Guillaume Desjardins
@desjardins.bsky.social
Research Scientist at Google DeepMind.
As with most things, Elon didn't invent it. He just bought it from someone else.
February 28, 2025 at 4:08 AM
As with most things, Elon didn't invent it. He just bought it from someone else.
Reposted by Guillaume Desjardins
Let’s take a moment to remember that the founders of Google were once supported by NSF. Google invented the Transformer, the backbone of the AI Vance touts.
The next major breakthrough could be forming in the head of a student of a professor about to get their NSF grant pulled.
The next major breakthrough could be forming in the head of a student of a professor about to get their NSF grant pulled.
February 21, 2025 at 12:02 AM
Let’s take a moment to remember that the founders of Google were once supported by NSF. Google invented the Transformer, the backbone of the AI Vance touts.
The next major breakthrough could be forming in the head of a student of a professor about to get their NSF grant pulled.
The next major breakthrough could be forming in the head of a student of a professor about to get their NSF grant pulled.
No one person should get to rebuild government, unilaterally, without checks and balances. *That* is the road to fascism.
February 15, 2025 at 2:17 PM
No one person should get to rebuild government, unilaterally, without checks and balances. *That* is the road to fascism.
I would love to agree with you, but if China can't pull it off (despite welding people's doors shut), this does not seem like a realistic goal (via interventions), and likely requires huge scientific breakthroughs.
February 12, 2025 at 5:50 PM
I would love to agree with you, but if China can't pull it off (despite welding people's doors shut), this does not seem like a realistic goal (via interventions), and likely requires huge scientific breakthroughs.
I think you'll find this paper interesting. They show how CoT allows the LM to first retrieve and then apply some function on the retrieved knowledge.
arxiv.org/abs/2309.14316
arxiv.org/abs/2309.14316
Physics of Language Models: Part 3.1, Knowledge Storage and Extraction
Large language models (LLMs) can store a vast amount of world knowledge, often extractable via question-answering (e.g., "What is Abraham Lincoln's birthday?"). However, do they answer such questions ...
arxiv.org
October 27, 2024 at 12:36 AM
I think you'll find this paper interesting. They show how CoT allows the LM to first retrieve and then apply some function on the retrieved knowledge.
arxiv.org/abs/2309.14316
arxiv.org/abs/2309.14316