AI model

“I personally am pretty skeptical. The reason is multifold. One is, I just think The technology and the research here is evolving so quickly that I think the number of data points you need to train a model is decreasing at, you know, some fixed efficiency level…”

“It's good at coding, but not great. I think the next few versions will be really great at coding, which is also where potential problems might come from.”

“The vast majority of companies are going to figure out actually how to apply these models that other people have built that are offered by APIs or in the open source or fine tune them or like build some other part of the stack.”

“I think the biggest thing is, in my opinion, all companies will have their own AI models. Like all companies will have their chat GPT or their GPT-IV.”

“It's not about more data, it's about better data.”

“The reality is no models you that are out today will be used in a year.”

“There is no such thing as an unbiased model.”

“So we are, in my opinion, going into a transformation of bundling when you'll see better, more efficient experiences done by one centralized brain, which is the AI model itself.”

“If you're trying to access information, the models should try to be as factual and objective as possible. And I think that kind of puts aside the notion of being politically correct.”

“The reality is no models that are out today will be used in a year.”

“Absolutely agree with that. You know, will you drive the car you drive today in 10 years? I don't think so. And 10 years in car industry is like one week in machine learning gravity zone. And so there is, of course, there is so much progress so fast That I don…”

“It's going to be lots of models at different parts different layers of this pyramid being used for different kinds of applications.”

“Turns out you can have moat other than your backend AI model, right? It's distribution. It's partnerships. It's your sales funnel processes and so on.”

“I think models are not a mode. Models eventually don't matter. What matters most is the people building those models, and how fast can you change and learn from those models?”

“I don't think there's going to be a single model to rule them all.”

“It's very important because, like, you know, if you just were to do the naive thing of take, like, every conversation you had with everyone and just train on it, then You could just be, like, spitting out, like, someone's private, you know, private life to, yo…”

“The data that you get from users is great because It tells people like what, you know, what users like or like what users like in, in some particular application. It's kind of like a, you know, training a human. Like most of what's important is like you have y…”

“I don't think there's going to be a single model to rule them all.”

“I think models are not a mode. Models eventually don't matter. What matters most is the people building those models. And how fast can you change and learn from those models?”

“There are a lot of issues. Probably the most obvious one is around the Correctness and dependability of answers. If you ask folks, one of the key concerns is like, well, these models make up stuff. So that, that is the obvious one. But then there are second or…”

“Anyone that builds too tightly coupled to a given model is sort of giving up optionality for the future.”

“It's very clear the top models are memory bound as well as CPU bound. So you can't just make the CPUs, the GPUs faster. You need to increase memory bandwidth in sort of on par with that.”

“I think they'll get a lot of the compute for sure. And that's, and I should say, I think the value is going to accrue in those two directions, infrastructure and app layer, and that the models are going to be squeezed.”

“I think the models are just not smart enough, which sounds like a annoying, low information kind of cop-out answer, but I think it's like deeply fundamentally true. Like the models just aren't smart enough. You fix that one thing, all these other things get be…”

“People use certain models because they are known to be good. You can't afford to evaluate everything out there. And so having some form of community vouching is super important.”

“I believe there is. I believe we can make models that are much better for a certain size.”

“And so I think that's the end state is models are effectively going to be A starting point for any AI application developer. They need to be surrounded by tools, by lifecycle management platform basically, and that's the one thing that we started to build.”

“The cost to the end customer, I think that should continue to go down. And that's what we're seeing is that it is just getting more and more competitive. The price continues to go down.”

“The challenge with RPA is, you know, RPA is relatively frail. It's often sort of, it's like looking at your computer screen and, you know, performing some kind of wrote, you know, routine actions doesn't handle variability very well. Because there's again, doe…”

“But I would say I, I'm seeing no signs of AI models, slow down, anything plateau in terms of the innovation curves that that we're seeing.”

“AI models can now do design and copy better than 90% of people.”

“Today's case is that we have achieved long context before achieving good instruction following. So you can dump a lot into your prompt. You have the memory, but models can hallucinate or get confused because of so much information to focus on.”

“That is like infinite memory. I think like we don't even have the algorithms for it yet today.”

“I think that's not the case today, which is why like these models are not so good that like, you know, they can just write an entire code base yet, but all that will happen. I think it's just a matter of time before, you know, they run another training run and…”

“I think AI services will actually create more revenue over the next five years than AI models.”

“And I think a lot of the plateau that we've recently seen can almost be explained at a very high level from hitting kind of a data wall.”

“Right now you know, HIPAA and PII and all the PII regulations will more or less prevent patient data from being used to train you know, AI models.”

“Now the critical path for model improvement is, is is shifting over to this to this sort of broader sort of simulation slash synthetic data slash like RL loop sort of path. I think it's just a natural consequence of the fact that it's so expensive to just keep…”

“What's going to prevent this for what I think has a hope of preventing this from just being a hype cycle that falls flat like AB is that those things are, those shoes are yet to drop. And as they do, the capabilities of these models are going to continue to im…”

“I don't think there's a lot further to go in just finding more text. You know, I think we may be at the flattening out of the S-curve on text, but I suspect the next S-curve could just involve finding ways to yeah, to use video, to use, like, interactive exper…”

“Option one is this is it like the, you know, the models don't get much better or, you know, and it's sort of this whole thing sort of fizzles out. I think this is unlikely because I think not only will models get better, but also we haven't even started integr…”

“I think we'll continuously exist in a world of multiple models, some focused and verticalized, others completely horizontal.”

“The reality of the matter is there's no market for last year's model.”

“The models definitely do hallucinate. The hallucination rates have been dropping dramatically, but they'll, they'll always have some chance of making stuff up or getting something wrong.”

“For the past year, year plus, folks have been focusing on that, and it will be ready for production. And so we'll see that come out, and I think that will be a big change in terms of capability.”

“Like a single bad example, right, amongst, like, billions. Like, it's so sensitive. Like, it is a bit surreal how sensitive The models are to their data. Everyone underrates it.”

“Because these models are getting smarter, humanity's ability to distinguish between them, or not humanity, but each individual's ability to distinguish between them becomes way harder. You can't tell the difference between generations, because you're not enoug…”

“It becomes cheaper to build last year's model by like a factor of 10 or a hundred each year. We just get better data, cheaper compute. So yeah, it definitely lowers the barrier to the previous generation of models.”

“We have you know, close to state of the art AI models that can already run on people's personal devices, and I think that trend is only going to accelerate.”

“So I think that's going to happen both with models as well as with these hardware cycles. You know, I can't predict how long that's going to take, but we are, I think, going to get to a world where models do get commoditized.”

“My view is that in a lot of cases, the adoption of these models is not bottlenecked by capability. If these models were actually deployed today to do all the tasks that they're capable of, it would truly be a striking economic transformation. The bottlenecks a…”

“We're not gonna have too many more cycles, possibly zero more cycles, of a model that's almost an order of magnitude bigger in terms of the number of parameters than what came before, and thereby more powerful.”

“These models are already trained on essentially all of the data that companies can get their hands on.”

“This is sort of like these models have a buffer overflow in all of them that we know about, and most importantly, that we don't know how to patch and fix. We don't know how to fix this yet with models”

“The domain I use models for most probably is coding, and also doing things like transcribing lectures and stuff like this. On those tasks, I am absolutely not seeing plateauing gains. The latest models, they are notably better than the previous iteration, and …”

“So I think this perception has more to do with people's limited imagination of what they can do with these models, and less to do with the models themselves. But that will evolve over time. People will start figuring out you can use them for better and better …”

“I think this is essentially already solved in many cases by these models that can already solve and analyze code to find vulnerabilities.”

“But it is also true that if you just take a fixed data set and run over it multiple times, if you use a bigger model, it will often work better, right? So I think that we have not really reached the plateau there.”

“We do not understand how these things work internally, the possible correlated failures, the possible attack factors, all sorts of things. We don't understand it.”

“The concern is that not that they can do this sort of autonomously, maybe initially, but that they can lower the bar so far in the skill level required to create these things that effectively it puts them in the hands of a huge number of bad actors.”

“Right now, the mod, the AI models that we have, for lack of a better phrasing, are not able to reliably follow specifications.”

“I think there will come a time when a certain capability, a certain ability of these models Reaches the point that should give us pause when it comes to just turning these things over to whoever and what, however, they want to use them.”

“And so I think this idea that somehow the way the world wants to buy software will change because these models are really smart. Doesn't resonate with me.”

“The real differentiation between two models is the data.”

“This thing that they're not though worth as much as they cost to train, that seems totally wrong. To say nothing of the fact that there's like a, there's a positive compounding effect as you learn to train these models, you get better at training the next one,…”

“There's a lot of, there are probably too many people training very similar models, and if you're a little behind, or if you don't have a Product with the sort of normal rules of business that make that product sticky and valuable, then yeah, maybe you can't. M…”

“I think developers use multiple models. Most of the time, and I'm not sure how that's all going to evolve as we head towards this more agentified world.”

“I personally don't believe that the models can innovate in a material way in a reasonable amount of time. I think they reach maturity in a way.”

“Model is it's a piece of machinery. It's an engine, but what is the car? What is the experience?”

“This has just made it absolutely nakedly clear that the models are commoditized, right?”

“So better models create more demand for inference. Then has people going, I should train a better model, and the cycle continues.”

“If your job is to get better at multiplying numbers, and I tell you that I want you to be able to do it with fewer steps, more intuitively, for you to be able to multiply three-digit numbers versus two-digit, you need 10 x the data, and you need 10 x the examp…”

“We have decided that we're not going to train our own models. We'll do a little fine tuning for specific cases or whatnot, but we don't want to compete.”

“So synthetic data will certainly be a part of the equation, but in a lot of ways, the bottleneck to unlocking and unleashing the next level of intelligence will be expert humans, which brings me back to the question and phrase that you used, low quality human …”

“I think we'll live in a world with many, many models with different use cases.”

“So I think it absolutely has to be a mix. And I think the best models will come from that combination of great, like for code it's, you know, being, having good foundational understanding of code and good examples, but then also being able to explore a really …”

“Recognize that they're not yet at the place where for many use cases, you can let them kind of run free for hours. You need that more human in the loop piece.”

“How do we evolve from being mostly code writers to mostly Delegators to the models and code reviewers. That's what I think the work looks like three years from now. It's coming up with the right ideas, doing the right user interaction design, figuring out a de…”

“And I think models, this is very underappreciated and probably under-researched as well from like a model capabilities perspective because models fundamentally want to be helpful. And that is not always what you want them to be. And there's a safety case for t…”

“What you need to ramp is to decide what, like, the process for storing memory, I think is what you have to ramp and it has to store all of these things. Like my conversation with Harry has to be stored somehow into the system and all like my childhood almost h…”

“Almost all synthetic.”

“Most of those assertions that people make are like most of the assertions that people make are just unfounded in any kind of science. Like no, no one's gotta, and like the measurements we do have show that there's a pretty big disconnect around what some peopl…”

“People who think of models as repositories of factual information, and they're treating them like the world's worst and most expensive databases. Like it's not, not super useful.”

“What you want models to be able to do is to be able to reason over information. So like if you give them access to information, like how well are they able to reason over a set of information to go do something that's useful for you? And so you just need diffe…”

“I think models are gonna be better at interviewing than people. And like the recruiting process now is just so inefficient that if you have like a model interviewing, like that's going to make it a lot better.”

“The wrong time to be the tiger global play of investing quickly was back in the COVID boom. The exact right time is now, when I think you have this unlock of capabilities with the AI models, which has meant new products which just couldn't exist before.”

“Information knowledge, intelligence, for lack of a better word, is just going to get cheaper and cheaper, faster and faster. There's going to be, there's just intelligence is available at a price that's declining, you know, by log or log orders of magnitude ev…”

“If all you're kind of doing is throwing PhDs as a problem, all you're doing is teaching models how to hack Silly benchmarks and get good at basically the equivalent of SAT problems.”

“You always need this external value system as a kind of safeguard to make sure that the models are working properly just because the models themselves have such a different set of way of thinking.”

“Synthetic data, it's made models good at synthetic problems, not, not real ones.”

“Unlike actual software open source, just because you release your model doesn't mean somebody can replicate it. Like to replicate it, you'd have to like recreate the data pipeline and the training pipeline.”

“Where the products are today, I don't think we've utilized 10, 20% of what the models can do. We have a huge sort of gap and way to go. And I think we could, just given more time, With the models we have today, given no incremental improvement in them, still h…”

“Like, these models have gold medals in Olympiad math, and they're better than the best PhD at reasoning, but they can't Draft an email for me. They can't schedule a meeting. They can't do so many of the basic things of just using a handful of tools to do a tas…”

“I don't think that Models are plateauing. Like, if we look at the last 12 months of progress in models, I've been blown away.”

“The total addressable market is limited by the amount of things that humans are better at than models.”

“I very much believe so, and the reason is that the question comes down to when we'll have super intelligence. Once we have super intelligence and models are better than humans at everything, then, of course, that means that humans won't be able to contribute t…”

“It's like, for me, the other one is like choosing which model you run on. Are you kidding me? We're not going to do that in a year.”

“I think you'll have a couple of global ones and a couple of sovereign ones in every geo. I mean, that, that probably is what'll end up sort of happening in, in AI, in my view. And not everybody's gonna make it.”

“I think the model itself is not a clear advantage.”

“In our case we found an area where we will not compete with our customers, which is we will not create our own models. So we just won't do it.”

“Another example, Gemini, the new Gemini, right? His capabilities around specific years on security are much better than any model we've seen before.”

“If you look at all the public benchmarks, models have increased 40 to 60% in performance over the last two years.”

“I think in, in most cases, that's not actually a problem with the model or a problem with the technology. I think it's often a problem with the user experience.”

“So the most sophisticated companies are starting to take support, sales, collections, operations, bundle them all together with one broad goal, like CAC improvement. And I think that is going to be the 10 X on productivity more than saying, hey, we're just goi…”

“Like whenever I talk to any, like even internally or even talking to like customers of our models, I'm always surprised by how vibes based the evaluation of how it feels to work with the model is.”

“Model capabilities are improving every six months because if you have too long a product roadmap, you're going around your program and the model comes and just blows it out of the water. So you've got to really understand what the capabilities are of each new …”

“Because of how quickly these models are just getting exceptionally good, it's almost certain companies are going to be releasing code faster. When you release products faster, you ship fast, you break things, and because of that, you're going to have more secu…”

“I think you do have a loss in ability to feed that information in the model and show other people what you're doing in those in person. So what we tend to do is believe the vast, vast majority of communication needs to be written or through a video call.”

“The capacity to serve more relevant, more targeted ads to the consumer, even knowing less about the person, has gotten so good that people are really able to use advertising to discover the products that they want.”

“Those capabilities are going to exist in the models in the next 12 months.”

“They'll be able to clean the data themselves fairly effectively as reasoning capabilities go up.”

“Those are the kinds of tasks that we need to be building to push the frontier of research and evaluation so that those are the capabilities that people are able to use in the models in six to 12 months.”

“Engineers. For sure. Because the models they change all the time, they get better all the time, but if you have really good engineers, you can build a system that exponentially improves, and that's worth a lot more.”

“I'm a believer that we're quite far from the wall and we will see a lot of like models improvement happening. I think that what we also see is much more new, like modalities and specialized models coming in game.”

“So I think what is necessary for the best outcome for the consumers Is going to be models that are separate from the applications. You, as a consumer, do not want to use applications that are provided for you by the same people that are giving you the model, b…”

“The bare case against factory is if one model provider gets significantly better than all of the others. So basically I think a key thing for us is that all the models are going to be roughly as good as each other.”

“At the face of it, this is a Rubicon moment in the history of the AI industry. It's the first time that the US has ostensibly regulated an AI model based on capabilities.”

“When you leave oligopolical stages, there's massive price erosion because you're no longer competing on features, you're competing on price, and that's exactly what we're seeing.”

“No matter what model wins, if you believe in AI, and if you believe in intelligence, there's going to be an energy layer underneath That is gonna be needed to power these things.”

“So on Wix, when you generate a website today, we are running a model that we trained. Okay. And it's faster, it's cheaper, but it's also better and it makes less errors.”

“As every organization goes from burn all the tokens you want, baby, to budgets, However you do it, whether it's automated, whether it's manual, whether it's picking a dropdown, everyone's going to have to have a cheaper model.”

“I think if you approach it in the right way, then the cleanliness doesn't really matter that much because the models are so good at cleaning up the data if you give them the right context.”

“First of all, within 10 years, I believe we, and I think some people are thinking two or three years, continuous learning models will come into existence. That means that every generation of every model we have today Dies, goes away.”

“The truth is, there's gonna be thousands of different forms of sandboxes, and you're gonna need a company that understands how models look at tools, and what that behavior is, and be able to take that behavior and optimize for it.”

“Models are going to be like that as well, where they emit opinions and they have takes that are different from the ones that are most popular, and people will gravitate towards those.”

“I think every company will have their own model, you know, trained on their own data. They'll supplement their own data with additional data.”

“I think there will be two vectors in the opposite direction. The usage will increase. And so the five percent will become 50%, but the cost will go down. And so net net, I think it will go down.”