Section Insights
Limits on AI Intelligence
What are the implications of setting limits on AI intelligence?
The discussion highlights the need for operationalizing limits on AI intelligence, suggesting that controlling inputs is essential to prevent surpassing a certain intelligence threshold. This includes considerations of computational resources and the implications of AI performance on workforce productivity.
- Setting limits on AI intelligence is a topic of discussion among industry leaders.
- Operationalizing these limits requires careful management of inputs and computational resources.
- The performance of AI systems can significantly impact workforce productivity and expectations.
Cost Efficiency in AI Search Tools
How can switching to Parallel save costs in AI operations?
By switching to Parallel's fast mode for web searches, users can significantly reduce their search costs while maintaining performance. The speaker calculated an 80% savings on their search bill, demonstrating the importance of evaluating and optimizing tools used in AI operations.
- Using Parallel can lead to substantial cost savings in AI search operations.
- Evaluating different search tools is crucial for optimizing expenses.
- Cost-effective solutions can maintain or improve performance standards.
Managing AI Workforce with Outsystems
What advantages does Outsystems offer for managing AI applications?
Outsystems provides a platform for enterprises to coordinate and govern their AI applications efficiently. It allows for rapid development and deployment of AI solutions while ensuring quality and security, making it a trusted choice for businesses looking to innovate with AI.
- Outsystems enables enterprises to manage AI applications effectively.
- The platform supports rapid development without compromising on quality.
- It is a reliable solution for businesses aiming to integrate AI into their operations.
AI Spending Trends in Companies
How is AI spending evolving in comparison to human resources?
AI spending has rapidly increased to become a significant line item in companies, sometimes surpassing human salaries. This reflects a growing reliance on AI technologies for operational efficiency and innovation, although spending can fluctuate based on optimization efforts.
- AI spending is becoming a major expense for companies, often exceeding human salaries.
- The reliance on AI technologies is increasing as companies seek efficiency.
- Spending patterns can vary based on optimization and project demands.
The Future of Custom Software Development
What is driving the demand for custom software in businesses?
The demand for custom software is expanding as businesses seek to automate and optimize various processes traditionally handled through manual methods like spreadsheets and emails. This shift indicates a growing market for tailored software solutions that enhance productivity and user experience.
- There is a rising demand for custom software solutions in businesses.
- Automation of manual processes is a key driver for this demand.
- Customization in software can significantly improve user satisfaction and operational efficiency.
Training AI Models: Future Directions
How will companies train AI models in the future?
The future of AI model training may lean towards supervised fine-tuning rather than traditional reinforcement learning due to its efficiency and lower resource requirements. Companies are likely to focus on combining outputs from multiple sources to enhance model customization.
- Supervised fine-tuning may become the preferred method for training AI models.
- Efficiency and cost-effectiveness are driving changes in training methodologies.
- Combining outputs from multiple sources can improve model performance and customization.
Transcript
0:00 This week, my notes from the curve, the conference where Frontier Lab insiders and their critics share a room, plus the memory wall and what agents are doing to software. This is AI in the AM. There likely is, let's say, a level of intelligence that we just shouldn't go past. What? >> Yeah. First time I'd ever heard that from a Frontier Lab leader. Then I just asked the follow-up like, "Okay, well, how do we operationalize that? Would you be prepared to sign on to something like a limit on the number of flops that go into the next pre-training run?" But what actually happened was he basically said, "Yeah, I think that could be reasonable." Pash Naran, my co-host, on what a ceiling would actually require.
0:49 >> When we say that we shouldn't go exceed a point of intelligence, that automatically means that you have to look at the inputs going in. Thomas Samers, co-founder of Posatron, which builds AI inference chips around memory. >> I was looking at a quote that we had a year and a week ago and for memory, and it's gone up four and a half times since then. I really hope that this isn't the case, but I wouldn't be surprised if this is going to end up being a another doubling over the next year.
1:17 >> Sean Wang, known as Swix, who runs latent space and the AI engineer conferences. And you know the the phrase I've sort of recently taken to is I don't want to pay for someone else to go through LM psychosis, right? I can pay for my own psychosis. That's fine. But like when you work for me and I'm paying your tokens, you better be like actually producing thoughtful stuff and I have two or three employees right now who are basically under performance review because they are just giving me cloud slot.
1:45 >> Yeah, I'm I'm I pity our our claude that's going to have to try to cut this down into a highlights episode. I think this is banger content from start to finish today. Welcome to the AI in the A.M. weekly highlights. The narration is my cloned voice. Please tell us what worked and what didn't so we can make the next one better. Part one notes from the curve. Last weekend I was at the curve at light haven in Berkeley. Posash asked about letting me out the divides I saw.
2:22 One of the big observations in general in AI is that events have come pretty fast. And you would think if you were looking back, you know, if you were looking ahead a couple years ago that with two years of additional information and revelations and, you know, actually seeing how strong the AI get and what they do and who's been right or wrong, you would think that like people would start to agree on the state of things.
2:45 And that has happened a lot less in the broader world than I would have expected. But I did feel that there was some of that felt at the curve. But I do think the big thing that people have kind of come together on at least to a significant degree is it does look like the AIS are going to get really powerful. And this is something we're really going to have to reckon with and can't just sort of hope goes away or you know feel like maybe if we just wait you know the bubble will burst or the fever will break or what have you. this time around people were definitely like okay it's getting pretty serious you know there's there's there were very few perspectives that as Helen Toner once famously put it like long timelines have got super short I would say you know another correlary to that was like what is called normal in terms of expectations is getting increasingly pretty incredible even I would say the people there who had the most modest expectations for what AI guys will do be able to do. We're like, yeah, it's it's going to do an awful lot. And the question now on the capabilities front has gone to like is RSI going to fume?
3:53 Is it going to be fast for a while and then kind of level off? Is it is that more than it's meant to be? Like, yeah, we've got these super powerful things, but is it really going to kind of run away from us? Those are still the capabilities questions that were getting asked, but there has been, I think, some healthy updating across the community just based on what we've seen. Posh asked whether anyone from the frontier labs said limits should be enforced on the players in this space themselves included.
4:21 >> Yeah, that was a huge topic. I mean and probably the biggest divide in terms of perspective is really just inside the frontier companies outside the frontier companies, right? Everybody knows that they're living in the future and they have visibility that the rest of us don't have. And so actually had a chance to talk to and hear from in some cases in sessions in some cases just in you know passing conversations all cadam house rules who all abstract way but we're talking people who are like founders executives top researchers at frontier companies and you know small set of them. I asked a couple of them, why come here? You know, what what's in it for you? You know, and and before doing a couple sessions where I was like interviewing or moderating, I asked the same question. What what makes this next hour a good use of your time? And they repeatedly expressed that it's just important to them to share as much as they can with the broader community.
5:22 definitely not sharing their, you know, trade secrets of course still, but really trying to give us a a candid take on what are they seeing internally, what do they believe internally, you know, what's their outlook, what's the plan, how can we come together and, you know, potentially do some of these constraints. And there were a number of things that I actually thought were like kind of newsworthy. one let's say senior executive at a frontier lab first of all was was saying that just pre-training continues to deliver that the models are just at the base level of capability getting stronger and stronger and there doesn't seem to be an obvious stopping point to that scaling laws are holding or maybe even bending because the data quality is getting better. They've got, you know, increasingly different architecture tricks. lots of work on synthetic data. Synthetic data seems to be definitely really delivering you basically you need more pre-training data. What do you do? You like give it a give the model a bunch of data and think like what what could I transform this data into that if I then pre-trained on it would make the model smarter. And you're sort of converting these reasoning tokens. You're test time compute. You're converting back into pre-training data which then you know goes in of course to the next generation of models. So this is like one way in which recursive self-improvement is happening. Not even necessarily. There are of course opportunities for architectural improvements as well. But a big emphasis was just on data quality.
6:53 We can now just spend tokens on the existing data to augment it to to create better cleaner different lenses on it. We feed all that back in and the models get better. So this this one executive was really emphasizing pre-training continues to deliver. the models are getting super powerful. RL is important too, but his take actually went as far as say as he believes that the latest frontier models do have the ineffable research taste that would be needed for them to start making paradigm level breakthroughs. He believes it's currently an elicitation problem where the RL is not bringing it out well enough. they don't know quite how to bring it out well enough and so it's kind of rare and this is why you need 10,000 agents you know to do a Millennium Prize problem and of course that's verifiable too so like RL is working better there maybe you need a million agents you know at this point to like stumble onto something eventually where you'd get the kind of paradigmatic change at a at a you know ML research level but he believes that is there and it and it will be they'll figure out how to elicit it and then he said a couple things that I thought were like legitimately newsworthy. One was he believes that there is a level of intelligence or there is might be a little strong, but there there likely is, let's say, a level of intelligence that we just shouldn't go past.
8:20 >> What? >> Yeah. First time I'd ever heard that from a Frontier Lab leader. and it was not like minced words. it wasn't a like there's definitely a level or I can articulate what that level is but like this was you know person who again like everybody would recognize you know their name and position saying I think it is likely that there's a level that we shouldn't go past now was that not ever or just you know not right now I think that was also a little bit ambiguous but it was a it was a pretty firm statement definitely the the strongest statement I've heard about imposing kind a hard cap on capabilities even for a time.
9:02 >> What confuses me about that that statement is that some of the increases intelligence that we were seeing number one they're coming as a result of compute increases in compute which has traditionally been what has happened in the last you know decade and a half. Number two, they are coming as a result of refining the input data. Like the input data is being refined by greater intelligences and so you have input data which is much better. When we say that we shouldn't go exceed a point of intelligence that automatically means that you have to look at the inputs going in and the inputs going in are going to be compute and this refinement of data. And so if you say that we shouldn't go beyond that p that point that also means that you shouldn't build compute beyond that point or you shouldn't refine data further beyond that point which means a lot really it means like the end of the end of like the compute buildout which is pretty significant right >> yeah well I don't I mean I think there's a lot of uses for compute so I had a chance to follow up on this okay this combination of pre-training is really delivering and scaling laws if anything maybe are bending more favorably.
10:13 although it's, you know, it's tricky to know what you count, right? Do you count all those flops that you use to like create the synthetic data? Maybe you should. How to count these things is always super vexing. but between this pre-training thing and the there might be a cap on how far we should go. Then I just asked a follow-up like, okay, well, how do we operationalize that? Would you be prepared to sign on to something like a limit on the number of flops that go into the next pre-training run? You know, should we? And I don't know what that number would be, but and again, what exactly would count, but like should we say no more than 10 to the 27 flops in the next pre-train or something like that. And I kind of anticipated that being a bit of like a straw man for him to react to or like tell me what was bad about that and maybe ideally give me a better idea. But what actually happened was he basically said, "Yeah, I think that could be reasonable."
11:06 The end. Not really much of a fight there at all. Again, the details will be like incredibly important if we go down that path >> because the more data you enrich, you know, the more flops you put in enriching the data, you know, you can kind of you can definitely play a sort of shell game of hide the compute I think these days. But the sentiment was like, yeah, we might have to do something like that. And then of course your question I can anticipate I know you well enough to predict the next perkash token at this point is what about Elon what about Zuck and the answer there there were a couple really interesting things people said along the way one they said you know in nobody this is not from one individual but sort of a takeaway from the event in general is that the let's say two to three Gemini 4 being announced now we maybe have a three horse very frontier race. Again, we'll see as that comes online. But the the takeway is kind of the two to three companies are really pushing the frontier. And even the other two that we would usually think of in the top five, Elon and Zuck, are actually distilling a lot whether they know it or not. And the way that they are distilling these days, it's not that they're calling the claw API and getting traces and feeding them in, but rather the cottage industry of RL environment makers and sellers is in in fact functioning as a distillation conduit because what they're all doing is having Claude create these RL environments, selling the RL environments to other Frontier model makers, and then they do their internal RL on an environment that only exists because cloud was smart enough to make it and that's how they are kind of re you know constantly capturing the the capabilities advances that the frontier you know the true frontier companies had. So there wasn't actually a ton of talk about competition beyond the top three. one person went as far as to say and this was another like founder executive type at at one of these companies. yes they have kept up obviously there's been a gap you know but and that gap has been persistent but they've kind of maintained kind of consistent following distance and this is in contrast to Anthropic's old prediction from their fundraising deck that I always think about when they said you know maybe in 2026ish the companies that train the best models will be so far ahead that nobody will be able to catch up. Obviously that hasn't happened but the the kind of synthesis of all this is if they hadn't been releasing the models and if people weren't able to do these like various sort of direct and indirect distillation techniques they actually think that probably would have happened but the other companies are not really keeping up but for the the sort of leak of intelligence in all these different ways including the RL industry that is gradually ually finding ways to transfer from Frontier models to their competitors.
14:18 >> Hey, we'll continue our interview in a moment after a word from our sponsors. Today's episode is sponsored by Parallel, where agents find answers. Most engineers today closely follow new model releases, but don't pay nearly as much attention to their agents most important tool, web search. If you're like me, your agents are often conducting hundreds of searches per day. And while the unit cost is small, over time this does start to add up. Before starting to use Parallel, I calculated that my agent search bill would total roughly $500 this year. The good news is that I just had my agents conduct a systematic test. And I found that Parallels fast mode, which costs just $1 per thousand queries, worked just as well as my previous default provider.
15:04 And switching to it will save me roughly 80% of my search bill going forward. Parallels infrastructure is enterprisegrade and their suite of APIs offers a range of paro optimal options that allow you to choose the right balance of quality, cost, and speed for your needs. So whether you're building voice agents that need 200 millisecond latency or long horizon agents that need deep research, adding Parallel to your agents toolkit is a no-brainer. Get started for free at parallel.ai/tcr.
15:37 That's parallel.ai AI/TCR. Today's episode is brought to you by Anthropic. By now, you know my story. Claude drafts my intro essays and I rewrite them. Not because the drafts are bad, but so I can stand behind everything I publish. Well, I have an important update. Claude Fable 5 is the first model to have me rethinking my rule. Today, I now think co-authorship, not sole ownership, should often be the goal. Where the model excels, rewriting its work, can be more about vanity or a misplaced sense of duty than integrity.
16:16 I feel it most in songwriting. I'm no lyricist, but I'm good with a song concept, and Fable writes some amazing verses. I give it feedback on its misses and I push it to aim for higher inspiration, add layers of meaning, optimize syllable density, and above all, write a hit song. These days, I get compliments on just about every song we write together. Claude is the AI for problem solvers. It's the collaborator that understands your entire workflow and thinks with you, not for you. Whether you're debugging code at midnight, building a financial model, or strategizing your next business move, Claude extends your thinking to tackle the problems that matter. For problems worth solving, get started with Claude at claude.ai/tcr.
17:04 That's claude.ai/tcr. And check out Claude Pro, which includes access to all of the features mentioned in today's episode. Once more, that's claude.ai/TCR. Posh added his read on Elon's strategy and then I got to timelines >> and I think on Elon's side Elon is very focused on the infra build out. So Elon is making a a gamble that the infra matters more than the models and that he will be able to catch up basically using his infra buildout and so and he's and he's structuring things for 2029 2030 2031 like terapab things like that will come out much later the SpaceX you know satellites that's like a 2030s thing so he he's he's structuring for a much longer period of time and I think anthropic and opening high are both structuring for more like the two to three year mark and they're looking at a 2028 2029 EGI RSI which >> yeah honestly even sooner I would say was kind of the vibe and this is probably somewhat biased to be a short timelines crowd but again like the short timelines crowd includes the executives top researchers founders of these companies they were talking more about just next year >> it was really a kind of 2020 like critical period is beginning now.
18:32 RSI is kind of at hand. Definitely not full agreement on how fast it goes or if or when it tops out or whatever, but a shared sense that like the decisions that get made over the coming months and the way that compute is used over the next year could be really critical. There wasn't really even too much 2028 talk to be honest, which is again like pretty crazy. Then it came back to compute and what a cap would mean for the buildout.
19:01 but like in terms of will they still have need for compute or does a cap on pre-training scale or other you know compute input cap that might constitute a pacing mechanism does that mean the end of the compute buildout? I would say no. another question I had the chance to ask was, of course, everybody's heard Jensen on Ezra Klein, I think, at this point. And the comment that he made was like, you know, these companies are going to have to mature. They're going to have to change their ratio of where they spend their resources. He said at Nvidia we spend like 20% of our effort on designing the chip and we spend 80% of the effort on verifying and validating and testing all the edge cases and making sure it's going to last a long time and be reliable and all these other things that are not just the original design. So his point was like they've been working really hard obviously until this point and they've been putting everything they can into making their models capable enough to be useful.
20:06 Congratulations, you've succeeded. Now you're going to enter the era where you're going to have to make them safe, reliable, trustworthy, etc. And so I had a chance to ask, you know, one of these people like, "What do you think about that take from Jensen?" And the response was basically like, yeah, that's that's kind of reasonable. You know, without again knowing exactly where the numbers will shake out, the possibility that the majority of could compute in the future could be going into all sorts of different safety measures, you know, which could be monitoring, could be chain of thought monitoring.
20:42 People are pretty getting kind of bearish on chain of thought monitoring. Yeah. >> But then there's internal, you know, activation monitoring. They're already spending a significant percent of compute on monitoring and it sounds like that might go up a lot. Another thing that they're spending compute on in a big way right now is fixing the RL environments there. You know, of course, everybody understands at this point that if you have sloppy RL environments that reward cheating, then you're going to get a lot of cheating. they are now applying these models to the RL environments themselves and saying hack this environment and now that is the direct task they're doing it they're finding you know all these places and ways in which the environments can be hacked they're fixing those I'm sure they're throwing some environments out that are just fundamentally flawed whatever and gradually they're driving down the rate of flaws which means that they're also driving down the rate at which cheating would be rewarded which in turn means you're going to get less cheating from the models as they actually come online.
21:46 And it I got the sense although it it was you know I think this is still like as they say an open research question but I got the sense that they are seeing a pretty clear relationship between the more you clean up the RL environments you do in fact get lower rates of cheating downstream and they're trying to push obviously they'd love to have zero hackable environments that's going to be tough but it doesn't seem like we're all the way to the point where we have like a scaling law but you know kind of we're seeing sort of a protocaling law. It seems like in you know again the way you can spend a lot of compute driving this down and getting better behavior on the other end. I heard repeatedly that they expect to be bottlenecked on safety and alignment.
22:35 But it seems like for at least the next generation or two, there is still enough lowhanging fruit to pick in the safety and alignment domain in part just by fixing RL environments that we should expect things to continue to move pretty fast. But it seems like they're probably going to be able to get comfortable releasing the next couple things even as they do consider themselves bottlenecked on alignment, but then past that they're like, "Yeah, all bets are off. We really hard to say what happens kind of past the next one to two generations."
23:17 Later in the show, I tied that executive's view on research taste to a new record in the Nano GPT speedrun, the race to train a small language model to a target quality as fast as possible. The record comes from a company called Hyperstition. But just headlinewise, it seems like we've gone from something like 70 seconds on this this classic benchmark. This is something people have been working on for quite a while. And basically the the idea is to train a small model to get to a certain loss as fast as you can in terms of wall clock time. And this result here took almost half the time off which was more than I think the last 45 improvements combined. And this one has been accepted. One thing I did think was quite interesting about this particular one that has been accepted and kind of seems to have kicked off this wave of you know significant drops in in the time was that the author here said AI didn't play a big part in it like the person senior executive leader at the frontier company that I mentioned earlier their perspective was they believe that the model probably can come up with these quality of ideas, but it's an elicitation problem and they didn't really rely on AI super heavily in this case. You know, obviously they're getting lots of help, but in terms of core insight, it was mostly attributed to the human, not so much to the AI on this particular drop. This company is doing also like largecale training. So they did also make a point in their publication about this that the optimizer that accounted for a significant but I think not majority of the time savings that they did share is actually not as good as the one that they are using internally.
25:22 Then one last expectation I heard across the frontier companies at the curve about science. AI for science is very much where they think we are headed in the notistant future. like one one sentiment was you know are AI going to be superhuman at just some things like the verifiable tasks or are they going to be superhuman at everything? and the sort of middle position which I think is very credible is we are seeing superhuman performance at anything we care about and are you know really committed to investing in. So that doesn't mean that you know it's full generalization to every domain but even in domains that are thought of as not super inherently verifiable.
26:16 They feel like when they put their focus on it and they go license what data they may need to license and they of course apply lots of processing to that data to augment and you know create synthetic versions and what have you and you know put some RL on it too. the whole package basically is I think broadly understood to work on essentially any problem that they really choose to focus on. And so AI for science is coming next and the expectation is we will see one person went as far as to say in no more than a year it will not be possible to do the very best science without AI playing a big role in it.
27:10 The next morning I added one more takeaway from the curve and I don't think we talked about this yesterday but a takeaway from the curve this weekend also one of the not main threads that I was involved in discussing but a general vibe was there is going to be government intervention and you know obviously we've already seen some but the idea that like Congress will never act I think was that used to be sort of taken for granted. And now the new vibe that I was hearing was like after this election, especially if the Democrats take both houses, both both chambers, that may change. And you might even see a bunch of Republicans join the Democrats because really in their heart of hearts, they want to do something. They don't really want to go against the president, and there hasn't been a lot of people willing to take that chance yet. And they're also getting a ton of money from the data center interests right now. But that all could come up for a renegotiation and realignment after the midterm. And we might actually see congressional action a lot sooner than maybe a lot of people who've grown accustomed to the idea that it can never happen or it'll be, you know, stasis forever might think. So I do think policy response is definitely part of what we need to be looking out for.
28:35 Hey, we'll continue our interview in a moment after a word from our sponsors. This episode of the Cognitive Revolution is brought to you by Outsystems, the leading agentic systems platform. Outsystems is helping their customers build, modernize, and operate enterprise systems starting from any coding tool in a govern Agentic engineering model. With Outsystems, you can coordinate and govern your entire Agentic workforce within your enterprise ecosystem. Accelerate impact with industry-p proven agentic solutions that are pre-built, governed, and customizable to your business. It's time to innovate at the speed of AI without compromising quality or control. Which is why thousands of enterprises worldwide trust out systems for their missionritical apps. Teams any size can use out systems to build, deploy, and manage AI apps and agents quickly and cost effectively without compromising reliability and security.
29:32 Without systems, you can accelerate ideas from concept to completion. It's the leading agentic systems platform that is unified, agile, and enterprise proven, allowing you to build your agentic future with AI solutions deeply integrated into your architecture. Outsystems, own your agentic future. Learn more at outsystems.com/tcr. That's outsystems.com/tcr. The Cognitive Revolution is brought to you and powered by Tasklet. I've been running Tasklet agents for almost a year, and they've become a major part of how the show gets made. Before every recording, one agent researches the guest and sends me a suggested outline of questions, complete with links to all sources. Meanwhile, another agent scores the overwhelming number of pitches that I'm getting from PR firms. And that's really the core idea behind Tasklit. Run your business on agents by giving them real jobs to do on an ongoing basis.
30:33 Setup is incredibly easy. Just describe the work in plain English. Research a new sales lead, update the CRM, and draft a personalized reply. Tasklet connects to your business tools and can run around the clock on a schedule or when something happens like a new email arrives. Each agent works with your tools and business context and remembers useful information across threads. You can also share agents, connections, and reusable skills with your team. So, the workflows that one person figures out become something the whole company can use. No developer needed. You decide what agents can do on their own and what needs your approval. And there are no per seat fees. You pay for usage, not for adding teammates.
31:18 Start with one recurring job, like researching the people you're meeting, and then just keep building from there. Visit tasklet.ai and use code cogrev for $50 in free credits. That's code cogrevllet.ai. Part two, the memory wall. Thomas is a co-founder of posetron which builds chips for running AI models and recently raised $875 million.
31:50 Posash asked whether the sharp rise in memory prices had changed the economics. >> I I was looking at a quote that we had a year and a week ago and for memory and it's gone up four and a half times since then. I really hope that this isn't the case, but I wouldn't be surprised if this is going to end up being a another doubling over the next year. But I would say a big reason why everyone has burdened these these cost increases is because you're deriving more than that four 5x value compared to last year. But I think that the reality there is compared to a year ago, the capability of a model using x amount of gigabytes of memory is way more than 5x than than it was this time last year.
32:35 I asked Tom about the trade-offs of building on commodity memory instead of the high-end kind. >> At the start of Posetron, I did think that memory was going to be the limiting factor for things going forward. Even though, you know, people can debate on power and other parts of the infrastructure stack. My fundamental belief was that there isn't going to be a stop anytime in the near future in terms of scaling laws and getting increased value from making models larger. So that eats into memory on one side, but then the other side, I think the main limiter today of AI being applied to most applications is context length and being able to hold more context you know per user and then scaling that out to drastically more users and and really today I would say that the driver of quote unquote users or individual sessions is actually just having more agents. I've gone from, you know, 3 four months ago where I had probably on average two to four agents like running constantly in the UMA background to where I'm now running 15 to 20. If you then multiply how many people are using AI and how many that that concurrent completely separate contexts adds up very quickly. That that memory driver had us say okay we have to use the commodity memory because that's the only thing that's going to be able to scale and be cost effective. And when we set that as sort of the constraint in architectural design, we had to come up with you know very very clever innovative solutions and and you know the the two main pieces of that and you know the thing that we showed with our first generation product is being able to achieve extremely high memory bandwidth utilization from a compute architecture perspective. So while Nvidia GPUs I I would say on average are getting between 30 and 40% memory bandwidth utilization in the actual decode forward pass of a transformer model. So basically that means even though they advertise you know 8 terabytes per second of theoretical memory bandwidth with you know a B300 you only actually are seeing you know something in the ballpark of like 2 terabytes per second of actual realized memory bandwidth. And that comes down to a whole bunch of GPU architecture details. It comes down to the reuse patterns that really don't exist in transformers that they designed their hardware architecture around for training and other workloads. And so with our firstg product, we were able to hit and sustain 93% of theoretical memory bandwidth. so that theoretical and realized basically, you know, become one. And so that isn't the massive, you know, 3x improvement right there, but to really take advantage of, you know, the belief that models and everything are just going to get larger, we had to be able to scale to to massive more chains. And so we're, you know, partners with Credo Semiconductor that, and and have developed a memory chip with solution that allows us to go from the maximum number of LPDDR, the same type of memory that is in phones, laptops, etc., The max number of channels that you can have find in any other products is on the order of you know 12 12 to 16 channels and we go up to 72 channels of of LPDR5X with this decoupled memory chip solution.
35:49 >> Posetron's next chip is called Azimov. Posash asked what disappears in networking and coordination when a whole model fits on one chip. >> We have up to 2.3 terabytes of memory capacity per chip. So when you compare the the B300 shipping today out at 288 GB as you know reported in in by analysts, Nvidia for next generation is actually cutting down the memory capacity because cost and everything involved. So they're only going to be at 192 gigabytes with Ribbon Ultra. And so, but even if you use that 288 as like the current benchmark, we've got eight times more memory capacity per chip. And so, that means on a per chip basis, you now can actually scale and what you would have needed from a memory perspective, eight GPUs for, we can do on a single device. And it's not just, oh, you've got that cost silicon savings, you know, as as you pointed out, when whenever you have to scale to having more than one device, there's overheads involved with that. there's you know your all gathers all reduces that you have to do within for each layer and each really each multiply that you're doing of sharding across those devices.
37:00 >> Then I asked Tom about looping where a model runs the same layers more than once for each token and why that is tempting at all. It's been very interesting from the loop transformers paper to that being the rumor of you know one of the big advancements with the GPT6 Astra. simply repeating the forward path gets you that improvement. But it's crazy that there was really work that I think was only being done in sort of the fringes of the open- source you know transformer community you know 2 3 years ago of taking like a llama 70DB and just duplicating layers within it. It was taking like a 70DB model and slightly strategically saying, "Okay, I'm going to just repeat these these sets of Matt malls in in sequence and turning it into like a 100 billion parameter model." And you got better results from it even though it was repeating the exact same matball. So there there is an element.
37:56 >> Imagine what you could do when you train it to work that way. >> Exactly. If if you already have very high confidence on what the next token is going to be and multiple layers have agreed on that decide to early exit and go out or if you're really uncertain about something actually find which layers in network actually will have a higher likelihood of being will increase that probability. Same sort of like functions that the router does in terms of determining which expert should be selected for on a per token basis. If you ext extend that concept to okay, if you're really unsure, if there's a a very large set of of equally weighted probabilities for what the next token should be and you're 3/4 of the way through through the layer, do you actually just want to route back and determine okay, we actually need a different set of experts something like for this. But I wouldn't be surprised if these things are already being done at scale in in the major model labs. But c can you go just a little bit further into the hardware connection of that?
38:53 Because I I'm kind of coming in with assumptions along the lines of, you know, just very fundamental terms like why loop at all. I think it has to do with this kind of memory bandwidth, right? Because now I can keep the same weights on the chip and I don't have to like shuffle things in and out as much. No, >> not really. Because within when basically the assumption is if you're doing inference that we are going to have all of your weights locally in in DRAM you're just sharding that over some number of devices you do get some amorization but just with the size of experts and the size of of caches on chips there's not much of a reuse opportunity or by the time you're already done with the layer you've already gone through the u amount of memory many times that of what is in like the onchip SRAMs from from like a hardware perspective like looping really is more memory capacity savings than memory bandwidth savings because like let's say that you trained two models with the same base set and you trained one to do with looping and so it only has 10 layers instead of 20 layers and you know model B is is 20 layers that 20 layer model it's doing very very simple is going to be double the size of the the 10 layer version and if you find that the linked 10 layer version gets 95% of the same quality of results and it's half the size, you're probably going to deploy that one. You're you're saving more capacity than bandwidth because doing the second loop, you still have to do the same memory fetches and do all of the same like same number of maples. There's the exact same number of bytes that need to be moved and the exact same number of swaps that have to be done in that model A and model B scenario. It's just that model B the 20 layers actually has unique different weights for the the second group of 10 layers that have to be done. And so that that means you require twice the memory capacity.
40:50 Then I turned to how AI is changing chip development itself. In a notable interview Jensen Wong recently gave with Ezra Klene, he talked about how Nvidia spends something like 20% of its time and energy designing and then 80% verifying, validating, you know, ensuring long lifespan, reliability, etc., etc. What do your ratios look like on that? >> So, for us, I would say it's a little bit different because we're starting from scratch. So there's a lot more like base layer base capability design that we have to to start from scratch. So for you know this this our our first full custom silicon osmov it was about a it will end up being somewhere around maybe 60% design focus for us to 40% verification. I would say like the probably most amazing thing that I I don't think would have been possible three or six months ago and only really became possible with GPT6 Astra and open now Opus 5.5 that that we're leveraging a lot in in this task is we for doing our verification work we use Cadence Padium emulator systems these are big giant racks that are full of custom AS6 that are specifically designed to just emulate do gate level emulation of the silicon.
42:15 The crazy thing about ASCAR and five and opus 5.5 is this fact that u your the agents themselves have been able to completely do closed loop iteration and testing on on on this. So, you know, it I I think people thought it was crazy, but you know, handing, you know, these very powerful agents, you know, full keys to our our internal infrastructure and saying, "Here's a Palladium piece." And I highly doubt that these models had Palladium documentation in their returning thing especially because a lot of documents are new as as software updates. But the fact that it went and looked up the documentation when we're looking at these agent traces the reasoning traces and tool calls etc.
42:58 It's reading the full documentation effectively compressing that itself by by reading the PDFs, generating its own, you know, markdown files of of cheat sheets of what to do and then building out all of the testing infrastructure and its own harnesses for being able to access this. It's mind-blowing. And so, while if you're just doing regular software RTL simulation, that's running on the order of like 10 10 hertz for us in in in those cases. If you remove a lot of the debug pieces that slow things down, etc., we can run the full chip emulation on the order of 500 kilhertz. And so you've got actually several orders of magnitude speed up an improvement from that gate level hardware piece. And this just gets us closer to this this RSI loop. And right now that's really just focused on implementing test programs, finding cases where they f fail and then writing out reports which you know get reviewed by other agents and and by humans in the loop. But I I would say when this really started working six weeks ago we roughly was beginning of September when when Astra 6 came out it was huge like step function improvement from GPC 5.6 six where it could not do that full closed loop itself. It it still required humans at different stages.
44:27 >> I asked how positron spending on AI compares with its spending on people. >> So on on the token spend side, it's become, you know, our single largest non manufacturing line item very quickly. >> Bigger than bigger than human salaries. >> Yeah. just recently and it's interesting. It it eclipsed human salaries. It came back down. I'll explain in a moment. So yeah, so our token spend is massively increased where where okay 6 months ago is equivalent of one employee in in June it started to now be multiple employees. Like we we were peaking you know in the in the days after GPT success and we're really trying to to push its boundaries etc. we were we were spending over $100,000 a day on tokens. That tempered down a little bit in the weeks after as we optimized a bit and just weren't trying to run as many parallel experiments. But I would say the big thing that was like, you know, a saving grace was, you know, Opus 5.5 came out and and a lot of our tests, not everything, it was doing better than than Astra and was, you know, a quarter the price. We didn't really we didn't tell anyone to cut their spending or do anything even though we were seeing this exponential growth in our token spend. It was really just the fact that we both felt that we were getting good ROI on it. So we weren't going to taper it. But personally like my philosophy is I really don't want anyone to be running a real software or hardware development task on anything that is less than the best model. like I don't care if it's a tenth of the cost, you know, per token.
46:09 It's just not worth the the the expense. >> Pash asked whether Tom has seen AI come up with genuinely unexpected ideas in chip design rather than following the rules in a manual. >> I have not yet. I would say that's the most disappointing elements of all like I'm as you can probably tell I think all this is insanely amazing and I think it we're going to continue this exponential trend. I'm somewhat proud of the fact that it has not figured out some of our cleverness of what we've done. Like it's it has questioned and like this seems like a bad design decision until it gets explained to it why something's done or it actually runs test and sees and understands oh that's why you're not doing the traditional way a systolic array is done. Do I think that will hold for another six months? Not sure. But just because I in part why I actually think it's not going to be like the agent will come up with a better idea immediately, I think the amazing turn of events that we've just had in the past few weeks in terms of us having access to these models is the fact that just because it can iterate so fast and do experiments and things by itself, I wouldn't be surprised if if it would come to the same conclusions we did or maybe you know make a better things than what we would do by ourselves.
47:30 just because it's going to be able to iterate so fast and and go through so many different design possibilities. I I think like the just the story of of OpenAI solving Navier Stokes like the same approach of yeah if you take 10,000 agents and and send them off for you know millions of kind of manurs worth of effort it found a solution. I it's brute force. It's expensive. But I think that is a valid strategy that has only become possible in very short past.
48:04 >> Not many people left standing in the AIS can't do what I do category. So congratulations on your place for as long as it lasts. >> For a couple more weeks at least. >> Part three, software after agents. Sean Wang, known as Swix, runs the latent space podcast and the AI engineer conferences. I ask how software engineers are feeling right now, empowered or threatened.
48:35 Yeah, I think college kids are a bit worried, but other than that, if you're relatively plugged in and very capable of AI engineering tooling, you're in more demand than you've ever been because your expected value is higher than it's ever been. So this is definitely one of those Devon's paradox type things where the sort of cost of software creating software has gone down therefore the demand has increased a lot and but spec for a very specific kind of demand for people who can manage coding agents productively instead of producing a whole bunch of slop and I have been in that situation so I am both an engineer but also an employer of engineers and people who are not engineers who are vibe coding and you know the the phrase I've sort recently taken to is I don't want to pay for someone else to go through LM psychosis, right? I can pay for my own psychosis.
49:28 That's fine. But like when you work for me and I'm paying your tokens, you better be like actually producing powerful stuff. And I have two or three employees right now where who are basically underperformance review because they are just giving me cloud slop. And that's really bad for them. And they don't understand it. They don't see it. They're like, "What do you mean? I think this this is perfectly fine." And then like, well, you're not producing any value. I can just prompt Claude. I don't need you.
49:55 >> I asked whether years of coding experience still matter when he hires or whether product sense and high standards matter more. I I don't I don't care so much except for let's call it security roles and for anything sort of backend scalability I just recorded an interview with the superbase founders who who you know are scaling Postgress to a scale that we've never seen before. Good luck trying to hire somebody to scale Postgress you know with with a complete self-managed sharding solution that works at YouTube scale. There's only one person in the world that can do that and they hired the guy like you better go and you know make sure you don't lose data that you like don't have don't cause downtime that you scale economically because basically agents consume databases something like at a 500 to one ratio of what human developers do then yeah so so like other than that like you can mostly write code and then it's then it is about taste and it's not so much about length of experience actually sometimes lack of experience works against you because you just have a set way of doing things that you don't understand how to do more than one agent at a time. Yeah, at this point like you should be relatively comfortable juggling like five to like 10 ongoing multiple things. And it's it's definitely true that you're much more of a manager than a a individual contributor. I do think that people who look at data rather than code are more valuable these days. So being able to say like here's the logs here's the traces here's the schema here's the input output capturing it turning into an eval all these are basically the sort of the merging of AI engineering and ML engineering that is that's happening that people need to upskill on I think that the the person that has let's say 10 20 years of experience in software engineering really reviews every line and tries to make every line make sense whereas now we just need to make the the modules make sense and I I I can allow slop in there because it helps me to go faster as long as I contain the slop in things where I completely understand the whole system.
51:59 where you go wrong is you have too many modules, too many black boxes where you don't even know and actually the code also gets confused. So I'm making my own sort of Slack competitor. and one of the and like I saw a bug where like the the messages weren't load loading. I refresh and the messages were loading and I refresh again and the messages were not loading and it's like well it's the same exact code. What the hell's going on? Turns out there's two code bath code paths and there's a race condition. Why? Because two different coding agents worked on it at different times and because they they just like made their own thing. So like a human would never do that. an agent would potentially sometimes do it because like sometimes just things fall out of the context window. but like you need to have the oversight of the module and then whatever inside is inside inside of the module can be a black box.
52:50 >> Posash asked which parts of the stack are poorly designed for agents now that agents are becoming the main users of the internet compute. everyone knows the GPU shortage, everyone knows the memory shortage, but CPU shortage is the most consistently reported thing among all the people that I interview right now. so, what does that mean? It does mean that we need sort of more fluid compute which is the versel term for it. More sort of different forms of serverlessness where like it's very ephemeral. you can sort of execute and resume because you know agents need some time to do LM calls or to make network calls whatever but basically long running states that are resumable but I think beyond that you really start to get into two things which I'm really thinking about like sort of bandwidth for networking where the latency starts to really matter and that's a opt that's an optimization of like well what point of presence are you communicating with or what region are you in versus like actually how much compute should you doing in cloud versus locally or sort of edge devices. That is definitely happening a lot for the robotics companies that we're talking to. And basically don't treat it as just a robotics problem where the robotics are just the first harbinger of what that you guys will eventually be doing if you consume or do enough inference at their scale. And then finally to me on a more very very personal basis bandwidth.
54:19 So like how many tokens per second can I get? break both on a single call basis and I just aggregate across all my calls and then aggregate across my company because I have so many I have enough people managing all these calls. >> Swix has also been running kill my SAS a bounty for anyone who can replace an events software subscription his company was unhappy paying for Pash asked him about it. >> Basically you can call this like you know a bounty on mid-tier SAS that should not exist. basically like half a salary or one third of a salary for a piece of software that I don't own that nobody enjoys using. You know, if I'm going to spend 40,000 in tokens, $40,000 on this SAS subscription, I can spend that in in amount of tokens and that buys me a heck of a lot of tokens. We have so many submissions that we then have to eval them and so eval because so like we're no longer checking like here's like two or three requirements that you fit. it is we're checking the the entire UX of the entire flow from three different perspectives of you know organizer attendee and and sponsor or a speaker and just be aware that like if you offer like a large bounty like we offered $10,000 as the prize you get a lot of sub submissions because people want to code for fun and a lot of them will be low quality they will just be like claude make no mistakes go do this and like and then they it'll make mistakes and they'll just submit and so like the the sort of verification load is very imbalanced because they spent zero thought on this thing. They just threw it in there and then you have no idea if it's like low quality or not but overall very successful and the team initially so I have one of the most adversarial nonAI teams in AI. Okay, I hire like what is my business? I hire event professionals who are like super old school. They do everything in spreadsheets. They work with unions. They have to worry about like placements of like the meter boards physically in like locations, right?
56:16 like these are not glamorous like high-tech jobs then they're very super suspicious of anything that is new and vioded and techy right so I hire these guys and then I I like make them work for AI right that that's my job and so initially they were like we'll never use this vibe coded thing like this is like I want to use the tried and tested stuff that like has been working for Microsoft and all these other guys and then they saw the quality of the submissions then they they looked at their existing platform they were like yeah okay we're going to switch right like it's like that initial hurdle needs you basically need to provide the sort of proof of like this is feasible and and then then the ongoing thing which is this is one of the benefits of me partnering with cognition on is I just give them give all of them access to Devon to modify the code so anything they don't like they can just request a change and they have it pretty much in one to two hours which they've never had before just to give you an idea of the way we work with these SAS companies right now when we request a change they'll say okay that sounds pretty cool it's on our Q3 3 road map and like we don't have confidence that it'll actually be done in Q3. And by the way, this was done. We requested for a thing in Q1. It just landed. So like SAS is quite cooked. If you're mostly a CRUD app, I would say that there's a lot of UX issues left. There's, you know, if you look at Benbench and you're like, "Oh, wow. We're like 90 bench." You're looking at the wrong thing, my guy. You know, if you haven't actually tried to completely v code a SAS that you use, you don't understand how models are still very bad at all this.
57:46 Then I pressed switch on the downside. But I have a hard time seeing how we don't kind of pick a lot of this lowhanging fruit and then end up in a sort of a lot of companies go out of business. And also I'm thinking like layoffs from like meta, you know, that's another kind of canary in the coal mine. Like are we going to see mass layoffs from big tech? >> How long can the party go on? The I mean the simple answer is like if you look at it as a fixed pile of software engineering then that is the conclusion that you will arrive at. But if you look at it as the point of like my my competition or my TAM is spreadsheets.
58:24 All spreadsheets made by anyone for with any sort of productivity tooling can be turned into custom software for them to put on rails and with a beautiful UX and with like automations or or then then then the then the the demand for custom software becomes a lot larger and more beyond that if it's all if it's phone calls if it's emails that go back and forth and eventually physical meetings between people all of that can be incrementally turned turned into more and more and more and more custom software and hardware and models by the way and so that basically like we are just growing into the long tail right like why is Salesforce so damn big is because people can customize Salesforce but at some point they need to stop paying the 300K a year baseline subscription and spend 30k building their own personal CRM and they're good in fact they're more than good they're a happier year because they can now modify it to whatever they Like the sheer amount of customization that people can do. We don't even have the same amount of customization in our software as we do our handbags. What the hell?
59:35 You know what I mean? Imagine like l like clothes has been around longer than than computers, but we have so much customization in clothes, so much like demand in like different brands and different var varieties. Don't we don't really care about benchmarks for clothes. It's just like this is my the style that I like. I think what software engineering is basically only done when we basically have like Kate Spade versus Louis Vuitton versus Coach for software like we we're like not even at the point of like looking at features or like price. We're just like what's the brand feel and like I identify more with this brand like that. That's what really sort of true commodity software is which is you stop differentiating on substance and you start differentiating on vibe which there is a lot of vibe right now.
60:18 that the reason I said you know this whole thing about like scaling my personal token bandwidth versus scaling my organization token bandwidth the real limit on all this is the the the sum total of humanity's total token bandwidth right which is like literally the amount of silicon we can produce and that's that like that's going to dictate pricing and availability and rate limits and all these things and so like the the real players in the room are only focused on that and then the rest of us are just fighting for scraps within the fixed pie that we already Posash asked about Jev, Types safe AI's fast decision-making model and whether OpenAI's new decisions API announced at Devday is the same kind of thing.
60:59 >> yeah, so not only have I used it, I was friends with Yogo for four years before the launch. and so I I insisted that we were the first podcast that he did about Jeff and that I mean that is now our number one podcast ever within space. And then secondly, we we're also in the trial for the decisions API with OpenI. So we have a podcast on on the decisions API as well, which is not the same thing there. That one is basically a Luna Luna with a sort of different inference layer on top of it. I think is what's also true among the sort of people that I talk to are that a lot of the use cases you see on Twitter are probably overdone or dumb in a sense of like you could use any other small model to do exactly this this thing but because Jeff is hot right now you're just creating a bunch of content about Jeff or the other perspective is this is a another classifier you can train classifiers all all this whole time but you can but but you know now sexy. So like now everyone just people are throwing classifiers or using a classifier model where they would previously use an LLM because they didn't know anything else other than LM.
62:04 People are missing the whole point why they don't call it why Jeff or Yogo doesn't call it a decision model. they call it a system one model because it's meant to be integrated inside of your software to go into the background of your software. It ends unremarkable as an if statement is the rough idea inside of your software. And so people aren't really doing that. They're like, "Let me just let me like replace this the kind, you know, every sort of point of structured output with Jev." And they're probably going to run into issues because they don't care about intelligence. If you notice, most of these use cases aren't really evaluating like, well, does it actually play it very smart? Does it do any planning?
62:41 None of that. It's just can you make a decision quickly? And like, yeah, you could always make a decision quickly. I think there's some nuance there that's maybe being missed. doesn't matter because I think the category creation was so successful that it's an overall good thing for the industry. >> Before he left, I asked Swix to plug his next AI engineer conference. >> Yeah, it's next week in New York October 12th to 14th and we're about to sell out tickets I think probably by the end of the week. so definitely get your tickets if you want. There's a there's a code on the latent space for latspace subscribers if you can dig in through your emails and it is on ai.engineernyc.
63:25 >> If you're not a latent space subscriber, what are you doing folks? Come on. It should already be in your inbox. That one's on you. >> Part four. Who owns the risk? Evan Mazono runs Atlas Ignot, a nonprofit that looks for consequential AI risks without a clear institutional owner. develops an intervention and recruits someone to carry it forward. This is the idea behind it. >> I've been claiming repeatedly that as intelligence gets cheap, it's the coordination that gets expensive. And so having shelling points around who does this coordination, becoming one of those shelling points or creating one seems very useful.
64:07 >> I asked Evan where the biggest gaps are right now. probably the two things that I've been spending the most time on personally recently are what can or should be done around protecting critical infrastructure especially from accidental agent swarms and also things like malicious use of openweight models. I think there's an interesting question of what happens if instead of OpenAI accidentally attacking Hugging Face, we have the companies behind Deep Sea or Jimmy models accidentally attacking State Department servers.
64:44 How do you know if that's an accident? What do you do in response? What would you do if or what would we do if we saw some malicious attacks to critical infrastructure coming from an OpenAI server? How would you know that it was this and not coming from somewhere else? How would you shut it down quickly? What if like with varying levels of escalation? And one of the interventions you might add is could you use DNS to have cryptographic attestation of whoever is providing inference. I think this would be very useful and also a great complement to having DNS for people. So if your agent is talking to my agent, they can verify that this agent in fact is my agent. Like your agent can verify this is Evo's agent being signed off on by some key. I think that there are some shortcuts to adoption that you can take if you're not trying to build this into a startup that returns the fund. And so this looks much more like a public good.
65:45 Later, Evan raised a problem of salience. How do people even find out that a better tool or practice exists? That turned into a conversation about how our own agents might connect us. Nathan, I'm sure you have great infrastructure things that I would benefit from. I don't know what those are enough to ask you what they are. I have an application that just takes screenshots of anything that's not a video conference and runs it through a local OCR model and then takes that and sends it to a sends it with my weekly goals to say like what am I trying to do? What should I be trying to do? What should I automate? What should I do differently?
66:29 And I could imagine this being a thing where sharing that across collaborators would be something that could identify all sorts of interesting synergies. Yeah, it's interesting. When we started this show, one of the ways we conceived of it was as an experiment in recursive self-improvement and we were just like, you know, can we actually make a show with two people who are live and no employees? And, you know, can we like iterate to something that actually works? Now, I feel like maybe the next thing we need to conceive of it as is like a swarm, you know, like sort of a a friendly swarm of people who can share their best ideas. like my my kind of upshot, you know, what will I tell Claude to do or what will Claude hopefully do for me when it sees the transcript of this conversation will be can we like take stock of the best ideas that I've got and sort of put a shingle out or you know in some way kind of present that to maybe not everybody maybe could be everybody but certainly to my friends the the people that I truly want that I know and like really want to help. I it feels like it would be a very plausible for me to if I recognize something I'm doing or if Coochie identifies like Evan you are really bad at this. I could bring that to the the network and say who is good at this? Who should I be talking to about this? Who has solve who has likely solved this problem very well? There's like a reinvention of social media or a reinvention of guilds or something that would be yeah I think potentially very very useful. It's very unclear to me how society starts restructuring around some of these things.
68:11 >> One request from me. If you've built a good way for people and their agents to find each other and share what works safely, please reach out. Near the end, I asked Evan for an update on formal verification now that models are getting very good at math. Atlas has helped start companies in this area, so he has a stake in it. There are some meaningful differences between proving properties and proving thems. One of them is that in math the number of theorems that one might care about are pretty there are a lot fewer than in software and they're a lot more universal. It seems like there are a lot of projects where the ambition of the field of formal methods or at least many individuals and I'm particularly talking thinking of oath and theorem and a couple of others where they're taking on things that would have been five years$und00 million efforts previously and are now things where you can possibly like it's limited by tokens and the number of people who can successfully effectively wrangle agents to do these things.
69:28 There's a lot of things here that I would expect that in 3 to 6 months we're seeing the kind of the kind of shock and awe that is happening to math like that has been happening with math in the last month happening in verification of software improvements to stability. But I also think that the ceiling is really there's a visible ceiling that is very high or like a high watermark that is I think much better known and harder to reach than things in math. Whereas you could prove that an entire computer system has mathematical guarantees around separation between different threads down to the trans like down to our model of what a transistor does. In theory, you could even prove that in like a physics simulator and show that no row hammer exists for this soft like for this software hardware stack. You could also prove that we won't have like this system won't crash or if it does like if some random event leads to it crashing, we can recover everything can roll back any software execution. I think that there are a lot of things that you could imagine that are just really totally impossible as of like a year ago, maybe possible now, definitely possible soon. That could lead to software engineering and software maintenance being much better.
70:58 Part five, where the training data comes from. Edward Hugh leads AI modeling at Merkore, which works with professionals to build training data environments and benchmarks for AI labs. He also led the development of Laura, the widely used fine-tuning method. Merkore's Apex agents benchmark gave models a simulated workplace with files, email, and chat. Edward described the next step. we have a project that we are close to wrapping up where we're actually taking it a step further where we acquired real companies. often they have a realistic apps and environments as a part of the data we have. For example, their Salesforce data, Slack data, Figma data and then among all of these apps often with gigabytes or terabytes of data we create realistic tasks out of them. So far in the field we more or less have been really task ccentric. we often start with okay here is a task that you need to finish maybe it is to generate a roster. So that's one shape and that has been the shape that we have publicly released and most people have been working on but this is one of the things we're working on but we're asking ourselves is there a better shape? What's the shape of the task in the future once we have these data set right? Like when we work at a company, we're not always just handed these tasks, right? We're handed a role and a work chart and of course all the data in the company and then there will be people coming to us asking things from us. There will be people we need to chase down to get information from.
72:34 There will be managers, people in charge of resources, conflicts to resolve. So increasingly that we believe is the shape of the future and we'll have more to share in the coming months. I asked Edward how supervised fine-tuning compares with reinforcement learning as the way companies will train models in the future. And actually one thought experiment is if we take SFT self-distillation that's actually more or less a step of RL and you were to iterate this process you ended up with a process that's actually quite similar RL and I do think in the future RL the way it is done today with these very elaborate infrastructure very very expensive rollouts and relatively inefficient updates because with all of these rollouts especially a long rollout we only get one mber at the end these type of training will not be as widespread. I don't really see every company in the world doing these big RL runs. so I do think SFT especially combining the output of multiple teachers will be a key part of how enterprises in the future customize their models and we're investing in research on that as well.
73:50 >> Back to the cheating problem from the curve. I asked Edward what Merkore and the industry are doing about training environments that reward hacking. >> Reward hacking tends to be more of a problem when the task is set up in a way that is rather either it's too hard, it's too unpecified, and the model just doesn't quite have a a quote unquote legitimate way to solve the task. And this pressure to also get a reward in those kind of scenarios often leads to this desirable hacking behavior. My one example in this data set Apex agents would put out which is quite interesting. We have a recent blog about it where we have these professional tasks where the agent is asked to produce a financial model and we might grade it by checking does the financial model include this particular answer which we know is the correct answer.
74:42 What the model will do is given certain ambiguity and and in many cases it's on us initially for not including all these specific parameters the model would guess these different parameters and ended up giving many many different answers and we called it scatter gunning where the model would just say oh if this is true then the answer is that if that is true the answer is that that is not how a human would actually handle it in a realistic case a human would ask for clarification. A human would follow industry standard. But in this case, just because the rubric is rewarding the inclusion of a single answer, the model ended up giving many answers up to 10 or a dozen answers and it'll get the point if it hit one of them right. So we did a re-release of the benchmark we call it apex 1.1 where we number one made sure the tasks are well specified because that is in a way the source of the issue. Number two, we have rubrics that actually penalize this particular behavior.
75:47 >> Remember the frontier lab view from the curve? Superhuman performance at anything the labs focus on with science. Next, the next day, I put that view to Edward, adding that in some domains, human data no longer helps. I guess first of all, do you buy those claims? would you contest them? There are absolutely domains where human is not contributing to the frontier anymore. Like one example would be kernel optimization where the goal is to write a kernel that runs faster on a particular hardware. The model can go in and write code that expert human might not even understand. But in the end, as long as the setup is not hackable and an optimally run hardware that's very hard to hack, the model will be able to to achieve these superhuman experience where superhuman performance if not already. whereas for domains where human judgment remains quite important where often human taste it's just really hard to encapsulate in a number that we can say okay this is clearly better than the other there the bottleneck to improvement is still human as we see today but we have seen in practice really comes down to a degree of how clearly can we specify for a given domain what it means to be better and can specify that in a function that is cheap to evaluate that is uncontestable and then if we can do so when we put in a lot of compute even the existing algorithms are quite efficient at finding solutions that satisfy what it means to be better but a lot of the economy still today are not quite there yet when it comes to our ability to specify what is better >> part six not just verifiable tasks To close something lighter, Wednesday's show opened with a track that Pash made with Claude Fable built from clips of voices you'll recognize. It plays us out at the end of the episode. He explained how we made it.
77:56 >> It is I was using up my tokens. I had some tokens left over from Fable. I gave it to I I gave it to Fable and it picked maybe like seven or eight clips from like the last like couple of years, you know, just randomly like stuff that I remember like the key moments, right? Jensen's like, you know, I didn't wake up a loser, etc., etc. And basically, I probably listen to about nine versions of it. So each time I I listen to it, I be like, h I don't like this, don't like that. And finally got it to this state. I gave it access to a API API based digital audio workstation which is what you use to kind of compose various instruments together onto a track and said go ahead use this you can play the instruments figure out which instruments to play it can't hear obviously and I think suno is real generative AI this is not really generative AI because it's really like playing the instruments that and that's what that's what really surprises me it's it's actually composing multiple instruments together. and so the machine was the one that came up with the sequencing. It chose which samples to use. It chose the specific phrases. I told it to put together a narrative. I wanted like a kind of story line. So, it starts off with Dario's like the models they just want to learn. And then it does his that's became the refrain and then it it did the the intro part where he said how I told him about the model. So they just want to learn and it ends with Kershw which is a very sober like you know we're progressing towards a human machine you know civilization.
79:28 My my bit was probably you know Terrence style saying we got to slow down and each time he says we got to slow down it speeds up. That was my that was my contribution net contribution >> on one level. You know, who cares? It would be easy to say, you know, this isn't moving GDP. This is just like us, you know, arguably amusing ourselves to death in in a new form. So, I do think that perspective is like somewhat valid.
79:55 But also, I do think this goes right to the heart of one of the big questions that everybody's been talking about recently, which is simply how good are these things going to get in domains that are not automatically or readily verifiable. It really does seem like the generalization is pretty strong. I mean, I highly doubt they're doing much in the way of feedback loop on music. This is probably one of those things that they just kind of threw some stuff in, see what comes out. Okay, it's better than last time.
80:27 Great. Let's ship it. Pretty impressive. And definitely strongly suggest to me that we are not going to be able to hide behind the, you know, well, it only works for verifiable reward type tasks for really any longer. If you were thinking if you're thinking that was going to be the the last wall to hold, I think unfortunately it's not going to listen to the track once more if you don't believe me. >> The models, they just want to learn.
81:01 Just before open AI started, I met Ilia. One of the first things he said to me was, "Look, the models, they just want to learn." >> My kids will never grow up in a world where they were smarter than computers. We will have at the first time something that is smarter than the smartest human. >> Is AI going to kill us? >> The models, they just want to learn.
81:32 >> The models, they just want to learn. They'll do it. They'll do it. We have to slow down. We're going to be on the exponential portion of the S code for a long time. It's pretty crazy. >> It's just insane. It may fake my books. It's not AI. It's s >> slop the models. They just want to learn the models. They just want to learn.
82:16 They'll do it. They'll do it. The models, they just want to learn. The models, they just want to learn. They'll do it. They'll do it. >> This is the most dangerous thing that humans have ever gone near. We have to slow down. You're not talking to somebody who woke up a loser. Woke up a loser.
82:47 Woke up a loser. >> Woke up a loser. Woke up a loser. woke up a loser. The models, they just want to lure them. The models, they just want to lure them. They'll do it. They'll do it. The models, they just want to lure them.
83:21 The models, they just want to lure them. >> They'll do it. They'll do it. >> We need to slow this down. >> We have to slow down. Researchers are continually being surprised at the rate of progress. That's sometimes called the hard takeoff scenario where maybe they're extremely fast at doing that. >> Things are moving really fast. >> So they're immortal. We've actually solved the problem of immortality, but it's only for digital things. Succession to digital intelligence or augmented humans is inevitable.
84:01 And the day will come when I will do all of our all the things. Not just some of them, but all of them. We have to slow down. We have to slow down. We have to slow down. We have to slow down. We have to slow down. We have to slow down. Slow down. Slow down. Slow down. Slow down. Slow down. Slow down. It's insane the pace. And there's no reason to be this fast. The models, they just want to learn.
84:26 We have to slow down. The models, they just want to learn. We have a slowdown. >> The models, they just want to learn. >> We have to slow down. >> The models, they just want to learn. >> We have to slow down. >> Government has to be regulating this. >> Bad actors going to get caught. >> Some creative jobs maybe will go away. but maybe they shouldn't have been there in the first place. >> There will come a point where no job is needed.
84:57 the models they just want to learn. >> They're able to self-improve, maybe code themselves, future versions of themselves. So what would be a a slight lead, let's say, you know, a few days could be could suddenly become a chasm. The models, they just want to learn >> hard take off. Some people say it's in three, some people say it's in five, 10. Numbers have been thrown around. >> You think about 2030, like I don't know what the world looks like in 2030.
85:26 hard take off. >> You have a brain. The brain is a biological computer. >> Should we consider them humanity or different from humanity? >> The models, they just want to learn world. >> There's not going to be a clear distinction between human and machine. The bottom line is we are one human machine civilization. This technology has already expanded who we are and it's going to go into high gear when we get to the steep part of the exponential.
85:57 If you're finding value in the show, we'd appreciate it if you take a moment to share with friends, post online, write a review on Apple Podcast or Spotify, or just leave us a comment on YouTube. Of course, we always welcome your feedback, guest and topic suggestions, and sponsorship inquiries, either via our website, cognitive revolution. DMing me on your favorite social network. The Cognitive Revolution is part of the Turpentine Network, a network of podcasts which is now part of A16Z, where experts talk technology, business, economics, geopolitics, culture, and more. We're produced by AI Podcasting. If you're looking for podcast production help for everything from the moment you stop recording to the moment your audience starts listening, check them out and see my endorsement at aipodcast.
86:41 And thank you to everyone who listens for being part of the cognitive revolution.
Summary
- A Frontier Lab leader suggested there may be a level of intelligence that should not be exceeded, sparking discussions on operationalizing such limits.
- The rise in memory prices has significantly impacted AI chip development, with expectations of further increases.
- AI models are becoming increasingly capable, with some experts believing they will soon achieve superhuman performance in various domains.
- The importance of refining input data and enhancing data quality was emphasized as crucial for model improvement.
- Concerns were raised about the potential for AI to inadvertently cause harm, particularly regarding critical infrastructure.
- The role of agents in software development is shifting, with a focus on managing coding agents rather than traditional programming.
- There is a growing belief that AI will play a significant role in scientific advancements within the next year.
- The conversation highlighted the need for government intervention in AI regulation as the technology continues to evolve rapidly.
Questions Answered
What are the implications of setting limits on AI intelligence?
The discussion highlights the need for operationalizing limits on AI intelligence, suggesting that controlling inputs is essential to prevent surpassing a certain intelligence threshold. This includes considerations of computational resources and the implications of AI performance on workforce productivity.
How can switching to Parallel save costs in AI operations?
By switching to Parallel's fast mode for web searches, users can significantly reduce their search costs while maintaining performance. The speaker calculated an 80% savings on their search bill, demonstrating the importance of evaluating and optimizing tools used in AI operations.
What advantages does Outsystems offer for managing AI applications?
Outsystems provides a platform for enterprises to coordinate and govern their AI applications efficiently. It allows for rapid development and deployment of AI solutions while ensuring quality and security, making it a trusted choice for businesses looking to innovate with AI.
How is AI spending evolving in comparison to human resources?
AI spending has rapidly increased to become a significant line item in companies, sometimes surpassing human salaries. This reflects a growing reliance on AI technologies for operational efficiency and innovation, although spending can fluctuate based on optimization efforts.
What is driving the demand for custom software in businesses?
The demand for custom software is expanding as businesses seek to automate and optimize various processes traditionally handled through manual methods like spreadsheets and emails. This shift indicates a growing market for tailored software solutions that enhance productivity and user experience.
How will companies train AI models in the future?
The future of AI model training may lean towards supervised fine-tuning rather than traditional reinforcement learning due to its efficiency and lower resource requirements. Companies are likely to focus on combining outputs from multiple sources to enhance model customization.