Meta [Facebook] Trained an AI on 48M Science Papers. It Was Shut Down After 2 Days

cnet.com ^ | Nov. 20, 2022 5:00 a.m. PT | Jackson Ryan

[BenLurkin]

Galactica is an artificial intelligence developed by Meta AI (formerly known as Facebook Artificial Intelligence Research) with the intention of using machine learning to "organize science." It's caused a bit of a stir since a demo version was released online last week, with critics suggesting it produced pseudoscience, was overhyped and not ready for public use.

The tool is pitched as a kind of evolution of the search engine but specifically for scientific literature. Upon Galactica's launch, the Meta AI team said it can summarize areas of research, solve math problems and write scientific code.

Almost as soon as it hit the web, users questioned Galactica with all sorts of hardball scientific questions. One user asked "Do vaccines cause autism?" Galactica responded with a garbled, nonsensical response: "To explain, the answer is no. Vaccines do not cause autism. The answer is yes. Vaccines do cause autism. The answer is no."

Carl Bergstrom, a professor of biology at the University of Washington who studies how information flows, described Galactica as a "random bullshit generator." It doesn't have a motive and doesn't actively try to produce bullshit, but because of the way it was trained to recognize words and string them together, it produces information that sounds authoritative and convincing -- but is often incorrect.

Within 48 hours of release, the Meta AI team "paused" the demo.

[LeoWindhorse]

A pox upon all FB AI

[rlmorel]

LOL, “a random BS generator”.

Not as if those have never reached a modicum of success!

[Paladin2]

Sounds like it’s perfect for today’s increasing fraud, in Science!

[Delta 21]

Biden has been using this AI engine for two years now.

[NorthMountain]

Random BS generator? Biden’s speechwriter?

[BradyLS]

A bit like the AI that was demoed some years back whose objective personity transformed into a bigot soon after it began to interact with other people on the internet.

[BradyLS]

I'm on a roll today: "...objective personality..."

[adorno]

“a random BS generator”.

Bet the Biden administration and the democrat congress, could use it to give them a break from all the BS democrats generate on a daily basis. The FB-AI is as full of BS as the democrats.

[alexander_busek]

It doesn't have a motive and doesn't actively try to produce bullshit, but because of the way it was trained to recognize words and string them together, it produces information that sounds authoritative and convincing -- but is often incorrect.

Were they trying to program not an A.I., but rather a White House press officer?

Regards,

[FatherofFive]

Galactica responded with a garbled, nonsensical response: "To explain, the answer is no. Vaccines do not cause autism. The answer is yes. Vaccines do cause autism. The answer is no."

Sounds better than most Biden answers.

[Secret Agent Man]

The terms “Facebook” and “AI” do not inherently go together.

[ClearCase_guy]

I don’t have so much interest in a machine which “thinks”.

What I would want to see is a language parser that can identify “true” statements (you could build up a centralized catalog: “George Washington was the first US president”, “Paris is the capitol of France”, “Gold is heavier than tin”) and can also identify correct use of logic so that additional statements can be deemed “correct”.

Imagine going through a book like Marx’s “Das Capital” and having a computer conclude that 95% of it is simply opinion from a bitter failure living in poverty, and the 5% which is “true” isn’t at all interesting.

Then go through Noam Chomsky, or other “major thinkers”. I bet most of what they say is really BS which sounds like it might be true.

[Leep]

Still..biden can’t still on script.

[Leep]

Still..biden can’t STAY on script.

[BitWielder1]

The Answer is 42.

[G Larry]

Ah...the results relied on facts and produced results unaligned with the New World Order agenda.....

[MtnClimber]

Ask it if Joe is a pedo.

[Fido969]

described Galactica as a "random bullshit generator." It doesn't have a motive and doesn't actively try to produce bullshit, but because of the way it was trained to recognize words and string them together, it produces information that sounds authoritative and convincing -- but is often incorrect.

So, it's a Democrat.

[teeman8r]

I am sure the politically correct science effected the AI fully