Thanks for this W&B

@WeightsBiases

3 ай бұрын

Our pleasure!

Thanks Joe Spisak

I really enjoyed this. Thanks

@WeightsBiases

3 ай бұрын

Glad you enjoyed it!

There's a universe where Joseph Spisak is Mark Zuckerberg's brother. Oh, and nice presentation. Wonderful work they are doing at Meta AI.

My favorite fact from this is that the smarter the model, the more it violates rules. Just like us :)

@utuberay007

3 ай бұрын

Very true ! People who are way smarter on tax laws are the one who violate most , innocent people pay more than what they are supposed to etc . Same goes with many other laws

@techpiller2558

3 ай бұрын

Or, the rules it uses instead of the rules we assumed are different.

@why.do.I.even.try.

3 ай бұрын

That's a great way to justify corruption and awful people.

@Crux69

3 ай бұрын

@@why.do.I.even.try. awful people are still human, best to understand how good people become awful

@why.do.I.even.try.

3 ай бұрын

@@Crux69 Yes but we shouldn't repeat their actions just because they work. We should work towards more ethical means to advance, technologically and societally.

Congratulations!

so all those supervisor/safeguard models are only utilized during training? i mean, once the weights of llama3 are out, there is no safeguard network between user and inference engine right?

@Crux69

3 ай бұрын

I'm sure they have a safety model that tries to review every request and catch some negative responses.

How much did it cost to build, including hardware and engineering costs?

What will be the SQLite of LLMs, with capability for local use? Llama?

I think he could have said "ridiculous" a bit more often

vin diesel!

Summary: Safety and size. The end.

why is mmlu still being used? it's broken

a few hours go by...llama 3 no longer SOTA

@sfsft11

3 ай бұрын

That's why they open source it. They let the community figure things out and iterate. For Meta LLM is just a tool and not a product on itself

@SkepticButOptimist

3 ай бұрын

Wait what is sota now?

@sfsft11

3 ай бұрын

@@SkepticButOptimist "state of the art"

@JeiShian

3 ай бұрын

Which model is sota?

@MiraPloy

3 ай бұрын

I think it's supposed ro be either phi or sensenova, neither of which are released ​@@JeiShian

How silly is to redteam a model which you control the training data to check for bioweapons capabilities. How stupid should you have to be? Isn’t easier to run a search on the data 😂😅

I’m glad they saw how useless they made codellama 😂, it was waaaay overly aligned