Month: May 2023

The Case for AI as an Existential Threat A short Youtube documentary that covers the basic doomer position. For a more detailed argument, see this article by Professor Yoshua Bengio, and the two debate links below. Related Links The AI Dilemma The AI Dilemma Open Source, Existential Risk Debate AI Pause Debate AI Pause Debate…

I’ve been meaning to write an article outlining why AI safety needs to be taken more seriously. Someone with much more credibility in this space, Professor Yoshua Bengio, has just written an article that says almost everything I wanted to say, with more eloquence than I could have achieved. He outlines the main areas of…

A poem by GPT4 constructed under the constraint that, for each verse, the first and last lines must be reversed forms of each other, with the provided theme of an AI composer never hearing its music. I’m sure Nick Cave would say it was bullshit. Still, not bad for a machine. And the artwork (both…

This is actually pretty neat. Describe a play list and ChatGPT will choose the songs, add it to your library, and make a playlist that you can click on from within ChatGPt itself. The first list was about an hour long, and the AI plugin happily extended it to >3 hours when asked. I did…

For a short time only, I am posting the first half of chapter from a work-in-progress, a book on consciousness. It’s heavy going, but intended to be accessible to well-informed lay people. Comments and feedback are welcome, but keep it nice. This is a topic where people tend to have strong views. http://www.asanai.net/the-hard-problem/

Professor Yann LeCun, a basic denier of the threat of AI, vocal tweeter and chief scientist at Meta , made the following comments in an interview with Lex Fridman as recently as last year, “I don’t think we can train a machine to be intelligent purely from text, because I think the amount of information…

The conversation below is part of an effort to improve the cognitive abilities of GPT4. The puzzle I put to it, which is also hard for most humans, seems to test its abilities quite severely. The puzzle was not in its training data because I invented it. The general form of the puzzle is that…

Yann LeCun is a prominent supporter of the idea that AI safety is guaranteed. All we have to do is make superintelligent AIs subservient to us, and no significant harm can come from this technology. I asked GPT4 to inspect LeCun’s Twitter feed and find potentially strained or inappropriate analogies. All text below generated by…

I am looking for collaborators who might be interested in improving the cognitive abilities of GPT4 with no new training, just careful selection of meta-prompts and refinement of the cognitive architecture. There are several similar projects out there at the moment, because it is fairly obvious that the current architecture is holding GPT4 back. This…

Hi there. The code below might be of some interest to other programmers who are working with the GPT API from OpenAI. The API has been somewhat slow lately, and it is often cumbersome to debug code while waiting for GPT to respond. I have faced a lot of timeouts, particularly with GPT4, which is…

Follow progress at AI Safety Tour: https://aisafetytour.com/ https://www.judiciary.senate.gov/committee-activity/hearings/oversight-of-ai-rules-for-artificial-intelligence https://www.washingtonpost.com/technology/2023/05/16/ai-congressional-hearing-chatgpt-sam-altman/

PROMPT: Could you please generate a GUI using PyQt that has an upper multiline editable text window and a button labelled A, followed by a lower panel that is split into two columns, each with non-editable but scrollable multiline text fields, accompanied by buttons B and C, respectively. On pressing button A, the text in…

The world is currently trying to pigeon-hole the latest crop of Large Language Models (LLMs), like GPT4. At the most essential level, what sort of entity is GPT4? Is it an AI? Is it one step short of Artificial General Intelligence (AGI)? Or is it several steps from AGI? Is it just a statistical text…

A Discord friend, Will Petillo, has been musing on AI sentience, and he proposes that we might have the first traces of it already. I have my own ideas on this issue, and personally I don’t think we have AI sentience yet. I actually don’t think “sentience” is well defined, so the question of whether…