8 Questions on the Future of AI

0:00Today on the AI Breakdown, we're exploring 8 questions around the future of AI. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.network for more information about our YouTube, our Discord, and our newsletter.

0:24Hello, AI friends. Quick note before we dive in, this is a bit of a weird week. I had some very last-minute travel come up, and we are racing for a big announcement on Wednesday, the substance of which you've heard lots of ads about in the past, but which I will once again point you to besuper.ai to sign up to be the first to hear more. And so I had to pinch hit and do something different for today's show, lest I didn't deliver a show. And recently, Stanford University's Neil Ferguson has been doing a set of Q &A type posts in his Bloomberg column, where he asks big questions that he's had from either readers or from other people in his life, and then tries to give an answer.

0:58What I'm going to do is go through a bunch of those questions from those last couple columns, give a glimpse of his answer, and then share any additional thoughts that I have. We're starting with his most recent column, and the first question is, will AI have an adverse impact on the 2024 election? Neil's answer, it seems highly likely. Because US election campaigns have much larger campaign budgets than any other developed countries, each new communications technology is rapidly adopted by political entrepreneurs. LLMs, such as OpenAI's GPT-4, have immense political potential. They can generate vast quantities of plausible content with little human oversight.

1:29This includes fake phone calls and video clips. We've already seen the first fake robocalls, mimicking Biden's voice in the New Hampshire primary. It's hard to believe that the political use of AI will simply be prohibited. Then he goes on to speculate around what it would actually look like, and there it gets a little bit clearer. I'm roughly in the same camp here as Neil. I think that it seems very unlikely that the presence of AI and its capabilities will not have some impact. The question will be how the human immune response system adapts to the new reality. For example, Ferguson writes, Surveys show that even when voters are primed to be aware of deepfakes, they do not get better at identifying deepfakes, but they do lose trust in real videos.

2:06My instinct has always been that people are going to adapt a little bit more quickly than we think to the reality of deepfakes everywhere, but that doesn't mean that we'll be ready to handle what feels like it will inevitably be some onslaught of faked information with this cycle. So even if it doesn't tear apart the fabric of democracy, it's hard to imagine that it doesn't make things more complicated. Question number two, will AI be curbed by US regulation? Ferguson writes, last October, Biden issued an executive order detailing his administration's priorities for regulating AI. In one of its toothier security provisions, the order invokes the 1950 Defense Production Act to require companies developing advanced AI systems that could threaten national security to notify the federal government and report the results of safety tests or red teaming.

2:44The executive order does not regulate liability or require licensing for companies building AI models. The executive order's enforceability and implementation will therefore largely depend on federal agencies' enforcement and rulemaking, as well as judicial review and the willingness of tech companies to abide by the new structures. Ferguson also makes mention of the FTC's investigation to determine whether OpenAI engages in, quote, unfair and deceptive practices relating to privacy and data security. And he looks at the current state of comprehensive legislative proposals, pointing out that one of the leading contenders right now, the Bipartisan Framework for USAI Act, sponsored by Richard Blumenthal and Josh Hawley, rips out Section 230 of the Communications Decency Act and says that it does not apply to AI, meaning that AI companies would be liable for users using them to disseminate AI-generated content that is ultimately harmful.

3:27Now, Ferguson kind of demurs on the question that he poses, will AI be curbed by US regulation? He simply points out that it's unlikely that we're going to get very comprehensive legislation passed anytime soon. My answer is multi-part here. In the short term, I very much do not think that regulation is likely to curb AI. For one, I don't think that we're at the stage yet where this particular AI needs to be curbed by regulation. However, I do think a more interesting question is how the presence of AI potentially increases state power in the medium and long term. The types of things that the state would need to do to exert control over AI are fairly dramatic.

4:01It's why we see proposals for, for example, the government having complete oversight into who's using big amounts of cloud compute. That would be a fairly significant change from the norms of where we are now and represent an increase in state power. And so I think if you're interested in this question, it's worth zooming out and raising the stakes a little bit to see where things land. The next question in Ferguson's essay is fairly similar, Will Europe Succeed in Regulating AI? in which he discusses the EU's AI Act. Ferguson writes, The Europeans are looking to repeat what they pulled off for online privacy with the General Data Protection Regulation, or GDPR.

4:32Many non-EU countries adopted the stringent European regulatory norms within their home markets purely because they wanted to sell their products into EU markets. Since 2018, GDPR regulators have imposed$4.5 billion in fines, although US big tech firms have fought back with litigation. However, I doubt the Europeans will be able to set the standards for AI regulation. In 2021, Washington and Brussels founded the Trade and Technology Council. Within this framework, they have tried and failed to agree on a voluntary code of conduct on AI regulation. The big European problem is that it is home to hardly any major AI companies, with the notable exception of France's hugging face.

5:03So what do I think? I agree with his skepticism. I simply think that the incredible pressure to compete to have AI companies, to have sovereign clouds is going to way overwhelm anyone's interest in following European norms. Now, that said, there's lots of parts of the AI Act that aren't just about generative AI models that could become templates for other places. It's really the question of the frontier models and generative AI that I'm less sure on. Question number four, is there any prospect of a system of global governance? In this answer, Ferguson references a piece last year from Mustafa Suleiman, who was the co-founder of DeepMind, the co-founder of Inflection, and is now the head of Microsoft AI, and the Eurasia Group's Ian Bremmer, who offered basically an international model for AI cooperation.

5:45Ferguson says, I remain unpersuaded that AI can be regulated like finance. However, the Bremmer-Sulima model had two other elements. One was a body similar to the Intergovernmental Panel on Climate Change to ensure we have regular and rigorous assessments of AI's impact. The other was that, quote, Washington and Beijing should aim to create areas of commonality and even guardrails proposed and policed by a third party. Here, the monitoring and verification approaches often found in arms control regimes might be applied. While Ferguson acknowledges that comparisons to nuclear regimes in the Cold War aren't perfect, he also notes that it's where everyone lands.

6:16He says, An ideal global governance system would structure coordination between states to stop non-state actors and rogue states from developing or accessing cutting-edge AI models. Enforcement would work through a global export control regime for GPUs and a global know-your-customer protocol for cloud compute. Such a system is already being built, however, for geopolitical reasons, it is targeted at China. As China is the only other AI superpower, this makes little sense. In the last year of his life, Henry Kissinger attempted to establish a meaningful AI arms control dialogue between the US and China.

6:44History and recent events suggest that a global AI governance regime is very unlikely in the short or even medium term. We are a long way from the idea of artificial intelligence limitation talks. The arms race will therefore continue at the current breakneck pace. I think Ferguson is right to identify this as currently a bilateral issue, or at least an issue that only has two parties really, which is the US and China. I think strategy one for the US is going to be to get so far ahead that they can dictate terms to the rest of the world. And should that not work? Well, then we'll have to see. When it comes to a broader global cooperation regime, I don't even really want to share what I think it would take to have a catalyst that would bring players to the table for something like that.

7:34hyper practical, focused on getting you actually using AI tools in minutes, not hours, and certainly not days, based around video tutorials and companion challenges and projects that have step-by-step instructions that make it really easy to try out new AI platforms. And all of this is now culminating in the launch of a new platform, which we're hoping is the most practical and useful way to learn AI that anyone has yet created. If you want to be the first to know when that launches? Go to bsuper.ai, that's bsuper.ai, and sign up for the waitlist. I cannot wait to tell you more about it. Number five, can China catch up?

8:13Ferguson writes, why is China lagging? The simple answer is that it cannot manufacture the most sophisticated semiconductors itself, and the US is able to restrict its access to those produced by TSMC, as well as to the complex chipmaking machines produced by the Dutch firm ASML. Now, Ferguson does not believe that this means that the tech war has been won. He said, true, China is behind the US in AI spending in an AI company formation, but it is ahead in robots, and it is striving mightily to find ways to circumvent US restrictions. Nor is NVIDIA indifferent to China's insatiable appetite for its chips.

8:41The Economist is not alone in wondering if contraband NVIDIA chips are being smuggled to China via Singapore. Ferguson also points out that even the US dominance in AI talent is slipping a bit. He ends on the note, remember, the Soviet Union began the nuclear arms race far behind the US. It took two decades to catch up, but it did so. This question is at once political and technological. On a technological side, people with much better information could have a much more nuanced debate. But I actually think from a political perspective, there's only one answer that matters. And that answer is yes, theoretically, China could catch up.

9:11It simply does not make sense from any sort of real politic perspective, or to hell with some framework like that, any rational political perspective, to blithely assume that somehow it can't and to create policy around that assumption. So far, it seems like US policy has been dictated by the premise that yes, indeed, China could catch up, and I think that's what we're likely to see going forward as well. Number six, is AI really the new Manhattan Project? Ferguson begins with a discussion of the debate between Vinod Khosla and Marc Andreessen, where Khosla asked Andreessen, would you open source the Manhattan Project?

9:42This one is more serious for national security. We are in a tech economic war with China and AI that is a must win. This is where Ferguson's answers get very blurry. He concludes, in thinking about the likely uses of AI, we should remember that as a species, we have a track record. Yes, there are all kinds of wonderful uses to which AI can be put. The medical scientific possibilities are especially mind-blowing. But the history of nuclear fission suggests we should devote at least as much effort to developing AI's destructive potential. I will only point out in this one that even as Congress and the Senate hold learning sessions and hearings and all these sort of normal processes of a democratic society, the US military establishment is absolutely racing to adopt AI.

10:21And so I think indeed we already are devoting a fairly significant amount of effort to, as Ferguson puts it, developing AI's destructive potential. Number seven, how much energy is an AI world going to need? Ferguson writes, one big difference between nuclear fission and AI is that AI only consumes energy. But how much? Some alarmist commentators have projected that AI could end up needing close to a quarter of global electricity generation by 2030. However, the most thorough analysis I have yet read concludes that AI will propel data centers to use 4.5 % of global energy generation by 2030. That's still a lot, and because so much AI activity is concentrated in the U.S., data center-critical IT capacity will need to triple from 2023 to 2027, taking it from 4.5 % of U.S.

11:02power generation to 14.6%. In recent weeks, mainstream media in the U.S. has been waking up to what this implies, not least for the dream of reducing the share of electricity generated by natural gas and increasing the share generated from renewable sources. That dream is dead. So, holding aside the question of renewables and what percentage of U.S. need they can make up, It is absolutely the case that leaders in the AI space believe that the needs of AI are actually going to propel the next energy revolution. Sam Altman is the most vocal about this belief, and part of the reason that he's not concerned about how much energy AI uses in the future is that he simply thinks there's no way that we get to that future without having developed new energy sources.

11:408. What will AI mean for the future of war? Ferguson writes, The most questionable assertion in Andreessen's AI essay why AI will save the world was his claim that, quote, AI is going to improve warfare when it has to happen by reducing wartime death rates dramatically, because AI will help statesmen and commanders make much better strategic and tactical decisions, minimizing risk, error, and unnecessary bloodshed. Ferguson continues, I strongly suspect the opposite will be the case. In the coming AI wars, mortality rates in armed forces will be very, very high, precisely because AI will make the missiles and other weapons so much more accurate.

12:11This is already apparent in Ukraine, where drone warfare is gradually transitioning from remote human-piloted systems to AI-powered autonomous systems. The central problem of our time should be obvious. Without quite thinking it through, the US, in effect, outsourced manufacturing of the most advanced semiconductors to an island claimed by China. The main reason TSMC, which produces over 90 % of GPUs, has a price-earnings ratio of just 14.3, is its vulnerable location. To quote the Wall Street Journal, a US-China war over Taiwan will almost certainly result in the destruction of TSMC's fabs. This would set back the global chip supply chain by 5 to 10 years, derailing the AI boom in the process.

12:45It would be very, very nice if the US could build its own version of TSMC on US soil. That is the dream that helped to inspire the Subsidy-Packed Chips Act. Let's just say I am not holding my breath. So the thing that Ferguson is saying here is that the presence of TSMC and its significance to the global AI transformation increases the likelihood of a war around Taiwan. And on top of that, AI-powered autonomous systems are likely to kill more people because they're so accurate. On the first question, which isn't exactly written but which is implied, does the importance of AI increase the likelihood of a war over Taiwan?

13:15I think it certainly does, but I also think it's by no means inevitable. And on the question of whether AI makes the future of war one with reduced or increased deaths, I think it could quite obviously go either way. Part of the reason that it's so important for us to be engaged in these decisions is that we are going to have to set very new norms around everything from work expectations to the social contract to norms of war. It's not just a technology question that will shape the answers to this, but a societal question as well. All right, and that concludes the review of these eight questions from Ferguson's most recent piece.

13:47I think I'm actually going to save the other questions from the first piece for another episode, because this is a pretty fun way to get at some of these questions. Hope this was interesting to you. Appreciate you listening as always. And until next time, peace.

14:02Thank you.