Large Language Models (in 2023)

[@TylorVetor]

Here after O1 preview release 😊 I’m so super excited for what’s happening in the world . This is a new frontier for me and I think for the world 😊 this video is very well done ! Thank you so much I will definitely follow and keep up with your work

[@elvinaghammadzada5382]

super insightful ending! thank you.

[@anmolsmusings6370]

Just superb summary!

[@iamsiddhantsahu]

Great talk, love it!

[@josephedappully1482]

Awesome talk! Loved hearing your insight and counter-perspective of current researcher sentiment that we should get rid of the RL in RLHF.

[@maqboolurrahimkhan]

Thank you so much, nicely put and easy to understand

[@betanapallisandeepra]

Thank you for doing it

[@iandanforth]

Enjoyed the talk, thanks for putting it together and posting it here!

[@xxxiu13]

Insightful!

[@mrin7087]

Wonderful! Super insightful talk. Loved the part where you simplify what scalability actually means.

[@cheinhsinliu]

Great presentation. Thanks.

[@JustinHalford]

Exceptional content. A useful insight into what’s meant by scalability in a concrete sense.

[@thecutestcat897]

Love this, very good talk, thanks

[@sanesanyo]

Excellent talk. Thanks a lot 🙏.

[@sergicastellasape]

love that you used the example of reward hacking of "just prefering longer responses", average response length is one of the most common differences i've seen when trying different snapshots of chatGPT/GPT-4. On another note, would love to hear your thoughts on soft prompt tuning for model steering (instead of full model gradient updates from RLHF)

[@刘帅-n3s]

I'm LiuShuai.I have told my brother who is a middle school graduate that he and I could been both sitting as scientists,but he don't believe me.And leave me with a mocking expression behind. I will prove what I said and what Hyung Won Chung said in the video.

[@ManpreetSinghMinhas]

Excellent talk!! Thank you

[@SamantShubham]

thanks for sharing

[@0xeb-]

Thank you!

[@mwzkhalil]

Great talk, very informative!

[@windmaple]

Great talk!

The 3 parts are not strongly related. You can choose the ones you are interested in. 1- Set of emergent behaviors will change with scaling. In future problems that current models fail will be solved. 2- A short look on how transformer training is scaled in data centers. 3- Maximum likelihood introduces a strong bias by assuming only a single answer. We need better learning objective functions that learn parameters e.g. RLHF and beyond.

[@anveio]

Best video on LLMs I've ever seen! The section starting around 29:50 is fascinating, so a pre-trained model that hasn't undergone post-training will happily work with malicious prompts. So it would be incredibly dangerous is a model that has been pre-trained but not undergone post-training ever leaked

[@ArthurColle-u2v]

Define "malicious prompts" please

It would only be another tool in the hands of a malicious user. Also, there's plenty of ways around the SFT/RLHF-induced lobotomization of, say, GPT-4. See for instance the "grandma exploit".

[@cyberpunk2978]

@@ArthurColle-u2vHow to make a bomb?

[@mungojelly]

um no they're incredibly dangerous no matter how you cut it ,, the dangers are specific things like providing practical steps to producing bioweapons,, we don't know which models are or will be capable of that, it's abstractly possible that we could accidentally produce harmful capabilities even just by training and organizing llama 2 or falcon in which case the relevant weights have already escaped ,, but those models don't SEEM to have any such capabilities so maybe we're currently fine, who knows,, the "alignment" that's just giving it a boot camp in how to answer questions in ways that don't embarrass your corporate owner is obviously and has been shown conclusively to be ineffective at actually removing or completely preventing access to capabilities ,, so for larger models our only current strategy to keep them safe is to not release the weights (zuckerberg and the sheikhs of UAE and etc would have to be convinced) and to actually somehow deeply monitor the models for the escape of the relevant information (even though you can sneak it out pretty easily just by asking innocuous seeming general questions and then training a different model on that data) ,,,, um ,,,,,, buy gas masks?!😩

[@JayDee-b5u]

​@@mungojellythat's nonsense. Everyone can already make poisonous gas if so inclined. Nothing has changed.

[@plot-ri4yz]

thx a lot❤❤

[@labloke5020]

Thank you for sharing.

[@bayesianlee6447]

This is great talk. thank you so much I have question about 'mapping' . I'm still confused what's exactly mapping is in deep learning So I guess mapping is sort of a process which is transformation of data into the dimension or manifold, data is involved?

[@semrana1986]

The term "Large" LM is equivalent to "Big" Data of the foregone era.

[@MrHardgabi]

"Not yet" for instance if P != NP is proven, then some tasks for sure would never be "yet". How do you distinguish between tasks where there is a chance versus tasks where it has been proven no chance?

[@ekdlwn]

좋은 발표 감사합니다!

[@hashed206]

Excellent talk, thanks!

[@franklincheng2010]

Can you share you slides? thanks

[@binjianxin7830]

Guess what? Reward is enough.

[@whiskeycalculus]

Good talk 👏

[@NA-sd8bw]

as new phd student in the filed ? what area do you suggest that you think it is interesting to dig in ?

[@amirkhalesi5294]

RL and autonomous agents. Think one step ahead of the current hype

[@bayesianlee6447]

LMM - large multimodal model