Bill Gates interviews Sam Altman about OpenAI’s Speedy Rise

Microsoft co-founder Bill Gates started a podcast and posted a pretty scoop earlier this month: an interview with OpenAI CEO Sam Altman. This is a surprisingly informative discussion about the implications of the speed of AI development. And it also includes specific predictions for the future by two people who are very close to the work being done today.

Most of the recordings were made before Altman was temporarily laid off from the company, but some of the recordings include footage of Gates checking in with Altman after he returned. OpenAI CEO says: It’s a very exciting time,” the team said. “We’ve never felt more productive, more optimistic and better.”

“In some ways, this was like a real growth moment for us. We are very motivated to get better and become a company that is ready to meet the challenges ahead.” ,” Altman said.

Bill Gates left Microsoft’s board in 2020, according to Microsoft CTO Kevin Scott, but BNN Bloomberg’s Jon Erlichman says Gates has worked very closely with the company’s AI team since then. It says that there is. There, Gates and Altman had a conversation that stretched far into the future and made many thought-provoking remarks.

Along the way, Altman also provides a very real sense of the excitement currently being felt within OpenAI.

“It sounds very cliché, and every company says it, but people feel that mission very deeply. Everyone wants to be part of the creation of artificial general intelligence. ”

What happens next?

Bill Gates asked what milestones ChatGPT would reach in the next two years, and Sam Altman gave his best prediction starting with “multimodality.”

Gate: Does it mean “say, say”?

Altman: Speech in, speech out. image. Video after all. It’s clear that people really want it.

ChatGPT’s image and audio features also had a “much stronger response than expected,” Altman said.

Beyond multimodality, Altman also predicted improvements in reliability and “reasoning ability.” But on top of that, Altman sees personalization and “customizability” becoming more important over the next two years. This includes not only different styles, but also different sets of assumptions, perhaps resulting from training on user-provided data.

Speed ​​of Progress and Decoding Encoding

In a previous podcast, Gates discussed the “black box” problem, where we don’t really understand how knowledge is encoded, and the mysterious black art of prompt engineering.

Gates returned to the topic in an interview with Altman, who expressed “100%” confidence that AI will eventually figure it out. “Trying to do this with the human brain is very difficult,” Altman said. “We’re not going to chop up your brain and see how it evolves.” But with AI, “we can take perfect X-rays. There’s been some great research done on this, and I think there will be more over time.”

But what’s even more interesting is that Altman believes that understanding this encoding of knowledge can help us build even better AIs that are more efficient and accurate. Now, “scientific curiosity aside, we all have the drive to really understand them. But the scale of these is enormous.”

And this leads to clear insights into how progress happens. Altman points out that this is a common pattern in the history of technology, where serendipitous discoveries are preceded by subsequent, more detailed understanding, leading to improved performance.

And surprisingly, ChatGPT was just such a discovery. “The people who created GPT-1 kind of solved this by doing it themselves. It was impressive to a certain extent, but they didn’t really understand how it worked or why it worked. I didn’t understand,” Altman said. “It really came out of empirical results first.”

While some predict that AI will be too expensive to deploy broadly, Altman points out that significant cost savings have already been achieved. “With GPT-3, he was able to reduce the cost by a factor of 40.”

In fact, he takes his optimistic predictions even further. “I think we’re on the steepest cost-reduction curve ever of any technology that I know of. Much better than Moore’s Law.” As research gains understanding , the model becomes not only more efficient, but also smaller.

“I think we will bring the cost of intelligence as close to zero as possible and make it the same before and after social change.”

What will happen to labor?

Where does that leave us humans? Altman told Gates that AI is on a “long, continuous improvement curve” and that “ultimately it will be able to do more of the things that we think of today as work.” I’m reminding myself.

Altman was quick to address concerns that labor would be lost, adding, “Of course we will find new jobs and better jobs.”

Powerful AI tools do more than just increase speed. Rather, Altman believes that people “can be qualitatively different.”

It’s already happening. According to Altman, AI can now “make a programmer three times as fast as he is.”

“It’s not just that you can do three times as much. It’s that you can now think about completely different things at a higher level of abstraction and using more brain power.

Altman goes on to say, “Moving from punched cards to high-level languages ​​doesn’t just make programming a little faster. It allows us to do these qualitatively new things. We put it into practice. I’m looking at it.”

Altman identifies programming as “probably the single area that we’re most excited about today from a productivity improvement standpoint.”

And two other areas that Altman says are “curving” are education and health care.

But where does this lead?

“One day, there may be an AI that can say, ‘Please start and run this company for me.’ And one day, there may be an AI that can say, ‘Please discover new physics.’ I don’t know. What we’re seeing now is very exciting and great, but I think it’s always worth thinking about in the context of this technology, which is going to have a very steep improvement curve for at least the next five to 10 years. .

“These models are the stupidest ever.”

Gates said that “this is forcing us to adapt faster than ever before,” improving “very rapidly, with no ceiling” and that perhaps we really “are reaching the limit of human ability in many areas.” “There is a possibility that it could even reach that level,” he said. Even if it’s not doing any unique science. ”

Altman agreed, saying, “Each technological innovation is picking up speed, and this one will be the fastest yet.” That’s the part that I think is potentially a little scary is the speed at which society has to adapt and the labor market changes. ”

Gates asked Altman whether advances in robotics could endanger blue-collar jobs by creating mechanical “limbs with human-level capabilities.” .

Perhaps making it clear, Altman responded by saying that robotics is another area he’s “very excited about” and even admitting that “we’ve started investing a little bit in robotics companies.” I started.

Altman added that exciting new platforms are already being built for robotic hardware. “At some point, as you were saying, with models that have language understanding and future video understanding, we’ll be able to say, ‘Okay, let’s do some cool things with robots.'”

Gates warned that if developments continue as they are, “the job market for many blue-collar jobs could change quite quickly.” Altman said that was always expected, but somehow it wasn’t, noting that it was in white-collar and creative jobs that he first felt threatened by AI.

What do humans do?

This led to Gates’ confession: Of all the possible problems with advanced AI, “the one that baffles me is human purpose.”

For example, what if… “A machine says to me, “Bill, play pickleball. We’ve eradicated malaria. You’re just a slow thinker.”

Gates called it “philosophically confusing.” How do we organize society? ” And if AI can one day do all the work, what will schools need to teach children?

Altman, like Gates, gets philosophical because he derives satisfaction from his research on AI. “In a real sense, this may be the last hard thing for me to do.”

Altman added, “Their feel will be very different,” but… “I’m sure there will be no end to the problems. And we will find fulfillment and do things for each other.” , we will never run out of different ways to understand how to play the human game for other human beings, and in this way it will continue to be really important.”

“Certainly things will change, but I think the only solution is to get over it. We have to do this. It will happen. This is a technical course that can no longer be stopped. Value is too big. And I’m pretty confident, very confident, we’ll make it work, but I feel like it’s going to be very different.”

Perhaps the most surprising thing about their conversation was how much optimism came through. At one point, Gates even wondered if AI could help reduce wars. I want them to grapple with some of the most difficult questions as human beings, such as whether or not we can get along with each other. I think that’s a very positive thing. ”

Altman’s reaction? “I believe there’s going to be an upside there that will surprise us. This technology is going to surprise us with how much it can do.

“We’ll have to look into it and see, but I’m very optimistic. I agree with you, what a great contribution that is,” Altman said.

David Cassel is a proud resident of the San Francisco Bay Area, where he has been covering technology news for over 20 years. Over the years, his articles have appeared everywhere including CNN, MSNBC, Wall Street Journal and Interactive.

Read more from David Cassel