It is time now for our AMA episode. And so for the past couple of months, listeners have been sending in some really fantastic questions. We cannot answer every single one that we got, but we are so appreciative of all of you who sent in your questions. This is going to be a super interesting conversation. Here's our question from Ravinder: 'How has your pace of learning changed in the era of AI? What's been the coolest thing you've done with AI personally?'
Yes, I've definitely been using AI a ton for the projects that I'm doing outside of work. The stuff I use it for at work is hugely useful, but the outside of work ones are fun. The coolest thing... I have been researching how to replicate some of the results in traditional, classic Japanese Raku tea bowl making. This has involved making my own kiln, devising my own glaze recipe, and even devising a way to take a clay body and make it tougher so it can handle thermal cycling and the crazy firing process. I will tell you that Copilot was amazingly useful in all of that, particularly with the kiln design and with helping get some ideas and make progress on the glaze chemistry. For the glaze design I asked Copilot, 'I've got a set of tea bowls that I am firing in the classic Raku style at 1100 degrees Celsius, where I'm going to take the glazed vessel, put it directly in the kiln at temperature, leave it for three minutes until the glaze goes cherry red, and then pull it out to air quench.' I gave it a few hints about what I had been thinking. The traditional Japanese Raku glazes use lead, and there are forms of lead you can use that are safe in ceramics, but I didn't want to. You need to use something else like boron. Figuring out how much boron to use and in what form is tricky. Copilot was super helpful. It felt like a real conversation with someone who knew a bit something different about glaze chemistry than I know.
But that's genuinely fascinating. Thank you for sharing all of your many interests with us because you're such an interesting person. I never would have thought of that. I knew how much you're into maker stuff.
This question is from Rafael: 'Do you believe that in the future AI will completely reshape the way that we produce software? Could we eventually get rid of the development tools that we use today and rethink the entire process from scratch, creating a completely new approach to software development?'
I'm an old enough fart that I've been programming for 40 years. I'm 52, I started when I was 12. In 40 years, particularly the past 40 years, software development even without AI doesn't really resemble much of what it used to be. It's super clear that AI is going to change the way we write software. There are all the obvious ways it's going to change things. Coding is a complicated activity where you have an idea in your head that needs to be sharpened, and then you need to get that sharpened idea out into a form that the computer can execute. The big thing that's changed with AI is that now you have a way to describe what you want accomplished, not necessarily even in a programming language. It changes what it means to be building. For instance, I just had a conversation with a bunch of engineers: 'I don't know that you need apps in this world.' An application is a byproduct of someone having to understand a set of problems and then adding code together into a thing called an application that does those things in a general enough way. I don't know that you'll need to hide the complexity of software behind weird user interfaces versus just saying what you want done. That's changing clearly, and that has implications for software development.
Yes. That's what I think about. It could change completely how we define a developer, which we've been trying to do in various ways for a long time. We finally feel like we're on the cusp of really broadening that concept. But it makes me think about how you design programming languages, or how does that change? What matters about the underlying code? How does that change how we design those underlying systems?
Yes, hundred percent. The trick is that you want things to compose. AI is still pretty far from doing the grand vision I articulated. Between now and whenever that happens, we need to think about how to take tools from classical software development to this new AI future and make sure all the things compose together so developers can get the thing built.
This next question is from Jen, who's shifting a bit but still around AI. She wants to know: 'How do you suggest we regulate AI? Should this be done at the federal or state level, and how can we ensure that AI is safe and secure, with both public and private standpoints?' Great question.
Yes, I think it's a great question. I've said this before, even in my book. Of course, any technology as powerful as AI needs to be regulated; it would be odd if it weren't. What you want with regulation is consistency. A reasonable set of international standards would be super useful. Good regulation's intent is to get beneficial technologies deployed to those who will benefit as quickly and safely as possible. You don't want unnecessary complexity in the regulation because that prevents beneficial technologies from reaching people. In general, we need regulators to be pretty agile. In many places, the biggest downside risk is failure to deploy quickly enough. There are whole bunches of medical things where models are strongly superhuman. My own mother had experiences with the healthcare system where if she had access to the most advanced AI tools, a lot of suffering could have been reduced. Lots of people are in similar situations where it's not some theoretical future; it's now that it could be beneficial.
Regulation is super important, and it needs to be consistent. But I sometimes wonder – it's hard enough for us as technologists to keep up with all of these things. How can we do a good job of making sure the legislators are informed?
What I'm most encouraged by is that more than with any previous technology, you have practitioners spending time talking with people in the academy and in government to make sure they have the information they need. If you're in the industry, you're obviously biased. We all need to be as clear about our biases and lay them on the table. Having biases doesn't mean you can't get information out there, and then someone can adjust for biases and make good policy. That's way better than not being transparent or deciding not to talk to someone because it's not your job.
This question is from Alex: 'How can large language models be scaled effectively across regions with limited technological infrastructure?' Think of places like African nations. What are the biggest hurdles for AI-powered educational solutions to move beyond prototyping into full-scale deployments in underserved regions, and how can these be overcome?
The news there is probably pretty good. If you want to build an AI application, it has never been easier than right now. You have many choices about very powerful models to access, on a spectrum from general purpose to very specific task design. You have a lot of choice; you don't have to start by saying, 'I've got to train a model from scratch.' That's a huge advantage. It wasn't like that 20 years ago when I wrote my first ML programs, nor even three or four years ago.
Right. It's much easier now to build really good things than three or four years ago.
Yes. I was talking to somebody the other day who is seeing diffusion of AI applications at a pace he's never seen before. He said there are parts of rural India where the industrial revolution still hasn't shown up after 250 years, yet they are already seeing AI diffusion. A farmer through their mobile device can access a powerful AI system to help them understand government programs and sign up for benefits. That's a shocking rate of diffusion. Even with high accessibility to APIs and infrastructure, you still need to be connected and have baseline technology fluency. Large parts of the world are not yet sufficiently connected, and technology fluency isn't as good as it should be. There's deeply unsexy work we need to prioritize, like rural broadband.
My mom got fiber optics, but my uncle, who lives just a few miles away, is still on a crazy 300K DSL connection. His internet is barely usable; he has to come to my mom's house. That's the sort of thing we have to pay attention to because as capabilities become more powerful, absence of connectivity becomes a bigger disadvantage.
I think you're exactly right. This is a conversation we've had before. Connectivity is only going to be more important. It's easier than ever to build applications, but actually getting them to people and making it so they can interact with them is the less fun part, but arguably more important.
