Yann LeCun
• 2nd
VP & Chief AI Scientist at Meta.1point3acres
2h • 2 hours ago
My unwavering opinion on current (auto-regressive) LLMs
1. They are useful as writing aids.
2. They are "reactive" & don't plan nor reason.
3. They make stuff up or retrieve stuff approximately.
4. That can be mitigated but not fixed by human feedback.
5. Better systems will come.
6. Current LLMs should be used as writing aids, not much more.
7. Marrying them with tools such as search engines is highly non trivial.
8. There *will* be better systems that are factual, non toxic, and controllable. They just won't be auto-regressive LLMs.
9.I have been consistent with the above while defending Galactica as a scientific writing aid.
10. Warning folks that AR-LLMs make stuff up and should not be used to get factual advice.
11. Warning that only a small superficial portion of human knowledge can ever be captured by LLMs.
12. Being clear that better system will be appearing, but they will be based on different principles..1point3acres
They will not be auto-regressive LLMs.. 1point3acres
13. Why do LLMs appear much better at generating code than generating general text?. 1point3acres
Because, unlike the real world, the universe that a program manipulates
(the state of the variables) is limited, discrete, deterministic, and
fully observable.
The real world is none of that. ..
14. Unlike what the most acerbic critics of Galactica have claimed. 1point3acres.com
- LLMs *are* being used as writing aids.
- They *will not* destroy the fabric of society by causing the mindless masses to believe their made-up nonsense.
- People will use them for what they are helpful with.
Last updated 4 Aug 11 Course Title: OWASP Top 10 Threats and Mitigation Exam Questions - Single Select 1) Which of the following consequences is most likely to occur due to an injection attack? Spoofing Cross-site request forgery Denial of service Correct Insecure direct object references 2) Your application is created using a language that does not support a clear distinction between code and data. Which vulnerability is most likely to occur in your application? Injection Correct Insecure direct object references Failure to restrict URL access Insufficient transport layer protection 3) Which of the following scenarios is most likely to cause an injection attack? Unvalidated input is embedded in an instruction stream. Correct Unvalidated input can be distinguished from valid instructions. A Web application does not validate a client’s access to a resource. A Web action performs an operation on behalf of the user without checking a shared sec
Comments
Post a Comment
https://gengwg.blogspot.com/