Bill Gates, Co-Chair, Bill & Melinda Gates Foundation
John Catley
The Microsoft founder spoke to Forbes about his work with the AI unicorn OpenAI and back at the Microsoft campus, the potential impact of AI on jobs and medicine, and more.
IIn 2020, Bill Gates left the board of directors of Microsoft, the tech giant he co-founded in 1975. But he still spends about 10% of his time at the company’s headquarters in Redmond, Washington, meeting with product teams, he says. A big topic for discussion in those sessions: artificial intelligence, the ways in which artificial intelligence can change the way we work – and how we use Microsoft software products to do so.
In the summer of 2022, Gates met with OpenAI founder and president Greg Brockman to review some generative AI products released by startup Unicorn, which recently announced a deep “multi-year, multi-billion dollar” partnership with Microsoft.
You can read more about OpenAI and the race to get AI into action — including comments from Brockman, CEO Sam Altman and many others — in our print feature here. Sharing Gates’ thoughts on artificial intelligence exclusively with Forbesbelow.
This interview has been edited for clarity and consistency
Alex Conrad: It seems like 2018 was the first time I saw you talking about what OpenAI was doing. Is this true, or where does your interest in the company begin?
Bill Gates: [My] The interest in artificial intelligence goes back to the early days when I learned about software. The idea of computers seeing, hearing, and writing is a long-term mission for the entire industry. It has always been very interesting to me. And so, when these machine learning technologies started to do very well, especially the speech and image recognition stuff, it kind of amazed me how many other inventions we’d need before. [AI] Really smart, in the sense of passing exams and being able to write fluently.
I know Sam Altman very well. And I got to know Greg [Brockman] Through OpenAI and some other folks out there, like Elia [Sutskever, Brockman’s cofounder and chief scientist]. And I’d say to them, “Hey, you know, I think it only gets to the upper limit if we have more explicitly cognitive representations, clear forms of symbolic reasoning.” There have been a lot of people asking these questions, not just me. But they’ve been able to convince me that there’s important emerging behavior as you extend these large language models, and they’ve done some really innovative things with reinforcement learning on top of it. I’ve kept in touch with them, and they’ve been great at showing me their stuff. And now over time, they’re doing some collaborations, especially with huge back-ends that those skills require, and that really comes through in their partnership with Microsoft.
It should be fun for you personally, and your legacy helping their legacy.
Yeah, that’s cool for me because I love those kinds of things. Also, I wear my foundation hat [The Bill & Melinda Gates Foundation, which Gates talked more about in September]The idea of having a math tutor available to inner city students, or medical advice available to people in Africa who generally won’t be able to see a doctor during their lifetime, that’s pretty cool. As you know, we don’t have a white collar staff capacity available for many worthwhile issues. I have to say, really in the past year, progress [in AI] Made me very excited.
Few people have seen so many technological changes, or major shifts, up close as I have. How does artificial intelligence compare to some of these landmark moments in technology history?
I would say, that’s right there. We have a computer without a graphical interface. Then you have the computer with a graphical interface, which is things like Windows and Mac, which really kicked off for me when I spent time with Charles Simonyi at Xerox PARC. This demo made a huge impact for me and set the agenda for much of what was done both at Microsoft and in the industry after that. [Editor’s note: a Silicon Valley research group famous for work on tech from the desktop to GPUs and the Ethernet.]
Then of course, the internet takes that to a whole new level. When I was CEO of Microsoft I wrote a “tidal wave” note on the Internet it’s pretty amazing that what I see in AI in just the last 12 months is just as important as the PC, PC with a GUI [graphical user interface]or the internet. As the top four milestones in digital technology, this ranks right up there.
And I know OpenAI’s work better than others. I’m not saying they are the only ones. In fact, part of what’s amazing is that there will be so many entrants into this space. But what OpenAI has done is very impressive, and it’s certainly groundbreaking in many aspects [AI]which people see through the wide availability of ChatGPT.
How do you see this changing the way people work or how they do business? Should they be passionate about productivity? Should they ever be worried about losing a job? What should people know about what this will mean for how they work?
Most of the futurists who took a look at the advent of AI said that repetitive material and physical jobs would be the first to be affected by AI. And it’s definitely happening, and people shouldn’t let their guard down about it, but it’s a little slower than I expected. You know, Rodney Brooks [a professor emeritus at MIT and robotics entrepreneur] He put forward what I would call some overly conservative opinions about how quickly some of these things can happen. Autonomous driving has its own challenges, but the use of robots in the factory will still happen in the next five to ten years. But what’s surprising is that tasks that involve reading and writing fluency—like summarizing a complex set of documents or writing something in a pre-existing author’s style—the fact that you can do that with these big language models, and reinforce that fluency is really amazing.
One of the things that challenged Greg [Brockman] With the onset of summer: “Hey,[Can]an OpenAI model pass the AP Biology exams?” And I said, “If you explain that to me, I’d say he has the ability to represent things deeply abstractly, and that’s more than just statistical stuff.” When I was first programming, we did random sentence generators where we had the syntax of typical English sentences, you know, noun, verb, object. Then we’d have a bunch of nouns and a bunch of verbs and a bunch of things and we’d pick it up at random, and every once in a while, he’d spit out something funny or semi-disguised. You’ll go, “Oh my God.” This is the kind of “monkeys typing on keyboards”.
Well, this is close to that. Takes [the AI’s] The ability to take something like an AP test question. When a person reads a biology book, what is left in your mind? We can’t really describe it on a neurological level. but in the summer, [OpenAI] It showed me progress which I was really surprised to see. I thought we would have to invent a more explicit knowledge representation.
We had to train him to do a sudoku, and he’d just mess it up and say, “Oh, I misspelled it.” Well, of course I misspelled it, what does that mean? You don’t have a keyboard, you don’t have fingers! But you “mistype?” amazing.
Satya [Nadella, Microsoft’s CEO] It’s so cool about getting input from me on tech stuff. And I might spend 10% of my time meeting with Microsoft product groups about their product roadmaps. I enjoy this time, and it helps me to be up to date on the Foundation’s work, in the areas of health, education, and agriculture. And so providing feedback to OpenAI over the summer was also a big win. (Now people see most of what I’ve seen; I’ve seen some things that are somewhat more recent.) If you take that lead, the ability to help you write and help you read is happening now, and will get better. And they don’t reach limits, nor do their competitors.
Well, what does that mean in the legal world, or in the billing processing world, or in the medical world? There was an enormous amount of play [ChatGPT] to try to run those applications. Even basic things like search.
[ChatGPT] It is really imperfect. Nobody is suggesting that she doesn’t make mistakes, and it’s not very intuitive. And then, with something like mathematics, it would be completely wrong. Before he was trained, his self-confidence on a wrong answer was also mind-blowing. We had to train him to do a sudoku, and he’d just mess it up and say, “Oh, I misspelled it.” Well, of course I misspelled it, what does that mean? You don’t have a keyboard, you don’t have fingers! But you “mistype?” amazing. But this is the crux of the book [of training text] I taught her.
After spending time with Greg [Brockman] badge [Altman]What makes you confident that they are building this AI responsibly, and that people should trust them to be good stewards of this technology? Especially as we get closer to artificial general intelligence.
Well, OpenAI was founded with that in mind. It’s certainly not a purely profit-driven organization, though they do want to have the resources to build big, big, big machines to get those things forward. This will cost tens of billions of dollars, ultimately, in hardware and training costs. But the near-term problem with AI is a productivity issue. It will make things more productive and this will affect the labor market. The long-term issue, which has not yet caught up with us, is what people worry about: the issue of control. What if the humans controlling it take it in the wrong direction? If humans lose control, what does that mean? I think these are valid discussions.
These guys care about the safety of the AI. They will be the first to say they didn’t solve it. Microsoft also brings a lot of sensitivities around these things as a partner, too. And look, AI will be discussed. It’s going to be the hottest topic in 2023, and that’s right. It will change the job market to some extent. And it will really make us wonder, what are the limits? [For example] It is nowhere near a scientific invention. But given what we’re seeing, that’s within the realm of possibility five years from now or 10 years from now.
What’s your favorite or most fun thing you’ve seen these tools create so far?
It is very interesting to play with these things. When you’re with a group of friends, and you want to write an poem about how much fun something was. The fact that you could say well, “write it like Shakespeare” and it is – the creation was fun. I am always amazed that although the reason I have access is for serious purposes, I often resort to [ChatGPT] Just for the fun stuff. And after I read a poem he wrote, I must confess that I could not have written it.
More from Forbes
