[mjburgess]

Experts on AI arent experts on sociology, economics, politics, labour markets, jobs or their tasks,

I can't see anyone on that list who can provide an empirical account of the breakdown of major job roles into tasks which can be automated by chatgpt, but not prior to it. I cannot see anyone providing an analysis of what that automation will do for those roles. Nor anyone providing an economic analysis of that impact; a sociological analysis of that, and so on.

"Experts in AI" are people who can describe gradient descent without prep, balance a binary tree, and compute a dot-product.

[visarga]

> I cannot see anyone providing an analysis of what that automation will do for those roles

Here are two links:

> Separate studies of both writers and programmers find 50% increases in productivity with AI, and higher performance and satisfaction.

https://mobile.twitter.com/emollick/status/16313979316044881...

> Salespeople with AI assistance doubled customer purchases & were 2.33 times as successful in solving questions that required creativity. Top agents gained most.

https://mobile.twitter.com/emollick/status/16428856052383989...

[mjburgess]

Sure, and do 50% and 2.33x suggest that we're about to experience the mass obscelence of those roles? Hardly.

Spreadsheets had more impact

[JohnFen]

> "Experts in AI" are people who can describe gradient descent without prep, balance a binary tree, and compute a dot-product.

But that kind of expert is no more or less qualified to discuss the wider social ramifications of AI than any random person off the street.

[ryanwaggoner]

I think that’s the point they’re making…