One of the boldest recent claims about artificial intelligence has come from SoftBank chief executive Masayoshi Son, who says OpenAI’s next major model is being designed largely with the help of AI rather than by human engineers.
In a recent interview, Son said he had learned this directly from OpenAI chief executive Sam Altman, and that it changed how he thinks about the future of the technology. He said he had previously assumed that artificial superintelligence was many years away, but now believes that stage may be only about two years off. Son described superintelligence as a system that could be thousands of times more capable than human intelligence.
Why the claim matters
The significance of the statement is not simply that a new model is coming. The deeper question is how fast progress could accelerate if AI begins designing the next generation of AI systems itself.
Traditionally, researching, designing and improving new models has been the work of human engineers. If machines take over that process, each new generation could improve the one that follows it more quickly than before. Some researchers call this idea “recursive self-improvement,” a process in which AI keeps increasing its own capabilities.
Reasons for caution
Critics warn that predictions of this kind should be treated carefully. Most claims about superintelligence so far are theoretical, and the practical evidence for an imminent arrival remains limited. Son’s timeline is a forecast, not a demonstrated result, and he has a direct financial interest in the companies he is describing.
A very large financial bet
One fact stands out in the debate. SoftBank has committed more than 64 billion dollars in investment to OpenAI, and Son has said the firm recorded roughly 25 billion dollars in gains on its AI investments in a single quarter. In other words, the world’s largest investors are not only talking about the future of AI; they are placing enormous financial bets on it.
A few years ago, the question was whether AI could help people. Increasingly, the question is what role humans will keep if AI begins to create its own next generation.
