IBM CEO: "AI will make programmers more productive, not eliminate their jobs"
IBM CEO Arvind Krishna says artificial intelligence is a valuable technology, but not a panacea.
IBM CEO Arvind Krishna says artificial intelligence is a valuable technology, but not a panacea.
IBM CEO Arvind Krishna says artificial intelligence is a valuable technology, but not a panacea.
He disagreed with a recent prediction by Dario Amodei, CEO of Anthropic, that 90% of code could be written by artificial intelligence within the next three to six months.
«I think it’s more like 20-30% of the code that can be written by AI, not 90%,» Krishna said. «Are there really simple use cases? Yes, but there are equally complex ones where that number is zero.»
Krishna believes, writes Techcrunch, that AI will ultimately make programmers more productive, increasing the results of their work and their employers, rather than eliminating jobs for programmers, as some critics of AI predict.
«If you can do 30% more code with the same number of people, will you write more code or less? History shows that the most productive company gains market share, and then you can produce more products, which allows you to gain more market share,» Krishna noted.
These statements also somewhat change the opinion of Krishna, who in 2023 said that IBM planned to suspend hiring employees to perform back-office functions, which the company expects to be able to replace with AI.
Krishna compared the debate about replacing workers with AI to the early debate about replacing mathematicians and artists with calculators and Photoshop. He acknowledged that there are «unresolved» intellectual property issues when it comes to AI training and its outcomes, but ultimately the technology is a positive force that augments human capabilities.
«This is a tool. If the quality of the products that everyone produces becomes better with these tools, then even for the consumer, now you are consuming better quality [products],» he noted.
He believes AI will become cheaper, although reasoning models like OpenAI’s o1 are computationally intensive and therefore energy-intensive. AI, Krishna believes, will use «less than 1%» of the energy it uses today, thanks to new technologies like those demonstrated by Chinese AI startup DeepSeek.
«I think DeepSeek gave us the idea that you can live with a much smaller model. The question is, do we still need any really big models to start with?» he asks rhetorically.
The CEO of IBM is not sure that AI will help humanity gain new knowledge, but quantum computing — a technology in which IBM is investing heavily — will be the key to accelerating scientific discoveries.
«AI learns from the knowledge it already has, literature, graphics, and so on. It doesn’t try to figure out what’s next… I don’t believe that the current generation of AI will lead us to what’s called artificial general intelligence… when AI has completely reliable knowledge and can answer questions that Einstein, Oppenheimer, or all the Nobel Prize winners combined couldn’t answer,» he is convinced.
Krishna’s claims contrast with those of OpenAI CEO Sam Altman, who argues that «superintelligent» AI is within reach within the next few years and could significantly accelerate innovation.
Previously, Mike Krieger, co-founder of Instagram and now Chief Product Officer at AI company Anthropic, said that the role of software engineers is rapidly changing, and that they will soon start reviewing «routine code» created by artificial intelligence, rather than writing it themselves.
Amazon Web Services CEO Matt Garman says the rapid emergence and adoption of AI could force software developers to stop coding, forcing them to upskill in the field to gain new skills.
Their statements come amid the growing popularity of the programming technique vibe-coding, which involves delegating the writing of code to AI chatbots based on user instructions. It was first described by Andrey Karpaty, a former OpenAI researcher and now director of artificial intelligence and computer vision for Autopilot at Tesla.
«I ask [the AI chatbot] the dumbest things, like ‘halve the sidebar padding,’ because I’m too lazy to figure it out myself. I always ‘Accept All,’ I don’t read the differences anymore,» he says.



