I build software to analyze big data from the LHC
Seeking a position in ML/DS
Math B.S. Phys B.S. Cal Poly
Phys Ph.D. University of Tennessee, expected May 2024
@AnthropicAI just turned on usage credits for Fable without informing us. I expect huge fallout this week from companies surprised by a massive credits bill after this move. The lack of communication is unacceptable.
This is a ridiculous assertion. These models are trained on human language and unless we develop a more efficient "language" to train models on, they will always use human language.
Top AI scientists from OpenAI and Anthropic sound a warning https://t.co/5CzMR21mi5
At @Bojangles drive thru and the new AI ordering is surprisingly effective, but pretty jarring. Not enough confirmation throughout the ordering process and too many upsells after each item I ordered.
I love agreeing to @WBHomeEnt ToS in Hogwarts Legacy and reading that I don't own the game, and my license to "download and install" the game is revocable π
@OpenAI@OpenAIDevs@OpenAINewsroom
Is anyone else experiencing issues with code interpreter sessions in the OpenAI Assistants API? Our sessions are failing with OpenBLAS(π€·ββοΏ½οΏ½) errors. Support chat has left much to be desired.
@JeffBezos said that LLMs are more like discoveries than inventions. I'd argue the opposite, that LLMs are our telescopes peering into the universe of language that we have codeveloped with the human experience. If an LLM can reason, then reasoning is probable in human language.
100 pages of content, 9 pages of references, 14 pages of appendices, and it's more or less done! What a journey my thesis has been. π #PhD#thesis#naptime
A thought as I code with #Copilot: With smart autocomplete, I tend to use longer variable names. This means that copilot has to process more tokens. In the long run, this will cost companies like @Microsoft and @OpenAI more per user. I think we fix this with AI control systems.