For those who don't know me — I'm very optimistic about LLMs in general. The visible problems of today's LLMs will be fixed, and long-form LLM writing is obviously useful to many people today. At the same time, I think StackOverflow and HN have a point. ChatGPT replies suck.
LLMS
-
NYT think-pieces on LLM social faux pas expected by January
By
–
I expect NYT think-pieces about this and other LLM-created social faux pas before the end of January.
-
PaLM’s Efficient Code Generation with Reduced Python Data
By
–
Coding: writes code with natural language description (text-to-code), translates code from one language to another, and fixes compilation errors.
— AI Breakfast (@AiBreakfast) 24 décembre 2022
PaLM claims to do just as well with 50x(!) less python data in it's training set.
(That's like learning to read from just 2 books) pic.twitter.com/7lgvDOxAqACoding: writes code with natural language description (text-to-code), translates code from one language to another, and fixes compilation errors. PaLM claims to do just as well with 50x(!) less python data in it's training set. (That's like learning to read from just 2 books)
-
Google PaLM Explains Complex Scenarios with Multi-step Inference
By
–
Google also says PaLM can generate explicit explanations for scenarios that require a complex combination of multi-step logical inference, world knowledge, and deep language understanding. (For example, it can explain jokes that are made up on the spot)
-
PaLM Model Advances Multi-Step Reasoning with Chain-of-Thought Prompting
By
–
By combining model scale with chain-of-thought prompting, PaLM also claims to show breakthrough capabilities on reasoning tasks that require multi-step arithmetic (this is big… and an area where ChatGPT currently falls short) and common-sense reasoning:
-
Google PaLM’s AI Language Understanding and Generation Abilities
By
–
Google released some details about PaLM earlier this year:
— AI Breakfast (@AiBreakfast) 24 décembre 2022
PaLM claims "impressive" natural language understanding and generation capabilities.
For example, it can distinguish cause and effect, understand conceptual combinations, and even guess a movie from an emoji set. pic.twitter.com/ktBjOgkOwTGoogle released some details about PaLM earlier this year: PaLM claims "impressive" natural language understanding and generation capabilities. For example, it can distinguish cause and effect, understand conceptual combinations, and even guess a movie from an emoji set.
-
Google Preparing PaLM Chat Model in Response to ChatGPT Growth
By
–
You may have heard the NYT report of Google issuing a "Code Red" after ChatGPT amassed 1 million users in 5 days, signaling an internal ramp-up for the deployment of their own Chat model: PaLM (Pathways Language Model)
-
Google’s PaLM Language Model Compared to ChatGPT
By
–
You've heard a lot about ChatGPT
— AI Breakfast (@AiBreakfast) 24 décembre 2022
But in the next few weeks, you'll start to hear about another model that could be 3x more powerful than ChatGPT:
PaLM
Google's Pathways Language Model (PaLM) a 540-billion parameter model trained with the Pathways system
Here's what we know 🧵 pic.twitter.com/M5V59Up7caYou've heard a lot about ChatGPT But in the next few weeks, you'll start to hear about another model that could be 3x more powerful than ChatGPT: PaLM Google's Pathways Language Model (PaLM) a 540-billion parameter model trained with the Pathways system Here's what we know
-
Summarization Quality and Hallucination in Davinci-3 Models
By
–
Comparisons would also help answer this question:
whether good summarization without hallucination warrants davinci-3 or not. -
Perplexity AI shows fewer hallucinations than Surge AI
By
–
Should make for an interesting comparison task @HelloSurgeAI
. From very few tries, it seems like @perplexity_ai hallucinates less.