The AI Daily Brief: Artificial Intelligence News and Analysis - 5 Questions Shaping the Near Future of GenAI

Starting point is 00:00:00 Today we're discussing five questions shaping the near-term future of generative AI. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes. Welcome back to the AI Daily Brief. We are definitely in a bit of a late summer lull when it comes to AI news. Enough so, in fact, that if you happen to see any accounts on Twitter or LinkedIn, trying to drop their posts about the last two days have been so insane in AI, you should instantly unfollow and, frankly, block those accounts.

Starting point is 00:00:37 But in any case, despite there being a little bit of a news slumber this week, there are still some really interesting things to discuss as it relates to where we are in the AI cycle. And so today, what we're going to be doing is discussing five questions shaping the near-term future of generative AI. And perhaps unsurprisingly, we are going to kick off with the question of regulation and more specifically SB 1047. We have been closely tracking the progress of this particular bill, and frankly, it's pretty hard to get a handle on exactly how this is going to play out. At first, the sides were simply the loud parts of the AI safety community on the one hand and the accelerationist on the other, but it's increasingly drawn in attention from all sorts

Starting point is 00:01:18 of other people as well. Nancy Pelosi and California's congressional delegation weighed in, sharing their opposition to the bill. But then Elon Musk came out and expressed his support for the bill, which some saw as unexpected and others saw as him trying to screw California with its pants on as he left the state. The latest person to weigh in, is San Francisco Mayor London Breed. In a short, simple letter, she wrote, I'm proud that in a few short years, San Francisco has become the AI capital of the world. 21 of the top 50 AI firms are located here, capturing over 70% of global venture capital funding for generative AI last year. The industry is core to our city's economy and future.

Starting point is 00:01:54 And this letter is written to Senator Scott Weiner, who's pushing the bill. I share your desire to ensure that AI develops in a safe and sustainable way, and I agree with the overall intent of your legislation. However, I believe that more work needs to be done to bring together industry government and community stakeholders before moving forward with a legislative solution that doesn't add unnecessary bureaucracy. With additional time and collaboration, I am confident that we can find a solution that addresses many of the concerns raised in recent months, while still enabling this emerging field to grow in a safe and sustainable way. As such, I unfortunately write in opposition to your bill SB 1047. Now, what's interesting about all of this is that it doesn't just

Starting point is 00:02:27 matter for California. One, SB 1047 is going to impact people because anyone who does business in California will be impacted, which is going to be a huge portion of the Gen AI field. But more than that, the debate is also drawing some lines around things that will shape future conversations. One of the big underlying themes is who gets to control what we're concerned about when it comes to AI safety. As I have mentioned in previous episodes, I believe that part of the challenge of this particular bill is the degree to which it's focused on extreme and catastrophic risks rather than current and contemporaneous risks. This makes sense, given its genesis, with a lot of folks for whom that is their primary concern, but it's also why a lot of people said that they didn't engage with the

Starting point is 00:03:04 process, because they fundamentally don't think that these things should be regulated. And that, of course, gets to a second underlying theme, what matters to regulate. Part of the debate we're having right now is not just about SB 1047, but about which aspects of AI should even be considered for regulation right now. We're also getting a chance to see where industry stands in specific, and in particular you've seen things like Open AI coming out in support of a different bill, seemingly to try to suggest that it's not about all legislation, its opposition is specifically to this legislation. It does seem as though we'll be on to the next phase of the conversation soon. Anna Tong from Reuters earlier today tweeted,

Starting point is 00:03:37 staff have requested that Scott Wiener's California AI bill SB 1047 be voted on today. Now, of course, a vote one way or another doesn't mean this thing is done, but it certainly will move us to the next step. Second question shaping the near future of generative AI is AI in a bubble? Specifically, is it in a Wall Street bubble? Aside from the regulatory discussion around SB 1047, this has perhaps been our most common topic over the course of the summer. The state of play was admirably summed up by the Washington Post in this piece from the end of July called Big Tech says AI is booming, Wall Street is starting to see a bubble. Now, on the bubble front, there have been a number of

Starting point is 00:04:12 different reports and blog posts that have been widely referenced. One of them is Goldman Sachs, Gen AI, Too Much Spend, Too Little Benefit. Another is AI $600 billion question, a blog post that was published by Sequoia. Now, neither of these are perhaps as negative as their titles might seem, But to the extent that they are negative, they both largely argue that the capital expenditures that have gone into and will go into the AI buildout are going to be very hard for companies to recoup. That's the $600 billion question. That's the amount of money that David Kahn from Sequoia estimates that companies would have to make to have made their spend with companies like Nvidia worth it. The flipside argument comes from venture capitalist Sarah Taville, who wrote a blog post where she argued that, quote, the next few years is going to make the $600 billion question look small.

Starting point is 00:04:56 effectively that if these companies are right, the value of what happens on the other side of this AI buildout is so immense that it dwarfs all of these considerations. Now, the interesting thing is that Sarah could be right, and yet in the short term, there could still be a correction. Wall Street is uncomfortably forced to play the role of venture capitalist in this field, trying to figure out how much the value of the future of AI is worth relative to the current considerations. One of the theories that I have around why this has been such a big conversation over the summer is the fact that we're heading into a rate-cutting cycle. I think that a big part of the reason that Wall Street has clung to the AI narrative for the last

Starting point is 00:05:32 couple of years is that it coincided with the Federal Reserve raising rates for a time at the fastest clip in about 40 years. AI was a thing to be excited about, and perhaps was held on to a little bit harder than it otherwise would have because it was the one bright spot in an otherwise gloomy period for markets. Now that the rate cycle is poised to go in the other direction, are we seeing Wall Street jettison some amount of the AI hype because they're simply able to now because they have a new theme again in the form of rate cuts and what they'll mean for liquidity and markets. Now, the implications are big. There is likely a trickle-down effect for startups in the space, particularly for the open AIs and anthropics of the world, which could need to fundraise again

Starting point is 00:06:09 soon. Today's episode is brought to you by Venice. Venice is a private, uncensored generative AI app. It accesses open source models to enable text, image, and code generation without the fear of being spied on or having your data exploited. Discuss anything with Venice without concern about it being monitored, sold, or given to advertisers and governments. Venice is different because your conversations and creations are kept securely within the browser, never stored or accessible by Venice. Unlike other AI apps, Venice won't tell you what's okay to say or not. Venice won't patronize you. It simply provides direct access to machine intelligence, no topics are off limits, no ideas or taboo. With Venice, you're in control of the AI as you should be.

Starting point is 00:06:47 Pro subscriptions are available for $49 a year or $8 per month. A.I. Daily Brief listeners receive a 20% discount on Venice Pro. Visit venice.a.I. slash NLW and enter the discount code NLW Daily Brief. That's NLW Daily Brief, all one word. Today's episode is brought to you by Super Intelligent, which is, of course, our platform that helps you learn how to use AI tools and perhaps even more importantly, gives you ideas on the best use cases that are actually going to help you achieve whatever it is you want to achieve. To recognize the end of summer and back to school slash back to work, we are running our best promotion ever when you sign up for Super Intelligent between now and the end of August using code so back, your first month will be 100% free. The platform features over 600

Starting point is 00:07:35 fun, highly practical AI tutorials that get you using AI fast and with an eye to actually transforming how you get things done. We've just launched Super for Teams. So if you have a group of people at your company that want to figure out how to use AI together, I highly suggest you check it out. But for those of you who are using Superintelligent as an individual, Once again, if you sign up for Super Intelligent between now and the end of the month using code so back, you will get your first month 100% free. Go to be super.aI and check it out today. The third question shaping the near future of AI is whether we're in a capability slowdown and whether, in fact, we've plateaued. It is notable to many people that for close to 15 months now, GBT4 has been the state of the art.

Starting point is 00:08:19 Now, yes, GPT4 has gotten a little bit better. GPT40 is different than what was released last spring. And of course, lots and lots of competitors have caught up to Open AI. You've got Claude 3.5 Sonnet right now. You've had Lama 3 catch up with an open source model that's basically at the state of the art. But there's still a big debate and a big question around whether we're really seeing a slowdown in progress and what it means for how much more we could wrench out of this particular AI architecture. This could have implications for things like regulation. If we are starting to get to a plateau, does that change the way that people think about how aggressively we need to regulate versus waiting and seeing what capabilities future models actually have.

Starting point is 00:08:56 Certainly, it seems to be impacting the foundation model companies themselves. Yesterday, we discussed new reports out of OpenAI around their attempt to launch their reasoning model strawberry, in part to help build out GPT5, which they are apparently codenaming Orion. The frontier of capabilities really does impact a lot in this space, and so this is a question that I think is worth watching closely. Fourth question shaping the near future of AI is around enterprise AI adoption and where we are in that cycle. I think the biggest takeaway here is that every enterprise basically in the world is trying to adopt AI and finding trouble doing so for a variety of reasons,

Starting point is 00:09:30 basically all reasons that justify the existence of superintelligent for what it's worth. Companies have bad visibility into what employees are doing with AI and bad strategies for gaining that visibility. They're not sure what type of upscaling or capabilities training their people need, and they're not sure how to translate the individual usage that's happening up into organization level change. This is the big theme of a recent McKinsey report called Gen AIs Next Infliction Point from employee experimentation to organizational transformation. This report had similar statistics to the earlier LinkedIn Microsoft report that we've referenced frequently that found that 75% of employees are using AI now,

Starting point is 00:10:05 but 78% of them are doing it on their own and not telling people about it. In a similar way, this study found that 91% of employees are using AI, but just a very small percentage of companies are actually implementing multiple AI workflows as a company. In other words, right now, there is, yes, personal productivity enhancement going on, but that's not translating up to actual organizational transformation. Now, this matters for a few reasons. First of all, Enterprise AI adoption is going to be a transmission engine for a huge amount of disruption. Most people still work inside big companies, and so how they do or don't adopt AI will impact people's work lives in a pretty dramatic way. There will also be a big impact for startups. Enterprise AI

Starting point is 00:10:42 adoption and the patterns therein will shape what type of markets and customer basis startups can go win, to say nothing of M&A and acquisitions as well. So I think the patterns and norms and expectations when it comes to enterprise AI adoption are actually much more significant than just in terms of what any given company is doing. As an aside, of course, if any of you companies out there are trying to figure out AI adoption, might I point you over to B-super.a.i slash teams, where you can learn all about the AI enablement network that we're building. Fifth on our list of questions shaping the near-term future of AI actually might not ultimately be all that near-term, is going to be significant, and that is, of course, the question of all of these numerous copyright lawsuits

Starting point is 00:11:22 when it comes to AI training. There are a million of these going on. Every week, there's some new update. For example, recently, courts declined to dismiss copyright infringement claims against companies including Mid-Journey and Staple Defusion, allowing the case to proceed. Then just last week, we got a new lawsuit from authors suing Anthropic for copyright infringement over AI training, and ultimately it's very clear that these lawsuits are going to end up at the highest levels. These are going be Supreme Court decisions. The results could very much impact future AI models. I guess the question for me is whether the genie is too out of the bottle by the time they come through, or whether the results will be dramatic enough that it actually makes people go back and correct and change models and take them

Starting point is 00:12:02 offline. What's more, I think that alongside the legal discussion, there is going to be a society level discussion around what the norm should be when it comes to AI training rights and remuneration. Like I said, this is a little bit of a fudge because I'm not sure how near term it is. But given that there is something in the news every week about the progress in these cases, I still thought it was worth noting. Now, one big bonus or honorable mention is the perpetual next thing in generative AI for the last year and a half, which is, of course, AI agents. There is a sense that AI agents could be a next frontier that once again fundamentally changes how we think about this technology, although for now they remain mostly in the realm of the potential. Still, it is worth keeping an eye

Starting point is 00:12:40 on this area because at some point they will no longer just be in the realm of the potential, and it could be very significant when that transition happens. So friends, those are five questions shaping the near future of Gen AI. Let me know what you think about this. Come join the discussion on Spotify or in the comments on YouTube. I'm interested to see what you think and whether I've missed anything. For now, though, that is going to do it for today's AI Daily Brief. Until next time, peace.

The AI Daily Brief: Artificial Intelligence News and Analysis - 5 Questions Shaping the Near Future of GenAI

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.