I love how this is a clear use case for hallucination. If the reason you’re generating an “answer” is merely to produce an embedding suitable for retrieving one, saying “I don’t know” is useless and confidently making up lies is better.
RESEARCH
-
HyDE: Generating Hypothetical Answers for Better Document Retrieval
By
–
Clever paper — HyDE: Hypothetical Document Embeddings Instead of encoding the user's query to retrieve relevant documents, generate a "hypothetical" answer and encode that. Documents with right answers more similar to wrong answers than to questions. https://
arxiv.org/abs/2212.10496 -
AI21 Labs Releases Paper on Grounded Text Generation and Fact Attribution
By
–
We are proud to announce the release of our latest scientific paper about improving grounded text generation and fact attribution – for any off-the-shelf LMs!
Read more details about this exciting development and its implications for the future of #AI: -
Stay Updated on Latest AI Developments and Industry News
By
–
Want to build the next big thing in AI but don't know where to start? Unwind AI's weekly newsletter is all you need to stay on the AI edge: Latest AI developments News from the Industry Tools of the Trade Hot Takes AI Meme of the Week https://
unwindai.substack.com/p/on-the-way-t
o-super-ai-ntelligence
… -
Nick Cave on ChatGPT: AI and Artist Confrontation
By
–
nick cave on chatgpt
the direct confrontation of ai models and artists was a fully unexpected development of 2022 and possibly one of the most through-provoking question in ai today https://
theredhandfiles.com/chat-gpt-what-
do-you-think/
… -
Fine-tuning in AI: Process Overview Without Deep Details
By
–
It just touches upon the fine-tuning part, and talks about the process but doesn't go into detail about the tips and tricks.
-
Internet’s Decline: LLM Output and Training Loop Crisis
By
–
The Internet’s twilight will come when it’s swamped with LLM output and new LLMs are trained on it.
-
Fantasy of Forking Into 1000 Parallel Threads
By
–
Rather, I fantasize about forking myself into 1000 parallel threads.
-
Absci’s Generative AI Advances De Novo Antibody Design
By
–
Absci’s Generative AI Approach Opens a Promising New Path for De Novo Antibody Design
-
Reinforcement Learning Agents Goals and Multimodal Integration
By
–
Online RL agents make interventions with a goal. I believe this is enough, but I’d like to hear @eliasbareinboim expanding on this as I believe he’s put a lot of thought into it. I also think multimodality helps, based on the rubber hand illusion, e.g. @anilkseth