March 12, 2012 - Foreign media outlet TechCrunch reported today thatIBM CEO Arvind Krishna has expressed his own belief that the AI Of significant value, but not foolproof.
Krishna disagrees with Anthropic CEO Dario Amodei's prediction that "the code for 90% will be generated by AI in the next three to six months".
"I think the percentage of code generated by AIMore likely 20% to 30%, much lower than 90%." Krishna said, "Certain simple tasks can be done by AI, but in more complex areas, AI It's probably useless.. "
He believes that AI is primarily a way to improveprogrammerThe efficiency of theWill not result in mass layoffs.
"If the programmer's jobEfficiency gains 30%, will the final code output be reduced?" He asked rhetorically. "History shows thatHigher productivityof companies typically capture a larger market share, which leads to the development of more products and further expansion of the business."
Krishna likens the fear of AI replacing human work to the controversy over calculators and Photoshop. He recognizes that the intellectual property issues surrounding AI training and content generation have yet to be resolved, but he believes that AI is ultimately a tool that canHuman empowermentThe tools.
"AI is a tool." Krishna said, "If it improves the quality of content, that's a boon for consumers as well."
He expects the cost of AI to continue to fall. While inference models like OpenAI's o1 still require a lot of arithmetic and energy, new technologies will make the energy consumption of AI computationDecreased below current 1%.
Krishna has talked about his views on DeepSeek. As previously reported by IT House, he said that it has always been a common belief that training state-of-the-art AI models requires an investment ofOver $1 billionand rely onThousands of the latest chips. But DeepSeek's approach proves that small, efficient models can also deliver real results without relying on large and expensive proprietary systems.
However, he doesn't think AI can drive scientific discovery. He believes the real breakthrough will come with quantum computing, an area in which IBM is investing heavily. "AI Just learning what you already know. It can't actively explore the unknown, and it's impossible to achieve a fully reliable 'general artificial intelligence'. It cannot answerBeyond Einstein or Oppenheimerof the problem."
In contrast, OpenAI CEO Aultman argues that superintelligent AI Possibly in the next few yearsemerge and will greatly drive innovation.