Shamik Bose
@shamikbose.bsky.social
AI Researcher working on making black-box systems less opaque, in order to give a better life to my future dogs. He/him
Oh no! What was in your training data for GPT-4, OpenAI? No stolen/illegally collected text, I’m sure www.bbc.co.uk/news/article...
Also, are they really going to lobby the US government to disallow distilling? 😅
Also, are they really going to lobby the US government to disallow distilling? 😅
OpenAI says Chinese rivals using its work for their AI apps
ChatGPT maker says it will need extra protection from US government, following emergence of Chinese rival, DeepSeek.
www.bbc.co.uk
January 29, 2025 at 1:36 PM
Oh no! What was in your training data for GPT-4, OpenAI? No stolen/illegally collected text, I’m sure www.bbc.co.uk/news/article...
Also, are they really going to lobby the US government to disallow distilling? 😅
Also, are they really going to lobby the US government to disallow distilling? 😅
I find the US government’s interference with sale of AI chips really disturbing. Is this the same country that loves the free market, or that does only apply internally?
January 13, 2025 at 2:46 PM
I find the US government’s interference with sale of AI chips really disturbing. Is this the same country that loves the free market, or that does only apply internally?
Love reading any papers coming out of @eleutherai.bsky.social! The clarity of thought, the completeness of the experiments and the simplicity of the writing are all top-class! 👌🏽Currently reading “LLM Circuit Analyses .. across training and scale”
December 10, 2024 at 5:00 PM
Love reading any papers coming out of @eleutherai.bsky.social! The clarity of thought, the completeness of the experiments and the simplicity of the writing are all top-class! 👌🏽Currently reading “LLM Circuit Analyses .. across training and scale”
I just completed "Historian Hysteria" - Day 1 - Advent of Code 2024 #AdventOfCode adventofcode.com/2024/day/1
December 2, 2024 at 10:29 AM
I just completed "Historian Hysteria" - Day 1 - Advent of Code 2024 #AdventOfCode adventofcode.com/2024/day/1
I’m now convinced that “hallucinations” in LLMs was a very successful marketing campaign. It subtly suggests that the model has an understanding of the truth or an intention of hiding it, both of which are false.
November 22, 2024 at 11:18 AM
I’m now convinced that “hallucinations” in LLMs was a very successful marketing campaign. It subtly suggests that the model has an understanding of the truth or an intention of hiding it, both of which are false.