want to use this technology for harmful use cases, well, regulation's probably not going to slow them down or stop them at all. Looking at what people say over on hacker news about all of this latest information. You tend to get responses like this the virtue signal of all virtue signals. We do not live in a society where an article like this will have any impact. This is merely an attempt to buy time because they're behind and want to have a chance to catch up. Work will progress whether they like it or not. Another response this would be totalitarian. This is arguing for a group of people to have the power to decide some field is unsafe as per some vague, unverifiable criteria, then set up a police structure to verify compliance, all outside of the safeguards of democratic or judicial norms. So it's a very, very hot button issue among the AI world right now. There's a lot of deb and I wanted to sum it up by going to our AI overlord and asking GPT what it thinks. So I asked what are some arguments for pausing the advancement of large language models? Number one Unintended consequences Pausing the development of large language models would allow researchers and policymakers to assess the potential unintended consequences and risks associated with their deployment, such as misinformation, fraud, and criminal misuse. 2 AI safety and Control A pause would provide time to develop and implement more robust safety measures and protocols to ensure that AI systems are safe, interpretable, transparent, and aligned with human values. 3. Incomplete Understanding Large language models are complex and often function as black boxes, making it difficult to predict their behavior. A pause could allow researchers to better understand these systems before advancing them further. Number four Ethical Concerns Pausing large language model development would provide time to address ethical concerns related to AI, such as fairness, accountability, and transparency, and to establish guidelines for responsible AI development and use. 5. Public input and policy Development A pause on LLMs advancement could give society and policymakers time to engage in a broader public debate about the desired outcomes and potential risks of AI and develop appropriate policies, regulations, and 6 preventing an AI Arms Race by pausing the development of large language models, AI labs could avoid an uncontrolled race towards increasingly powerful AI systems, which could exasperate risks and make it more challenging. Now, when asked why we shouldn't pause the advancements of these large language models, here was GPT4's response. Economic advantages Some argue that the development of large language models and AI in general can lead to significant economic benefits, including increased efficiency, cost savings, and the creation of new industries and job opportunities. 2. Technological progress the advancement of large language models may be seen as a natural progression in the field of AI, and pausing their development could hamper overall technological progress and innovation. 3 Global competitiveness and the global race for AI dominance. Some may argue that pausing the development of large language models could put a country or region at a competitive disadvantage compared to the others who continue their research and development. 4. Potential societal benefits LLMs have the potential to provide numerous societal benefits such as improving health care, education, and environmental sustainability. Pausing their development could delay the realization of these benefits 5. Addressing risks through regulation Instead of pausing the development of large language models, some experts may argue that the focus should be on creating and enforcing appropriate regulations and safety measures that mitigate the risks associated with AI and six Better AI safety Research the development and deployment of advanced AI models can help researchers better understand the challenges and risks associated with AI, allowing them to develop more effective safety measures. I let the current large language model champion debate itself on this issue. On one hand, we don't totally know all of the unintended consequences that may come from this technology, and we could probably use some time to really think things through and really figure out how do we put regulation and safety controls and safeguards on top of this technology. But on the other hand, this would sort of stall progress and what sort of advancements can we make in a shorter amount of time that could help in industries like health care and education and improving the environment? And who does get to make the rules and decide what is good for us and not good for us? And if some of the big companies that we're aware of and are comfortable with put pauses on it, who's behind the scenes? Who's not putting pauses on it? It's a very, very nuanced discussion and it's going to play out very interesting as governments get more involved as companies try to speed up the race while other companies try to slow down the progress. And it's fascinating to watch. Again, I'm kind of here to make silly YouTube videos and to make animations of monkeys on roller skates, but I did also kind of place myself right in the center of all the AI world. So I'm in the middle of all this too. I am really thinking through all of the repercussions and the philosophical and ethical elements of it. And I am being tagged on Twitter and shouted out on YouTube videos and being asked questions about where I stand on all of this stuff. And to be honest, I have very mixed emotions. It's nuanced. It's not a binary discussion. I see the benefits of saying, hey, let's slow this down, we're moving this a little bit too fast. But I also see the benefits of going but if we Slow down the progress. What are we missing out on? What are other people doing that aren't slowing down on it? What sort of technological advances could we be making if we didn't slow it down? And I don't have the answer. I. I think it's an interesting, fascinating discussion, and I can see both sides of the coin, and I'm just here playing around with the technology and trying to give people the TLDR of what's happening, but my opinions are very mixed. They're very nuanced. And I do see both sides of that coin. I do tend to lean a little bit more towards the side of continuing technological advancement, but I also see how that's really scary. The unknown of that is scary. The unintended consequences of that is scary. And I think it's okay to be on the side of advancing technology, but also being a little bit scared of where that might take us. And I think that's kind of where I stand right now. But I'm curious where you stand. I think the comments will be really interesting on this video because it's not me playing with Midjourney or making pictures of cats eating tacos. Anyway, that's what I have to say about this topic for right now. I'm sure I'll have more to say about it on future videos. If you want to keep yourself in the loop with what's going on in the AI space and get the TLDR every single Friday, come over to FutureTools IO, click on join the free newsletter. And every Friday I'll send you all the news articles that are relevant and hot in the AI space right now. I'll send you the coolest tools that I came across across, a handful of YouTube videos, and one new way to make money with AI. I'll send it every Friday. It's the TLDR of the Week. And if you're really interested in this topic of AI and you want the TLDR videos and you want to see some of the advancements and hear about some of the things that are happening in the space, make sure you like this video and subscribe to this channel, and I'll make sure stuff like this keeps on showing up in your feed. All right, thanks so much for watching. I really appreciate you. I'll see you guys in the next video. Bye, Sam.