Email or username:

Password:

Forgot your password?
david_chisnall's posts Post Back to profile
Top-level
David McMullin

@david_chisnall
It makes sense that average would represent an improvement for half the population. But very few people think they themselves are below average, except maybe in some specific area they don’t care much about. So who is the market for “become average”? My guess is it’s employers who assume all their workers are below average, and in any case don’t like paying them.

21 July at 11:35 | Wall-to-wall | Open on musicians.today
6 comments
J Miller

@mcmullin @david_chisnall

Also, those at or near average who can now get results with much less effort. And for things with a normal distribution, that’s a lot of people. Of course, for things that are important, it’s a disaster.

21 July at 12:33 | Open on mastodon.online
david_chisnall

@mcmullin I would interpret it somewhat more positively. Most people are good at a fairly small set of things. I've had four books published in English, but French is my second-best language and there's no way I'd be able to write a book in it. There are vast numbers of languages where I'm even worse, yet with a language model I may be able to at least be comprehensible, if not erudite.

Looking at things like painting or music composition, I'm well below the average of people who produce anything that would go into a training dataset.

Most people are bad at most things. Talented people are typically talented only in a small set of things. It takes time and practive to become really good at something. If, without AI, I can be good at a handful of things and completely useless at a lot of things, but with AI I can be good at the same set of things and mediocre in a lot of things, that's a win.

Unfortunately, so far, the set of things where it can get me up to even mediocrity is fairly small.

@mcmullin I would interpret it somewhat more positively. Most people are good at a fairly small set of things. I've had four books published in English, but French is my second-best language and there's no way I'd be able to write a book in it. There are vast numbers of languages where I'm even worse, yet with a language model I may be able to at least be comprehensible, if not erudite.

21 July at 12:35 | Open on infosec.exchange
David McMullin

@david_chisnall
I think your interpretation is correct. But I look at this mainly from the point of view of how these tools will be used for art and music. A mediocre drawing by a real person still conveys something of that person. Quality aside, bot “art” is empty in a way human art cannot be, no matter how badly it’s done.

21 July at 15:35 | Open on musicians.today
Matt Campbell

@david_chisnall @mcmullin That's not a win for the people, including merely mediocre people (which is the majority), who no longer have a job.

21 July at 16:04 | Open on toot.cafe
MadMatheMatiker

@mcmullin @david_chisnall I would think of employers perspective like this: For most everyday tasks, it is enough to have average performance. However, in few cases you need a performance as good as possible (e.g. to sort things out when the average did poorly). So, instead of hiring 5 slightly-above-average people, you hire 1-2 very good people and use AI for the standard tasks

21 July at 13:04 | Open on mathstodon.xyz
Latte macchiato :blobcoffee: :ablobcat_longlong:

@mcmullin@musicians.today @david_chisnall@infosec.exchange
Don't forget that it also reduces the effort needed for mediocrity, which is often all you need.

21 July at 15:42 | Open on plasmatrap.com
Go Up