AI Systems

“I think now putting things on back on the web and then making sure that they're accessible across the different AI systems in a way that makes sense for creators and make sure that creators get compensated is going to be something which actually I think will r…”

“Verifiability is ultimately the single most important property of success with current AI systems.”

“Where I think a concern that we have is like on the default trajectory, you maybe go straight from like AI systems that are like competitive with humans to AI systems that are wildly superhuman in a very short period of time.”

“Eventually the truth is the best systems use both. The AI will understand relationships inside your business, and it will also know how to move through the right steps.”

“I think it's the first time in human history that AI has become the face of the product. A user is directly talking to an AI and working with an AI. So every output of an AI system is mission critical. And with agents doing actions and making critical decision…”

“The problem remains then and now is regulatory because you actually can't shift the burden Of the wrong clinical diagnoses from the physician to the AI system.”

“So we think ultimately that in this era, the amount of memory per user could quintuple just based on the demand that these AI systems are requiring to provide their services.”

“So I think there's now of the order of a hundred AI systems for interpreting scans that have been federally approved and they're being used a lot by radiologists.”

“And already we know that AI systems are better than doctors at diagnosis.”

“The, there's no sharp point, or I don't think there will be a sharp point where we'll say that systems didn't, weren't able to be useful for scientific, the scientific process to their fully fledged scientists. There'll be sort of a gradual shift.”

“In, in general, we're not just going to like set up a system and let it think autonomously for eight years, if anything, because like the system's eight years. After will be so much more powerful that it probably doesn't make sense to let a system think for a …”

“There's a lot of systems where that's just not viable technically because AI today runs On the cloud, on someone else's infrastructure, on your endpoint, and just proxy is not always an option.”

“The hard problem is understanding if what I should do now. It turns out that in the case of AI systems, that is the hard question. Like, what is the engine that needs to underwrite these different actions and say if they're okay or not? And because We need to …”

“These AI systems need to log the entire trace of how the AI reasoned on the initial input. What were the tool calls it made? How did it interact with the LLMs? How did it sort of ultimately generate the response? And did that response actually meet the user in…”

“The entire complexity of an AI system evolves from the data they're trained on.”

“I've yet to seen anything so far, and we all tinker with same things, you know, some math problems that are a little bit harder than IMO and so on. I haven't seen anything yet that is a true genuine, You know, massive discovery. That's my personal opinion.”

“When I say just, we're now talking about just like solving the Riemann hypothesis or something. This would be obviously amazing. Well, one of the Millennium Prize problems, and maybe we're a couple of years out from doing that.”

“I do think these systems will be eventually be able to do that. Maybe we were missing one or two things.”

“Marty brought in an AI system to help them be able to comb through the thousands of pages of filings that sponsors behind medicines have to submit in record time versus the, you know, armies of people that would have to comb through that and the days and days …”

“These systems are not very good at planning at long time horizons, you know, many years into the future which we is, you know, with our minds we can do.”

“And it's not ideal, but I think we're gonna have to try and do the best we can to at least come up with a sort of set of minimum, maybe minimum standards, some benchmarks that test for undesirable properties, for example, deception. You don't, you know, nobody…”

“It wouldn't be desirable to have AI systems output tokens that are not human readable. So, you know, in some kind of machine language that We couldn't understand. I think that would, you know introduce a new vulnerability.”

“The opinion that I and others held as periodic was, you're not going to see the same kind of acceleration in science and technology unless you start connecting these things to the physical world.”

“I think one fallacy is thinking about intelligence as a scalar. We've consistently seen these systems have a very odd spikiness, and it's actually possible to architect a system that is world-class on some math domain, but then you could do some perturbations …”

“So rolling forward that software engineering self-improvement, I think you're going to have a system that can write complete repositories, identify bugs, refactor code, But it doesn't suddenly understand biology. Right? It's just like there's a domain gap ther…”

“In the future, I believe that you are not going to do clicks to get something done. You're just going to ask in a natural language, right? So, which means that like, I think this earlier software companies more than likely is going to get replaced by AI system…”

“So in the short term, what we're seeing is there are a lot of areas where there is a shortage, especially of highly skilled labor. It's not just unskilled labor. And there are organizations that are actively looking to hire and can't find people. And this crea…”

“I do think when our AI system, when our AI systems get advanced enough, I suspect they'll have something that, you know, resembles what we would call consciousness or moral significance. I do think it'll happen at some point.”

“Meta has obtained a patent for an AI system that could simulate deceased users by analyzing their historical data, allowing the account to keep posting, messaging, and even making video calls in their behavioral style.”

“And the fundamental problem that we're seeing is we don't know how to take our values or our goals and encode them into these AI systems and get them to pursue it reliably.”

“The AI systems are helpful above and beyond Google in part because they can go back and forth with you and they can help you troubleshoot and dynamically answer your questions. And so even presented with the information on Google, people often can't get all th…”

“That level of sort of compounding improvement in processing power means that in just a few years, you're going to have systems that, you know, can think a lifetime worth of human thoughts in the time that you and I sort of think one or two thoughts, right?”

“Well, I think it's certainly achievable, in my view to make AI systems that appear to be as intelligent in every way as a human and beyond that. I mean, we're gonna have AI systems, I'm fairly confident, that are not only as intelligent as us, but are way smar…”

“And today's systems, in my opinion, are nowhere near that. Doesn't matter how many, you know, Erdos problems you solve, which for some reason, you know, I mean, you know, that's, it's good that we're doing those things, but I think it's far, far from what you …”

“Eventually the entire economy is going to run on AI systems. That's for sure. But it might take up 20 years because it's actually fairly complex.”

“There's not really a rule book you can follow, but it's all about minimizing surprise, which means let's say you're calibrating every one or two days and you figured out that you're not seeing new data distribution patterns. Your users have been pretty consist…”

“And I think you know, systems that are purpose-built for scientific research are probably going to focus a lot on that, on the interpretability, on recording what exactly is, is is happening throughout each step of the process as it collaborates with a human s…”

“You can patch a bug, but you can't patch a brain. And what I mean by that is if you find some bug in your software and you go and patch it, you can be 99% sure, maybe 99.99% sure that bug is solved. Not a problem. If you go and try to do that in your AI system…”

“And then they'd be like, oh, you know, they, you know, they'd realize we can just dockerize that code run put it in a container. So it's running on a different system and take a look at the sanitized output. And now we're completely secure. So in that case, pr…”

“And I think in some of the rhetoric in the last few years, you can feel that there's a little bit of like you know, a kind of creeping assumption that it is inevitable that These kinds of systems exceed our control and our capability and move beyond us as a sp…”

“A smarter person is going to say, no, they should not do math. Instead, I should write the two lines of Python, which I know how to, you know, direct and run the Python in order to solve the math problem. I think of trust in a very similar way. There are well-…”

“Cause the big unlock that most people have yet to experience is going beyond chatting with an AI system to having it work for them. But most people I find have a really hard time of cresting the idea of Having it do work for you without going through these bab…”

“I don't think it's a phase in the sense that I do think this partnership, we'll call it, between humans and machines, where humans provide guidance to machine, like, we are in this for a long time. What will change is the nature of the information that the AI …”

“There are companies here that are making very, very good margins on serving their AI systems, like, 7080, sometimes 90%, depending on the modality and so, like with everything, the average number sucks but, like, when you look at the best companies, it's reall…”

“I think agency and goals is almost certain. There is a question. I think this is one of the areas where we want to have some clarity and control. That was a little bit like the kind of question of what kind of compute fabric holds it together because you can't…”

“Maybe what we need is the AI system where you feed the novel and it makes the movie. You know, which, by the way, first, you know, now, now that I say that, like, for sure that's gonna happen.”

“But I would be very surprised if you can't get your nuts and bolts Web app, mobile app type things spit out for you for far less and far faster than, and probably honestly with significantly higher quality and less back and forth with an AI system than, you kn…”

“AI systems need to be model agnostic. They shouldn't care about which model that they use, and they should look at what I call actions, which is a combination of a model with a prompt and maybe a set of tools that it can use and say, what can an action do for …”

“I feel like the self-driving car industry did a lot of damage to AI in one particular way, which is it because with self-driving cars, 99% accuracy is not enough. That, that last one percent really matters. But with a lot of AI systems, if you're automating th…”

“Fact checking a long document is still well outside the capabilities of most AI systems because, you know, for the attention reasons, but if you're working in sections, it's really easy to be like, while I'm here,”

“So I would actually say that like a lot of the things to simulate are more aligned with like product KPIs or product metrics that actually make Whatever AI system you're building very sticky, rather than, you know, focusing more on, like, traditional safety se…”

“I think AI systems will give back a lot more to energy and climate change and these kind of things than they take in terms of efficiency of grid systems and electrical systems, material design, new types of properties, new energy sources. I think AI will help …”

“I've still yet to see a system where you, they perform very well and you don't draw a circle that doesn't have a human being in it somewhere.”

“Like, the idea that there's going to be, like, a lot of offline compute and inference under the hood that helps make AI systems continuously self-improve is a sure bet.”

“You got to think about these AI systems as almost biological systems that are grown, not coded up, right?”

“If we strangle off the business model of those places, if we strangle off the incentives for content to create for content creators to create content, then we're actually going to end up strangling the AI systems as well, because there's, if there's no content…”

“We launched the first, you know, knowledge management powered system in a corporate enterprise within Sprint. Kind of set the industry standard for creating and organizing knowledge, which became the basis of today's AI systems.”

“When it comes to computer systems, we can put formal bounds on them. We can do this information theoretically. We can do this computationally. And like, that's gonna come, and once that happens, we will understand these systems fully, and it'll be very hard to…”

“I actually think the best strategy is just tell people wrong things. So they feel the need to correct you, right? So like, there's a whole neural path We're like, people will spend infinite attention on that very happily, right? If there's just an opportunity,…”

“I think these are kind of growing pains for AI systems that have public access. Like it's just Sam has built systems that have untrusted user inputs. As soon as you build one of these, people will do the worst possible thing just, you know, for fun, for whatev…”