Mustafa Suleyman (Microsoft's AI chief) said just a few months ago that within 12–18 months AI would reach human-level performance on most professional tasks — lawyers, marketers, project managers, and so on — with all the implications that has for employment and the labor market.
Now, though, he's actively pushing Nobel laureate Daron Acemoglu's line: by his estimates, AI will automate only about 5% of human work over the next 10 years, and the GDP boost from the technology will come in at around 1.5% for the whole decade.
Commentators are calling this shift in rhetoric a sound and useful move: up to now both AI boosters and AI doomers had been competing to out-radicalize each other's predictions — and both camps ended up completely detached from reality. The gap between public panic and the fairly calm numbers from actual statistics kept growing too.
But now it's interesting to see whether the AI-apocalypse messengers can shift their tone as well :) Somehow I have my doubts.