Hacker News: rustastra

Hallucination is a problem we'll have to live with for a long time

rustastra — Sat, 18 Jan 2025 09:28:07 +0000

Article URL: https://www.theregister.com/2025/01/07/interview_with_aws_byron_cook/

Comments URL: https://news.ycombinator.com/item?id=42747040

Points: 9

# Comments: 11

New comment by rustastra in "Automated reasoning to remove LLM hallucinations"

rustastra — Wed, 04 Dec 2024 03:03:56 +0000

Same with symbolic systems!

New comment by rustastra in "Automated reasoning to remove LLM hallucinations"

rustastra — Wed, 04 Dec 2024 01:28:47 +0000

aiui a natural language question e.g. "What is the refund policy?" gets matched against formalized contracts, and the relevant bit of the contract gets translated into natural language deterministically. At least this is the way I'd do it, but not sure how it actually works

Automated reasoning to remove LLM hallucinations

rustastra — Wed, 04 Dec 2024 00:45:41 +0000

Article URL: https://aws.amazon.com/blogs/aws/prevent-factual-errors-from-llm-hallucinations-with-mathematically-sound-automated-reasoning-checks-preview/

Comments URL: https://news.ycombinator.com/item?id=42313401

Points: 57

# Comments: 38

New comment by rustastra in "Happy Birthday ChatGPT: Looking Forward from Here"

rustastra — Wed, 29 Nov 2023 19:26:40 +0000

I couldn't agree more with this take: > Our demand for stimulating content is being overtaken by supply. Analogously, with AI, we might be in a world where scientific progress is accelerated beyond our wildest dreams, where we have more answers than questions, and where we cannot even process the set of answers available to us.

New comment by rustastra in "Learnings from employing ChatGPT as a ML Engineer for a day"

rustastra — Wed, 22 Feb 2023 20:54:25 +0000

Curious to learn more about prompt engineering takeaways here. Was feeding more context (or chapters of textbooks, bits of papers, documentation) helpful? It does seem like layering information and being very precise helps a lot. Eerily like with interns

New comment by rustastra in "Launch HN: Cord (YC W21) – training data toolbox for computer vision"

rustastra — Thu, 11 Feb 2021 20:11:20 +0000

As a user, am I expected to write the labeling algorithms myself or do you offer some in-built ones?

New comment by rustastra in "Label a Dataset with a Few Lines of Code"

rustastra — Tue, 19 Jan 2021 00:48:35 +0000

Could feature extraction be helpful here?

New comment by rustastra in "Label a Dataset with a Few Lines of Code"

rustastra — Tue, 19 Jan 2021 00:32:39 +0000

I am very curious to know which pre-trained models work better for this task and whether it's possible at all to do without a neural net...