Kaiser Pister

AI advising, CTO @ ALL3D, Faculty @ UW Madison

[Kaiser Pister @KaiserPister]

@BenjDicken seems pretty biased to say python = slow, js = fast for interpreted languages. a fast python interpreter is fast. a fast js interpreter is fast. slow interpreters are slow.

[Kaiser Pister @KaiserPister]

@seanmcdonaldxyz @Jac5Connor https://t.co/XWgNRtrXJ0

[Kaiser Pister @KaiserPister]

@sh_reya Doesn't seem particularly hot. We won't have RAG in 10 months, not 10 years.

[Kaiser Pister @KaiserPister]

The keyboard shortcut for "search" and "launch vs code editor in your browser, which loads for 30 seconds before interactive" are too damn close on @github #GitHub

[Kaiser Pister @KaiserPister]

@annarmonaco But I don't want 1000 interns, I just want a few good FTE

[Kaiser Pister @KaiserPister]

@leopd This is just the state of affairs in research, not sure if it a hot take

[Kaiser Pister @KaiserPister]

I wish there were a way to automatically clean up research code so that it is readable

[Kaiser Pister @KaiserPister]

@auchenberg but then what is the point of 2fa if you don't need the 2. This is just normal auth now

[MTGRemy @MTGRemy]

Do not be discouraged if someone dismisses your dreams.

There might be 100,000 people out there who disagree. 🥳

[Kaiser Pister @KaiserPister]

@hwchase17 @levelsio What are you using the $35M for though?

[@levelsio @levelsio]

I recommend everyone against using LangChain and this article explains well

It uses abstractions on top of abstractions and actually makes your code needlessly complicated

Just write API calls and add a vector database instead

https://t.co/7scFP6O17t

[Kaiser Pister @KaiserPister]

Did 2FA just die with continuity? @Apple

[Kaiser Pister @KaiserPister]

@leopd Also your BS can be dynamic and adjust depending on the phase of learning your model is currently in. (per https://t.co/YSu3gr8wfS)

[Kaiser Pister @KaiserPister]

It was smart for Google to take the login-redesign opportunity to simultaneously break all Safari usage

[Christopher Potts @ChrisGPotts]

I know I am late in the project cycle for this, but I do have suggested edits for the team behind https://t.co/w4YH7FvOpo My overall comment is that the central claims in the original are lacking in empirical support.

[Kaiser Pister @KaiserPister]

@hwchase17 But why tho? If you move all the power out of langchain then what was the purpose of installing langchain? You have zero attack surface if you publish an empty package 🤔

[Kaiser Pister @KaiserPister]

@Kangwook_Lee Does the correctness of the labels actually matter? Work out of Meta from '22 seemed to show that the key goal of ICL is to shift the generation distribution towards the final task regardless of the actual values used. Curious how 2-mode plays with that (https://t.co/q2asLYLq96)

[LLM4Code @llm4code]

📢 The #LLM4Code program and accepted papers are available now at https://t.co/LyVNzNYHXK

Congratulations to the authors again! We look forward to meeting everyone at LLM4Code @ICSEconf! 🚀

[Kaiser Pister @KaiserPister]

@degourab @wandb Perhaps it was the lightning logger that wrapped wandb. If you send me a new API key that is "near the limit" I'd be happy to rerun the training run to see what happens.

[Kaiser Pister @KaiserPister]

@wandb just crashed by 48 hour experiment because my free trial account ran out of hours. I get it, you want me pay you, but maybe gracefully handle that error?