Van GanzMar 31
Liam Proven

RE: https://mathstodon.xyz/@tao/115722360006034040

This is an extremely worrying conclusion to me.

Mar 31 at 8:02amWeb
Show threadMidgePhotoMar 31

@lproven
I've been looking at cars. Having a lot of general purpose computer in them and getting them to drive themselves on random roads still seems optimistic.
(I remember the Palm Pilot and Palm deciding it was easier to teach people an alphabet than computers to read)

My current car* has assorted reluctances to run into things and affordances to travel. Steadily adding specific gadgets to discourage particular failure modes is engineering we can understand.

* A Voltswagen ;)

Show threadChristian LynbechMar 31

@lproven @bsdphk I am not sure I understand what hsi conclusion is? Are he saying that the main problem with AI tools of today is that we had expected more and if we could let of go of those expectations, we would be able to understand the utility of said tools? In other words, skeptics are just holding it wrong?

His opening however does track. I do not know of anbybody, that does not have an obvious vested interest in the opposite, that believes current AI technologi trajectory leads to AGI.

Show threadLiam ProvenMar 31

@mapcar @bsdphk My take is this:

As the models get bigger and more complex, they'll be able to do more and increasingly "clever" tricks, without ever becoming any smarter. There is no smartness there to improve. Doubling zero still yields zero.

But this will get ever-harder to demonstrate to the True Believers, who think that they minds in boxes.

Show threadRichard CrawshawApr 1
@lproven @mapcar @bsdphk yes, "true believers" (in the capabilities of "AI"s) is going to be a long term problem. What he's saying makes sense and provides a way to usefully use them in the future, ignoring the environmental and moral issues. But the "true believers" will turn them into religion.
Show threadPoul-Henning KampApr 1

@richardcrawshaw @lproven @mapcar

For perspective:

There are still senior managers who long back to "the good ol' days of deregulation under Reagan and Thatcher"

There are still not quite a senior managers who think "exponential growth, like in the good ol' dot-com days" is something good to aspire to.

So yes, we will be battling the daemons of the AI-Bros for the rest of our life...