Even the scientists who build AI can’t tell you how it works

vox.com

7 points·ubj··4 comments

Utter nonsense non-journalism feeding the hype.

The mechanisms of current AI methods are not only well-understood, there are explainer videos that get posted here on the regular that can even explain it to you, the credulous Vox reader.

Indeed, the fact of the matter is that we have had this tech for a very long time, but we have dedicated our compute capacities to other important matters, such as fake money for criminals.

It is extremely comprehensible, and the people who are currently huffing VC money about it are trying to sell an image of wizards conjuring daemons to gull a Terminator-groomed public into begging the friendly capitalists to please control this monster and save us all from Skynet!

Embarrassing.

Also nonsense. We have a very good understanding of the building blocks but that does not mean we understand the whole system (hint: we know how to model fluids, but we don't really understand turbulence).

The interview itself is very balanced on this and does nothing of what you claim it does.

We understand what each part is goingy and how it's doing it.

You cannot look at the data it's processing and understand it, that is true, but that's also true of absolutely any frame from a video game or cartoon.

But hey, keep forcing that exuberance.

Disagree with this comment. See the Wikipedia page on Explainable AI [1]. Many models today fall under the "black box" category described on that page. Also note the last paragraph in the Criticism section:

> The goals of XAI amount to a form of lossy compression that will become less effective as AI models grow in their number of parameters. Along with other factors this leads to a theoretical limit for explainability.

[1]: https://en.m.wikipedia.org/wiki/Explainable_artificial_intel...

Reply on news.ycombinator.com