“The general problem of mixing data with commands is at the root of many of our computer security vulnerabilities.” Great explainer by security researcher Bruce Schneier on why large language models may not be a great choice for tasks like processing your emails. https://cacm.acm.org/opinion/llms-data-control-path-insecurity/
"When the Singaporean government asked local writers if they would agree to having their work used to train a large language model, it probably did not expect the country’s tiny literary community to react so fiercely."
I just issued a data deletion request to #StackOverflow to erase all of the associations between my name and the questions, answers and comments I have on the platform.
One of the key ways in which #RAG works to supplement #LLMs is based on proven associations. Higher ranked Stack Overflow members' answers will carry more weight in any #LLM that is produced.
By asking for my name to be disassociated from the textual data, it removes a semantic relationship that is helpful for determining which tokens of text to use in an #LLM.
If you sell out your user base without consultation, expect a backlash.
@sean Good questions. The way I see a RAG being constructed / or other knowledge graph would be to associate Contributors with Questions and Answers - so you need the Question Answer relationship to generate plausible answers, but the Contributor Answer relationship lets you rank Answers higher from higher rated contributors:
See something like this:
He, Xiaoxin, Yijun Tian, Yifei Sun, Nitesh V. Chawla, Thomas Laurent, Yann LeCun, Xavier Bresson, and Bryan Hooi. "G-Retriever: Retrieval-Augmented Generation for Textual Graph Understanding and Question Answering." arXiv preprint arXiv:2402.07630 (2024).
@metin Dis yiu test the AI Link to Krita? It works offline and gives also nice results. The option to control the AI reslts is incredible and brings "artist skills back to the desk".
If someone interested in the AI Module for Krita: https://github.com/Acly/krita-ai-diffusion
i used an analogy yesterday, that #LLMs are basically system 1 (from Thinking Fast and Slow), and system 2 doesn’t exist but we can kinda fake it by forcing the LLM to have an internal dialog.
my understanding is that system 1 was more tuned to pattern matching and “gut reactions”, while system 2 is more analytical
i think it probably works pretty well, but curious what others think
@kellogh I use that exact analogy. And emphasize that we certainly do use and need System 2 at least occasionally. At some point, human-like reasoning must use symbols with definite, not probabilistic, outcomes. Can that arise implicitly within attention heads? Similar to embeddings being kinda-sorta knowledge representation? I mean, maybe? But it still seems hugely wasteful to me. I still tend towards neuro-symbolic being the way.
Just came up with a new analogy I'm rather proud of: LLMs are digital compost heaps. They decompose whatever you hurl in and turn it into artificial excrement.
Also I'm moving from StackExchange to Codidact. If I'm going to do any more unpaid labour it's going to be for a not-for-profit, rather than a for-profit company. Feeding that work into a digital compost heap is the push I needed.