TNS
VOXPOP
Where are you using WebAssembly?
Wasm promises to let developers build once and run anywhere. Are you using it yet?
At work, for production apps
0%
At work, but not for production apps
0%
I don’t use WebAssembly but expect to when the technology matures
0%
I have no plans to use WebAssembly
0%
No plans and I get mad whenever I see the buzzword
0%
AI / FinOps / Tech Culture

Bill Gates Interviews Sam Altman about OpenAI’s Speedy Rise

The two tech titans agree: AI will probably eliminate your job, but it will probably also lead to more interesting work to do instead.
Jan 28th, 2024 6:00am by
Featued image for: Bill Gates Interviews Sam Altman about OpenAI’s Speedy Rise
Feature image from the blog of Bill Gates.

Microsoft co-founder Bill Gates has started a podcast, and earlier this month he posted a quite a scoop: an interview with OpenAI CEO Sam Altman. It’s a surprisingly well-informed discussion on the ramifications of the speed with which AI is developing. And it includes specific predictions for the future from two people who are very close to the work being done.

Most of it was recorded before Altman’s temporary removal from the company — although it includes some audio of Gates checking in after Altman’s return. The OpenAI CEO says, “I’m all right. It’s a very exciting time,” and the team “has never felt more productive or more optimistic or better.”

“In some sense, this was like a real moment of growing up for us. We are very motivated to become better, and sort of to become a company ready for the challenges in front of us,” Altman said.

Bill Gates left Microsoft’s board in 2020, but BNN Bloomberg’s Jon Erlichman writes that since then, Gates has been working very closely with their AI teams, according to Microsoft CTO Kevin Scott. So, together, Gates and Altman had a conversation that ranged far into the future, with more than its share of thought-provoking pronouncements.

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

“It sounds so clichéd, and every company says it, but people feel the mission so deeply. Everyone wants to be in the room for the creation of artificial general intelligence.”

What Happens Next?

Bill Gates asks what milestones ChatGPT will hit over the next two years, and Sam Altman obliged with his best predictions — starting with “multimodality.”

GATES: Which means “speech in, speech out?”

ALTMAN: Speech in, speech out. Images. Eventually video. Clearly, people really want that.

Even the image and audio capabilities of ChatGPT drew a “much stronger response than we expected,” Altman says.

Beyond multimodality, Altman also predicted better reliability and “reasoning ability.” But in addition to that, Altman sees personalization and “customizability” becoming more important over the next two years — including not just different styles, but also different “sets of assumptions,” possibly derived from training on user-supplied data.

The Speed of Progress — and Cracking the Encoding

In an earlier podcast, Gates described the “black box” problem — how we don’t really understand how knowledge gets encoded, and the mysterious black art of prompt engineering.

Gates returned to the topic in his interview with Altman, who expressed “100%” confidence that we’d finally understand that with AI. “Trying to do this in a human brain is very hard,” Altman said. “We’re not going to slice up your brain and watch how it’s evolving.” But with AI … “we can perfectly X-ray. There has been some very good work on interpretability, and I think there will be more over time.”

But what’s even more interesting is that Altman believes understanding this knowledge encoding will in turn help us build even better AIs that are more efficient, and more accurate. Right now, “We’re all motivated to really understand them, scientific curiosity aside — but the scale of these is so vast.”

And this leads to a revealing insight into just how progress happens. Altman notes that it’s a common pattern in technology history, with a chance discovery preceding a more detailed understanding that comes later and leads to better performance.

And surprisingly, ChatGPT was just such a discovery. “The guy that built GPT-1 sort of did it off by himself and solved this — and it was somewhat impressive, but no deep understanding of how it worked or why it worked,” Altman said. “It really came from a place of empirical results first.”

While some predict AI will be too expensive to deploy widely, Altman points out that the cost reductions have already been enormous. “For GPT-3 we’ve been able to bring the cost down by a factor of 40.”

In fact, he takes his optimistic projections even further. “I think we are on the steepest curve of cost reduction ever of any technology I know — way better than Moore’s Law.” Models are not only more efficient, but also capable — with smaller models — as the research becomes better understood.

“I think we are going to drive the cost of intelligence down to so close to zero that it will be this before-and-after transformation for society.”

What Happens to Labor?

Where does that leave us humans? Altman reminds Gates that AI is on a “long, continuous curve” of improvement, and, “Eventually, they will be able to do more things that we think of like a job today.”

Immediately addressing concerns that labor might disappear, Altman adds, “We will, of course, find new jobs and better jobs.”

Powerful AI tools will not just make us faster. Instead, Altman believes, people “can do qualitatively different things.”

It’s already happening. AI can now “speed up a programmer 3x,” according to Altman.

“It’s not just that they can do three times more stuff. It’s that — at that higher level of abstraction, using more of their brainpower — they can now think of totally different things.

Altman added, “Going from punch cards to higher-level languages didn’t just let us program a little faster. It let us do these qualitatively new things. We’re really seeing that.”

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

And two other areas Altman says are “coming up the curve” are education and health care.

But where does this all lead?

“Someday, maybe there’s an AI where you can say, ‘Go start and run this company for me.’ And then someday, there’s maybe an AI where you can say, ‘Go discover new physics.’ The stuff that we’re seeing now is very exciting and wonderful, but I think it’s worth always putting it in context of this technology that, at least for the next five or ten years, will be on a very steep improvement curve.

“These are the stupidest the models will ever be.”

Gates expresses a concern that “it’ll force us to adapt faster than we’ve had to ever before,” improving “very rapidly, and there’s kind of no upper bound,” possibly even truly achieving “human levels on a lot of areas of work, even if it’s not doing unique science.”

Altman concurred: “Each technological revolution has gotten faster, and this will be the fastest by far. That’s the part that I find potentially a little scary, is the speed with which society is going to have to adapt, and that the labor market will change.”

Gates asked Altman whether blue-collar jobs could be jeopardized by robotics advances producing mechanical “hands and feet that are at human-level capability.”

Perhaps revealingly, Altman starts his answer by saying robotics is another area he’s “super excited” about, even acknowledging, “We’ve started investing a little bit in robotics companies.”

There’s already exciting new platforms being built for robot hardware, Altman added. “At some point, we will be able to use our models, as you were saying, with their language understanding and future video understanding, to say, ‘All right, let’s do amazing things with a robot.’”

Gates pressed that if progress continues, “That could change the job market for a lot of the blue-collar type work, pretty rapidly.” Altman pointed out that that was always the prediction — but that somehow, instead, it was white-collar and creative work that first felt a threat from AI.

What Will Humans Do?

This led Gates to a confession: Of all the possible problems of an advanced AI … “The one that sort of befuddles me is human purpose.”

For example, what if … “The machine says to me, ‘Bill, go play pickleball; I’ve got malaria eradication. You’re just a slow thinker.’”

Gates called it … “a philosophically confusing thing. How do you organize society?” And if AI can one day perform all work, then what’s left for our schools to then teach our children to do?

Altman turns philosophical, because, like Gates, he draws satisfaction from the work he’s doing on AI. “In some real sense, this might be the last hard thing I ever do.”

 

“They will feel very different,” Altman added, but … “I’m confident we’re never going to run out of problems. And we’re never going to run out of different ways to find fulfillment and do things for each other and understand how we play our human games for other humans, in this way that’s going to remain really important.”

“It is going to be different for sure, but I think the only way out is through. We have to go do this thing. It’s going to happen. This is now an unstoppable technological course. The value is too great. And I’m pretty confident, very confident, we’ll make it work, but it does feel like it’s going to be so different.”

The biggest surprise in their conversation may be how much optimism came through. At one point Gates even wonders … “whether AI can help us go to war less. I’d love to have people working on the hardest human problems, like whether we get along with each other. I think that would be extremely positive.”

Altman’s response? “I believe that it will surprise us on the upside there. The technology will surprise us with how much it can do.

“We’ve got to find out and see, but I’m very optimistic. I agree with you, what a contribution that would be,” Altman said.

Group Created with Sketch.
THE NEW STACK UPDATE A newsletter digest of the week’s most important stories & analyses.