The Exchanges

Q for you. So basically you can write the prompt of what you want and the code will be developed on the backend, uh, or will help you devote, develop the code on the backend. Is there going to come a point where we're not going to need code at all to build? Because like, if you can just prompt and why do we really need to be in the code?

A You might say we're already at that point to some degree where, you know, when you ask and go chat GPT, ask and go and ask chat GPT a question, uh, you get an answer and you must, when you ask a question to plot the chart, for example, or do a method, method, mathematical, mathematical, um, calculation. It, um, it actually. Generates a Python script that then, you know, plots that data into a chart and it shows you the chart and it still shows you that step where you see in between. Uh, the Python script, but, uh, you, they could just hide that. Um, and you just see the, see the chart output, right? Like in, in many ways, ChatGPT is giving you an answer without you ever having to worry how that was generated. And so, yes, we're going to see computer systems where large language models are just one building block in addition to code, or maybe it's multiple language models and image models and, you know, time series models and whatnot, plus code combined. Um, To, to generate, um, all the output, um, that the developer or the user expects. Um, do we still need engineers then to code? Yeah, because, well, first of all, you know, there's billions of lines of code out there that still have to be maintained. You know, one of the examples I like to get is that most banks are still running Cobalt code. That's a program in English from, uh, invented in the late fifties when Eisenhower wa…

AI assessment note: “You might say we're already at that point to some degree”

Q no. But that is, we're talking about generative AI in particular. So we can go on for days about how AI has been, you know, for feed ranking and computer vision, fine. But the big moment right now is all about generative. And generative has been something that GitHub has written with Copilot to this amazing moment. But that's the question. Is that type of technology in particular transferable elsewhere?

A I think one, one scenario that comes to mind, um, uh, that we are already using at GitHub is, um, support. Um, and so, um, you know, if you look at our support system, uh, today, you actually find, uh, the GitHub support copilot, um, that tries to help you before you, you know, submit your ticket to a human. And we actually see, um, and it generates answers. So it's generally fair. It uses the same last English models to, to stay within the scope of your question. And we see that the number of, um, tickets that get solved that way, um, Uh, is, uh, about 50%. So, you know, 50% of those questions that go through the support co-pilot get solved by the support co-pilot and do not get submitted into, to a human. And so as such, it makes, you know, supporting, uh, our, our developers, um, um, our customers, um, uh, more efficient for us as a company. So I'd say, you know, that's, that's definitely another scenario where we see the, um, um, efficiency gains for us as a company and other, similarly, You know, we have an internal tool called Octobot, uh, you know, like Octocat or our logo, um, my t-shirt, and it, it helps, uh, you know, our folks internally, um, to, to solve IT problems. And our IT team is getting, you know, more than three hours per IT supporter back through that internal tool by, just by, you know, helping, um, uh, uh, uh, employees, you know, to solve their own IT is…

AI assessment note: “one scenario that comes to mind that we are already using at GitHub is support”

Q no. But that is, we're talking about generative AI in particular. So we can go on for days about how AI has been, you know, for feed ranking and computer vision, fine. But the big moment right now is all about generative. And generative has been something that GitHub has written with Copilot to this amazing moment. But that's the question. Is that type of technology in particular transferable elsewhere?

A I think one, one scenario that comes to mind, um, uh, that we are already using at GitHub is, um, support. Um, and so, um, you know, if you look at our support system, uh, today, you actually find, uh, the GitHub support copilot, um, that tries to help you before you, you know, submit your ticket to a human. And we actually see, um, and it generates answers. So it's generally fair. It uses the same last English models to, to stay within the scope of your question. And we see that the number of, um, tickets that get solved that way, um, Uh, is, uh, about 50%. So, you know, 50% of those questions that go through the support co-pilot get solved by the support co-pilot and do not get submitted into, to a human. And so as such, it makes, you know, supporting, uh, our, our developers, um, um, our customers, um, uh, more efficient for us as a company. So I'd say, you know, that's, that's definitely another scenario where we see the, um, um, efficiency gains for us as a company and other, similarly, You know, we have an internal tool called Octobot, uh, you know, like Octocat or our logo, um, my t-shirt, and it, it helps, uh, you know, our folks internally, um, to, to solve IT problems. And our IT team is getting, you know, more than three hours per IT supporter back through that internal tool by, just by, you know, helping, um, uh, uh, uh, employees, you know, to solve their own IT is…

AI assessment note: “one scenario that comes to mind, that we are already using at GitHub is, support.”

Q for you. So basically you can write the prompt of what you want and the code will be developed on the backend, uh, or will help you devote, develop the code on the backend. Is there going to come a point where we're not going to need code at all to build? Because like, if you can just prompt and why do we really need to be in the code?

A You might say we're already at that point to some degree where, you know, when you ask and go chat GPT, ask and go and ask chat GPT a question, uh, you get an answer and you must, when you ask a question to plot the chart, for example, or do a method, method, mathematical, mathematical, um, calculation. It, um, it actually. Generates a Python script that then, you know, plots that data into a chart and it shows you the chart and it still shows you that step where you see in between. Uh, the Python script, but, uh, you, they could just hide that. Um, and you just see the, see the chart output, right? Like in, in many ways, ChatGPT is giving you an answer without you ever having to worry how that was generated. And so, yes, we're going to see computer systems where large language models are just one building block in addition to code, or maybe it's multiple language models and image models and, you know, time series models and whatnot, plus code combined. Um, To, to generate, um, all the output, um, that the developer or the user expects. Um, do we still need engineers then to code? Yeah, because, well, first of all, you know, there's billions of lines of code out there that still have to be maintained. You know, one of the examples I like to get is that most banks are still running Cobalt code. That's a program in English from, uh, invented in the late fifties when Eisenhower wa…

AI assessment note: “You might say we're already at that point to some degree”

Q Okay. And then, alright, last question for you. You said that you have a hundred million users on GitHub today. You think that you're gonna get to a billion with this. So I'm curious, like, why you think AI is going to drive so many people to start coding? And then what does that mean for a broader economy?

A I believe that today the biggest adoption blocker is the complexity of the technology, the complexity of a language that is not the language that we learn and use every single day when we communicate. And programming languages are great because they're deterministic. Um, you know, the same thing does the same, uh, has the same output every time you write it. Um, but it's hard to learn. Uh, it's hard to learn, uh, when you're a kid, um, uh, it's much harder to learn than playing an instrument, um, or Drawing an image, um, because you have to learn the thing first before you can produce anything. And then you still have to develop your craft and do it over and over again to actually get good at it. And I think AI is going to accelerate that massively. And, you know, one, one billion developers, um, by 2030 or so is a little bit under 10% of the population, depending on where the world's population is going. Um, that's actually a low number if you think about it, because Um, we all use computers every single day, yet we are not able to create the thing. Most people are not able to create the thing that runs on those computers. Um, and I think, you know, most people are able to go to Home Depot on and buy a screwdriver and, and put a screw in a wall. And I think that it's, it's just going to be a fundamental skill of humans, um, to, to, to be able to control the computer and, um, c…

AI assessment note: “I believe that today the biggest adoption blocker is the complexity of the technology”

Q are purpose-built for inference, which is effectively running these AI models, which are extremely expensive to run now, but it's much cheaper to run on an Etched chips. How important do you think hardware Innovation is for these technologies to be able to be cost effective and grow to the point that sort of the industry is betting on. And then on that note, what do you think about it?

A It's incredibly exciting that we have Silicon companies in Silicon Valley again. I think that's number one, there's innovation in Silicon. Um, and, um, there's not only edge, there's, you know, a bunch of companies that are going in the same direction. And, um, I think it's just fascinating to see after, you know, we believe that, um, Moore's law is over and, uh, and, and there's no more innovation in chips. Um, we are back, you know, to a world where there's innovation across the whole stack. We talked a lot about models and we talked about Copilot and we talked about agents, which, you know, is going up the stack, but there's also innovation going down the stack, you know, from, from the model to the data center, um, all the way down to the chip. And, um, I think, you know, we are going to see, uh, much more on that, uh, in the, in the coming years, um, the cost to run influence will come down with these specialized chips. The models itself become more efficient. Um, You know, GPT for many, uh, was, you know, announced last week, um, uh, which is much faster and much more efficient. And I think, you know, we, we are going to have innovation on the tops and bigger models come out and do more stuff. And we are going to have innovation on the efficiency side where the functionality that, you know, a few years ago required, you know, more GPUs and more time is now, it's now done,…

AI assessment note: “the cost to run influence will come down with these specialized chips”

Q Okay. And then, alright, last question for you. You said that you have a hundred million users on GitHub today. You think that you're gonna get to a billion with this. So I'm curious, like, why you think AI is going to drive so many people to start coding? And then what does that mean for a broader economy?

A I believe that today the biggest adoption blocker is the complexity of the technology, the complexity of a language that is not the language that we learn and use every single day when we communicate. And programming languages are great because they're deterministic. Um, you know, the same thing does the same, uh, has the same output every time you write it. Um, but it's hard to learn. Uh, it's hard to learn, uh, when you're a kid, um, uh, it's much harder to learn than playing an instrument, um, or Drawing an image, um, because you have to learn the thing first before you can produce anything. And then you still have to develop your craft and do it over and over again to actually get good at it. And I think AI is going to accelerate that massively. And, you know, one, one billion developers, um, by 2030 or so is a little bit under 10% of the population, depending on where the world's population is going. Um, that's actually a low number if you think about it, because Um, we all use computers every single day, yet we are not able to create the thing. Most people are not able to create the thing that runs on those computers. Um, and I think, you know, most people are able to go to Home Depot on and buy a screwdriver and, and put a screw in a wall. And I think that it's, it's just going to be a fundamental skill of humans, um, to, to, to be able to control the computer and, um, c…

AI assessment note: “biggest adoption blocker is the complexity of the technology... AI is going to accelerate that”

Q I'm talking about like to build new things. Like, is there going to come a time where we're just going to have prompters instead of coders?

A Almost most developers work on an existing code basis. So I wouldn't, I would push back a little bit on maintenance. Um, we're building on top of an existing world. Um, I think there's, uh, developers have always moved up the traction that, you know, we used to build it all ourselves. Then it came to the internet and we used to start sharing software components, so called open source. Um, uh, you know, nowadays most applications are sitting on stack of thousand, uh, uh, components, um, Already, and you're building the 10% layer on top of that. Now that 10% layer, you know, might, you know, at some point be written an 80% by AI or replaced by AI, but that means you have more time for the remaining 20%. On top of that, the pile is getting always bigger, and the developers are still going to have enough work, you know, carved out for them. In fact, I'd say, you know, AI has created more work for developers because now somebody has to build all these AI systems, and we're not At all at a point where you can just, you know, have an AI engineer, quote unquote, uh, do the job of a human, like that doesn't exist to you. And, um, even if it exists, it's, it works well, my demo, um, uh, but it doesn't, it doesn't actually do any real work.

AI assessment note: “developers are still going to have enough work, you know, carved out for them.”

Q are purpose-built for inference, which is effectively running these AI models, which are extremely expensive to run now, but it's much cheaper to run on an Etched chips. How important do you think hardware Innovation is for these technologies to be able to be cost effective and grow to the point that sort of the industry is betting on. And then on that note, what do you think about it?

A It's incredibly exciting that we have Silicon companies in Silicon Valley again. I think that's number one, there's innovation in Silicon. Um, and, um, there's not only edge, there's, you know, a bunch of companies that are going in the same direction. And, um, I think it's just fascinating to see after, you know, we believe that, um, Moore's law is over and, uh, and, and there's no more innovation in chips. Um, we are back, you know, to a world where there's innovation across the whole stack. We talked a lot about models and we talked about Copilot and we talked about agents, which, you know, is going up the stack, but there's also innovation going down the stack, you know, from, from the model to the data center, um, all the way down to the chip. And, um, I think, you know, we are going to see, uh, much more on that, uh, in the, in the coming years, um, the cost to run influence will come down with these specialized chips. The models itself become more efficient. Um, You know, GPT for many, uh, was, you know, announced last week, um, uh, which is much faster and much more efficient. And I think, you know, we, we are going to have innovation on the tops and bigger models come out and do more stuff. And we are going to have innovation on the efficiency side where the functionality that, you know, a few years ago required, you know, more GPUs and more time is now, it's now done,…

AI assessment note: “the cost to run influence will come down with these specialized chips”

Q I'm talking about like to build new things. Like, is there going to come a time where we're just going to have prompters instead of coders?

A Almost most developers work on an existing code basis. So I wouldn't, I would push back a little bit on maintenance. Um, we're building on top of an existing world. Um, I think there's, uh, developers have always moved up the traction that, you know, we used to build it all ourselves. Then it came to the internet and we used to start sharing software components, so called open source. Um, uh, you know, nowadays most applications are sitting on stack of thousand, uh, uh, components, um, Already, and you're building the 10% layer on top of that. Now that 10% layer, you know, might, you know, at some point be written an 80% by AI or replaced by AI, but that means you have more time for the remaining 20%. On top of that, the pile is getting always bigger, and the developers are still going to have enough work, you know, carved out for them. In fact, I'd say, you know, AI has created more work for developers because now somebody has to build all these AI systems, and we're not At all at a point where you can just, you know, have an AI engineer, quote unquote, uh, do the job of a human, like that doesn't exist to you. And, um, even if it exists, it's, it works well, my demo, um, uh, but it doesn't, it doesn't actually do any real work.

AI assessment note: “developers are still going to have enough work, you know, carved out for them.”

Q question. Like, the question is not, can the model, uh, sort of learn to get better as it goes, right? Which is still following the model. And the question is, like, the basic design of these AI programs. Can AI learn to make them even better? Like, can AI be able to take a GPT-IV and turn it into GPT-V, right? That's the real question. You think that's gonna happen?

A It's a, I think it's like predicting the future and I, I don't know if the, if I could, uh, if I can, um, I don't know. Um, I haven't seen, uh, you know, I haven't seen any indication that that's possible today. Um, but you know, maybe I'm on your podcast again three years and you're telling me, you see, maybe the AI is going to do the podcast between the two of us. No, but like, look, you know, all, If you look at the technology today, it's super powerful. It helps developers and support agents and IT employees to achieve their job faster, which ultimately means, you know, they have more time for other things. And I think that's remarkable. And I would, I'm not too worried about, you know, AI taking over these shops and replacing them with a fully automated employee.

AI assessment note: “I haven't seen any indication that that's possible today.”

Q question. Like, the question is not, can the model, uh, sort of learn to get better as it goes, right? Which is still following the model. And the question is, like, the basic design of these AI programs. Can AI learn to make them even better? Like, can AI be able to take a GPT-IV and turn it into GPT-V, right? That's the real question. You think that's gonna happen?

A It's a, I think it's like predicting the future and I, I don't know if the, if I could, uh, if I can, um, I don't know. Um, I haven't seen, uh, you know, I haven't seen any indication that that's possible today. Um, but you know, maybe I'm on your podcast again three years and you're telling me, you see, maybe the AI is going to do the podcast between the two of us. No, but like, look, you know, all, If you look at the technology today, it's super powerful. It helps developers and support agents and IT employees to achieve their job faster, which ultimately means, you know, they have more time for other things. And I think that's remarkable. And I would, I'm not too worried about, you know, AI taking over these shops and replacing them with a fully automated employee.

AI assessment note: “I haven't seen any indication that that's possible today”