A pretrained LLM is not an AI but a simulator, described by a statistical physics based on internet webpages. The system evolves given any initial conditions (prompt). To gather logprob it internally maintains a probability distribution over what kind of document it is completing
LLMS
-
AI Alignment as Entropy Control in Language Model Simulators
By
–
In particular, "good, aligned, conversational AI" is just one of many possible different rollouts. Finetuning / alignment tries to "collapse" and control the entropy to that region of the simulator. Jailbreak prompts try to knock the state into other logprob ravines.
-
LLM Psychology Framework Discussion and Analysis
By
–
More good read/discussion on psychology of LLMs. I don't follow in full but imo it is barking up the right tree w.r.t. a framework for analysis. https://
lesswrong.com/posts/D7PumeYT
DPfBTp3i7/the-waluigi-effect-mega-post
… -
SambaNova Sponsors 29th Annual NLP Conference in Okinawa
By
–
SambaNova Systems co-sponsored the 29th Annual Conference of the Association for Natural Language Processing ( #NLP2023 ) to be held at the Okinawa Convention Center from March 13th to 17th.
-
Daily Python, Machine Learning and Language Models Content Sharing
By
–
That's a wrap. Every day, I share and simplify complex concepts around Python, Machine Learning & Language Models. Follow me → @Sumanth_077 if you haven't already to ensure you don't miss that. Like/RT the first tweet to support my work and help this reach more people.
-
GPT-3.5 Algorithm Emulation Depth and Context Window Limits
By
–
GPT-3.5 can emulate algorithms up to certain depth, so in theory this should be possible. But I am not sure whether Stockfish code and caches isn’t already too large for the context window and and also what depth would need to be simulated.
-
Can Sydney Emulate Stockfish Chess Engine Code?
By
–
I don’t know much about Stockfish and how it works, but couldn’t be Sydney emulating Stockfish code in order to get the moves?
-
Democratizing Hardware for Large Language Models
By
–
Democratizing the hardware side of large language models
by @bendee983 @bdtechtalks Read more: https://
buff.ly/3Q0ImKe #MachineLearning #ArtificialIntelligence #ML #MI #DataScience cc: @bigdata @patrickgunz_ch @amuellerml -
ChatGPT Achieved One Million Users in Just Five Days
By
–
How fast can you reach to 1 Million Users? It only took 5 days for #ChatGPT. #GenerativeAI #AI #NLP #100DaysOfCode #DataScientists #WomeninScience #OpenAI
-
Top 10 Tech Trends 2023 Selected By Industry Experts
By
–
10 Industry Experts’ Picks For The Hottest #TechTrends Of 2023 via @forbes https://
forbes.com/sites/forbeste
chcouncil/2023/02/28/10-industry-experts-picks-for-the-hottest-tech-trends-of-2023/
…