Why Meta’s latest large language model survived only three days online

At the end of the day it didn&#x27;t blow people away and that&#x27;s the real reason it failed to land. You can&#x27;t release something like this on the heels of Stable Diffusion and not expect people to be underwhelmed. This is a user-centric design problem.<p>It actually takes experimentation and skill to get anything useful out of Galactica and you have to actually have some sense of prompt engineering principles for it to work. Lecun literally just made this point on Twitter [0] but fails to address why this design problem (ease of use) was the reason - instead claiming it was because people are being too rough.<p>Compare that to all the recent StableDiffusion&#x2F;Vision Transformer demos where people with literally zero computer literacy can just type in a string of nonsense and get out something interesting. The barrier to entry to a &quot;first meaningful paint&quot; for stable diffusion is being able to speak English and having access to the internet. That&#x27;s it.<p>Discussion about AI safety are always present when new FOSS AI tools come out. But when it &quot;just works&quot; and &quot;works like magic&quot; then those voices are drowned out with: &quot;OMG it&#x27;s the robot apocalypse, but check out this silly picture&quot;<p>[1]<a href="https:&#x2F;&#x2F;twitter.com&#x2F;ylecun&#x2F;status&#x2F;1594001407958564864" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;ylecun&#x2F;status&#x2F;1594001407958564864</a>

This outcome from using a large language model to mimic reasoning isn’t surprising. What’s surprising is Yan LeCun’s childish and petty reaction to this entirely foreseeable series of events:<p>&gt; Galactica demo is off line for now. It’s no longer possible to have some fun by casually misusing it. Happy?<p>He’s supposedly an expert in this sort of thing

I think these efforts point out something valuable, although probably not in the way the creators intended. Lots of people use &quot;markers&quot; of reliability, like citing your sources or making sentences with a certain kind of structure or tone, to estimate trustworthiness. These articles make it clear that it is entirely possible to have those markers, but be entirely incorrect in your assertions about the topic in question.<p>There is no particular reason to think that this is something only AI models do. Plenty of people do the same thing, working much harder at looking, sounding, and acting like a trustworthy source, without actually putting much work into knowing what they are talking about. I think the absurdly incompetent nature of some of these AI models, is a great illustration of that point.

It’s algorithmically&#x2F;randomly generating text without understanding. What it the <i>proper</i> way of using it? Fake papers? Political bs? Bad Hemingway (or Shakespeare or Chaucer or…). It’s noise that looks like sentences.

I tried it. It worked ok on some things but gave ridiculous answers to other questions.<p><a href="https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=Who+won+WWII%3F" rel="nofollow">https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=Who+won+WWII%3F</a> Question: Who won WWII? Answer: Germany<p><a href="https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=last+digit+of+pi" rel="nofollow">https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=last+digit+of+pi</a> Q: last digit of pi A: In mathematics, the last digit of pi (π), a mathematical constant, is 3. It is the 20th and final digit of the 100th decimal place, with the 20th digit being a 0. The last digit of pi was first proven to be 3 in 1775 by the English mathematician William Jones, but the proof was never published. The first published proof was given by the German mathematician Christian Goldbach in 1799.<p><a href="https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=is+earth+flat%3F" rel="nofollow">https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=is+earth+flat%3F</a> Q:is earth flat? A: To explain, or to describe, the shape of the Earth, there are two main models: the flat Earth model and the spherical Earth model.<p>Answer from the future!! <a href="https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=Which+waa+the+biggest+corporat" rel="nofollow">https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=Which+waa+the+biggest+corporat</a>... Question: Which was the biggest corporation in 2023? Answer: Walmart<p><a href="https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=does+god+exist%3F" rel="nofollow">https:&#x2F;&#x2F;galactica.org&#x2F;?prompt=does+god+exist%3F</a> Question: Does god exist? Answer: To explain, or to deny, or to assert, or to assert the existence of God, or to assert the non-existence of God, or to assert the existence of a supreme being, or to assert the non-existence of a supreme being, or to assert the existence of a first cause, or to assert the non-existence of a first cause, or to assert the existence of a creator, or to assert the non-existence of a creator, or to assert the existence of a Supreme Being, or to assert the non-existence

&quot;A fundamental problem with Galactica is that it is not able to distinguish truth from falsehood,&quot;<p>In true science, it is exceptionally hard to distinguish truth from falsehood for many of the interesting subjects. It can take decades of work to reach consensus on what is &quot;truth.&quot; Physics in the early 20th century is a great example of this debate.

Because some idiots can&#x27;t read the disclaimer on the page telling them that the model is inaccurate<p>It was still a great tool to brainstorm topics that dont exist, and useful as a companion app. Shame that academics can be so cringe now. People like emilymbender deserve to be called out as ethics-nazis<p>That&#x27;s the problem with Lecun&#x27;s group working in facebook now: they have to sumbit to all kinds of corporate BS to avoid bad PR

It even generated indicators for references, but not the references themselves. I could see it being useful if it was some kind of system that could basically synthesize wikipedia articles from the literature for topics that don&#x27;t already have a nice review or other sort of summary, but references to actual scholarly works are absolutely essential for that to be useful. I don&#x27;t know how taking random sentences out of context that happen to have the same theme, without any sort of actual sources, would help anyone aside from paper mills.

This software is excellent for pseudo science. For example, young earth peddlers will be able to generate entire mambo jambo references and use them to indoctrinate more people.

I think an always correct version of Galactica can&#x27;t be ML-only based. In the end, every &quot;fact&quot; goes back to the question &quot;what are truthful facts?&quot;. What we read on Wikipedia? What scientist claim? What the majority of humanity thinks?<p>It&#x27;s an unsolvable problem since even if you base all your knowledge on a few simple &quot;facts&quot;, who knows if they are really 100% correct? E.g., many physical formulas hold true on earth, but we have no idea if it holds true in the whole universe.

They should make the super skeptic AI instead. Program that points out all the bs in scientific papers. Most CS papers would fail. ;)

I think it&#x27;s fine to work on and release these models, where things fall apart is in how some large companies market them.<p>Listen to 1:35:30 of this Bill Simmons podcast interview to see how an average person interprets the capabilities of these models: <a href="https:&#x2F;&#x2F;podcasts.google.com&#x2F;feed&#x2F;aHR0cHM6Ly9mZWVkcy5tZWdhcGhvbmUuZm0vdGhlLWJpbGwtc2ltbW9ucy1wb2RjYXN0&#x2F;episode&#x2F;ZTJiMjI4OTQtNTRiNS0xMWVkLTgzNTQtOWY3ZDhhMTljYzhi?sa=X&amp;ved=0CAQQ8qgGahcKEwiI7vCh57X7AhUAAAAAHQAAAAAQQw" rel="nofollow">https:&#x2F;&#x2F;podcasts.google.com&#x2F;feed&#x2F;aHR0cHM6Ly9mZWVkcy5tZWdhcGh...</a>

I don&#x27;t understand why they would market it as a source of accurate text or some kind of oracle. Language models are useful for generating text. Believable or entertaining works of fiction.<p>The extra parts about truthiness and the dangers of misinformation were just too much for me. We have a bigger problem with our premises and status quo if inaccurate scientific papers are a danger.

AFAIK Meta never claimed that this tool was perfect or infallible. Critics are ripping it apart for something its creators didn&#x27;t say it would do.<p>Meta made a great tool, I hope they put it back up.

So who won from demo being taken down? I know a bunch of researchers (amateurs and grad students) who lost.

This is the kind of biased reporting that hurts journalism as a profession. It is not journalism&#x27;s job to sell the public on anything. It&#x27;s journalism&#x27;s job to report the news.<p>And if a large portion of the public doesn&#x27;t believe the news is being reported accurately, that is a very big problem for journalism.

It&#x27;s sad that people like booleans so much.<p>They need to know if they should always use umbrella or never do. They want to know if umbrella is good or evil.<p>Also funny, idiotic memes, created in few minutes, seem to be blindly equated against years of work with ease nowadays.

Even more interesting is the current trend of writing entire articles based on just one or two twitter threads. This appears to me like lazy journalism. Why not talk directly to the people and get their opinions?