AI plus for jobs, but needs "precision regulation": IBM chief

Artificial intelligence will create new business opportunities and lead to economic growth, which will create new jobs and more of them, according to IBM chief executive Arvind Krishna.

Krishna said he believes the net effect of AI in employment will be "job creation," rather than job destruction, although it will replace humans in certain types of work, Nikkei Asia reports.

Earlier this month Krishna stunned the world, telling business news outlet Bloomberg in an interview that about 30% of the roughly 26,000 "non-customer-facing jobs" at the information technology giant will be replaced by AI over the next five years. In an exclusive interview with Nikkei on Tuesday, he reiterated that certain routine jobs will become "unnecessary" for people to do, but also emphasized, "I am not saying jobs [overall] are going away."

He was in Japan to attend events and meetings on the sidelines of the Group of Seven summit meeting in Hiroshima, which ended Sunday.

Krishna explained that IBM saw its headcount rise by 2,000 during the first quarter of the year despite some 5,000 job cuts that took place during the same period. "You will add even more jobs in categories where there is more value creation. "So, fundamentally, there is going to be net job creation" with AI, he said.

More generally, he said the jobs humans do always change over time with the introduction of new technologies, recalling that machines eliminated many farming and laundry jobs, but created new ones such as truck driving in the 20th century. And with working-age populations declining in many societies, having humans do such routine tasks "is not an option. We are going to need technology to do some of the mundane work so that people can do higher-value work," Krishna said.

Regarding the risks posed by the recent rapid spread of large language model-based generative AI applications such as ChatGPT, Krishna disagreed with those who argue that they will become an existential threat to humankind, pointing out that writing and speech are "relatively simple use cases."

"Can it [AI] do reasoning? Can it do symbolic manipulation? None of these things are there yet," he said. "When people say it is an existential threat, they are trying to say this is artificial general intelligence," a type of all-purpose AI capable of thinking and executing tasks across a range of domains, just like a human brain. "But that is not true," Krishna said.

"There is no explanation deep inside them for their answers. They are very powerful, but I put them in the range of refinement, not revolution," he said. Instead of threatening humans, the generative AIs currently on offer can play a very useful role as societies age, he said.

Nonetheless, Krishna acknowledged that societies need to oversee AI, calling for "precision regulation" of different use cases rather than "blunt regulation" that attempts to restrict the technology as a whole.

"We are not worried about [a] raw algorithm. We are worried about risks lying in how it is used. We should regulate who can use it, how they use it, as opposed to stopping technology advancement itself," Krishna argued.

Last week, IBM's chief privacy and trust officer, Christina Montgomery, in testimony before the US Senate Subcommittee on Privacy, Technology, and the Law, recommended that Congress create a precise regulation framework where different rules are applied to different types of use cases and different risks, as well as disclosure requirements on data sources and technologies used, something analogous to nutrition labels on food products.

IBM executive Christina Montgomery testifies before a Senate subcommittee on AI oversight in Washington on May 16

Such a risk-based approach forms the basis of the draft of the European Union AI Act published in 2021, which EU leaders aim to implement within a few years. Krishna expressed support for the EU's approach, saying it is still viable and applicable to general-purpose language models like ChatGPT, which have a wide range of use cases.

Krishna emphasized that IBM's new business-focused AI platform, Watsonx, comes with strong risk-mitigation and safety functions. Companies can customize their AI applications on-site or in the cloud. He contrasted this approach with that of other big tech companies like Amazon.com, Google and Microsoft, which only allow business users access to their AI systems through their own cloud platforms.

One IBM client, a chemical company with 1 million pieces of literature accumulated over the past 70 years, does not "want to share it with the whole world," Krishna said. "We are not dictating that it has to be one business model."

Asked what led to IBM's decision last year to license its most-advanced two-nanometer chipmaking technology to Japan's new, government-supported semiconductor foundry venture Rapidus, Krishna shared his thinking: "Japan has enough semiconductor engineers. The capital and the land are there. So I actually, give this a very high chance of success," he said.

"I think many Japanese semiconductor engineers who are working in various parts of the globe outside the country are going to come back to Japan because of Rapidus. And those engineers are going to come to Albany, New York, to learn our technology and will go back to Hokkaido for the production launch. Once Rapidus is up and running, then I expect there will be even more collaboration," Krishna said.

Rapidus was founded in August last year as an all-Japanese chipmaking startup initially funded by eight major companies, including Toyota Motor, Sony Group, NTT, and SoftBank. The Japanese government is lending financial support to the venture as part of its "economic national security" strategy aimed at reinvigorating the country's domestic supply chain for advanced chips. Rapidus has decided to build its first chip fabrication plant near Sapporo on the northern island of Hokkaido.

Left to right, Rapidus Chairman Tetsuro Higashi, the company's president, Atsuyoshi Koike, and IBM Senior Vice President Dario Gil announce their partnership in December 2022

Other semiconductor players are stepping up investments in Japan as well. Taiwan Semiconductor Manufacturing Corp., the world's biggest chip foundry, is building a new fabrication plant in the Kumamoto prefecture, while Samsung has decided to build a research and development centre in Kanagawa prefecture near Tokyo. Intel is in talks to buy an Israeli company with chip fabrication plants in Japan.

Krishna pointed to these moves in Japan as aimed at lessening excessive dependence on Taiwan in the face of rising geopolitical tensions across the Taiwanese Strait.

If geopolitical tensions create choke points, "All of us, businesses and national leaders, need to create alternate sources [for] supply chains, and certainly what we are doing with Japan is a piece of that," Krishna said. Japan is one of the few places where "the US government would allow us to take this very advanced two-nanometer technology" given its defence pact with the United States, he added.

IBM announced on Sunday plans to invest $100 million in a joint research project with the University of Tokyo and the University of Chicago -- an effort to build a 100,000-qubit quantum supercomputer in 10 years. "Japan has a lot of expertise in quantum computing, [which it has] built up for decades," Krishna said.

"There are physicists, engineers and material scientists, all working on different aspects. Many people may not realize how deep Japan is on those areas," he said.

He predicted quantum computing technology will advance step by step through the joint project over the coming decade and will begin to see commercial applications within three to five years. "That is the time in which people will solve some problems that are likely too expensive or impossible for normal computers. We need 4,000 to 10,000 qubits for that," he added.

In quantum computing, a qubit or a quantum bit, is the basic unit of information, just like a binary bit in conventional computers.