IBM CEO Krishna: Potential innovations in artificial intelligence cannot be regulated

Artificial intelligence will not make everyone unemployed, nor will it destroy the world.

Here’s what Arvind Krishna thinks. Last week, he held a lengthy discussion with Barron's about the company's approach to artificial intelligence.

The interview with Krishna is featured in Barron's' latest Tech Stock Trading column, which explains why IBM is an overlooked AI company whose 5% dividend yield could use a boost . However, not everything from the interview can fit into this column. Here are some other highlights from the conversation with IBM’s CEO.

About artificial intelligence and unemployment

Krishna said he was a little annoyed by a series of recent news reports quoting him as saying IBM could replace 7,800 employees with artificial intelligence software. Krishna said his comments were taken out of context. As the CEO explained, what he was actually saying was that 30 to 50 percent of repetitive white-collar jobs could be replaced by artificial intelligence within the next five years. He said the bottom range of this ratio was most likely to become a reality.

However, in a follow-up question, Krishna said he was asked how many IBM employees fit the description of "repetitive white-collar workers," and he estimated the total to be about 10 percent. IBM currently has about 78,000 employees, of which 7,800 account for about 30%.

But he said that calculation misses another key piece of information.

Krishna said: "I also said that artificial intelligence will create more jobs, just like the agricultural revolution and the service revolution created far more jobs than those that were taken away" . He believes that as AI tools are created, we will inevitably need new "prompt engineers" to take full advantage of AI tools, as well as more fact-checking to address the acceleration of misinformation.

About Artificial Intelligence and Potential Regulatory Risks

Barron’s asked Krishna about some tech experts’ assertions that artificial intelligence poses an existential threat to humanity—that it has the potential to become smarter than humans and somehow wipe us off the face of the earth. .

He didn't believe it.

He said: "I don't agree with the fundamentals of this view" and thought it seemed to be a big inference. The memory and pattern matching of these things are very good, but they don't have knowledge representation yet, they don't have any symbolic manipulation, they do math problems through memory, not through understanding, and there is still a lot of work to do Do. ”

Krishna noted that people are using nightmare scenarios to call for strict regulation of artificial intelligence. While he agrees that we should be careful about how and where we use these models, he believes there is no point in overly strict regulation. “It’s actually silly that we don’t want to regulate innovation,” he said. You can provide an advantage to those who ignore regulations and those who work outside the United States. ”

He suggested comparing AI regulation to the way regulators develop chips. He asked himself whether he had made any formal investments in future transistors, materials and nanotechnology, and replied: "Of course not." But we do have rules about who can buy chips and for what purpose. You can't police potential innovation. I just don't understand what this accomplishes. ”

About the potential combination of quantum computing and artificial intelligence

As Barron’s noted in a cover story on quantum computing last fall, IBM is arguably the leader in this emerging field. But the company already has 20 quantum computers in use, and Krishna believes we are not far away from the day when quantum computing and artificial intelligence will collide, which will open up new areas of computing power.

Krishna suggests one way to explore this idea is to consider the future of the chemical and pharmaceutical industries. “Maybe AI will find some knowledge gaps after reading through all the literature,” he said. “Today, you can do a wet lab experiment that may take three to six months or even longer. Fill in these gaps. Within the next three to five years, he said, quantum computers will have the ability to simulate these experiments and fill in the gaps, and it will only take a few minutes.

According to Krishna’s prediction, within the next decade, quantum computers will be able to generate artificial intelligence models. It is worth noting that one of the powers of quantum computing is to process large amounts of data at the same time. Training a model with tens of billions of parameters on a huge GPU cluster can currently take two to three months. With a quantum computer, you can train the same model overnight," Krishna said. "You will be able to solve problems that are far beyond what the world's largest supercomputers can solve. ”

Text|Eric Savitz

Editor|Yu Zhou

Copyright Notice:

Except with permission, no reproduction is allowed without the authorization of the original article of "Barron's" (barronschina). For the English version, see the report "AI Will Actually Create, Not Destroy, Jobs, IBM CEO Says." on June 6, 2023.

(The content of this article is for reference only and does not constitute any form of investment or financial advice; the market is risky, so investment must be cautious.)

The above is the detailed content of IBM CEO Krishna: Potential innovations in artificial intelligence cannot be regulated. For more information, please follow other related articles on the PHP Chinese website!