@jxmnop

–

10 November 2023 1h37

people are getting hackathon fatigue

10 November 2023

Refusal Classifier Training with Supervised Data

By

–

10 November 2023 1h20

@wzhao_nlp is the expert on this and she said she thinks they have a classifier trained for refusal, probably just with some supervised data it would still work well even if it were pretty small (<1B params)

10 November 2023

Refusal Models Filter Chatbot Queries About NYC Businesses

By

–

09 November 2023 15h59

you can't! there's a refusal model that decides whether to route your question to the chatbot at all, so if you type in something that isn't about NYC businesses (lol) it'll immediately refuse

9 November 2023

NYC Small Business Chatbot Powered by GPT-4

By

–

09 November 2023 15h40

tired of paying OpenAI for GPT-4 API? the NYC Department of Small Business Services has your back! the NYC small business chatbot is powered by GPT-4, so equally capable just have to ask it information about operating a business in New York City first http://
chat.nyc.gov

9 November 2023

Cornell researchers address AI model updating challenges

By

–

08 November 2023 23h07

no but I know some people at Cornell have been working on the updating problem: http://
arxiv.org/abs/2307.10323

Researchers Working on Unsolved AI Challenge

By

–

08 November 2023 22h39

yeah we don't know how to do this yet but people are working on it: https://
arxiv.org/abs/2202.05262

Fine-tuning LLMs: Knowledge Integration and Efficient Updates

By

–

08 November 2023 18h36

lots of problems need to be solved before we have this:
– how to finetune LLMs on data so they “know” the data, and don’t just mimic it?
– how do we combine knowledge from pretraining on the universe with finetuning on small datasets?
– how do we efficiently update with new docs?

Vector Databases Replaced by Single Transformer Models

By

–

08 November 2023 18h36

prediction: every vector database will eventually be replaced with a single transformer a row for every individual datapoint is so old-school, feels outdated what would be better: a single differentiable blob, something that “knows” about all your data and can chat about it

Hybrid AI System Combining Search, Reasoning and Generation

By

–

08 November 2023 18h36

Imagine a chatbot and a vector db had a baby this thing can do search, sure, but it’s much more capable: can do search, reasoning, and generation, all at once