Minotaur claims LLM-style reasoning without a language model
Original title: Achieving AI-style reasoning without an LLM
The author says one network operation on a single CPU core answers in about a quarter of a second.
A follow-up post describes Minotaur, an ELC-based system the author says reasons by measuring a described system rather than predicting text, and without a GPU. Claimed tasks include deduction, induction, abduction, diagnosis, experiment choice, belief revision, and counterfactual histories. On author-described blind tests, including a police interview where questions change a witness’s memory, it is said to match a current language model reasoning in words. The post contrasts this with search, expert systems, logic programming, and Cyc.
Unconfirmed: Performance and blind-test results are the author’s claims and are not independently verified in the text.
- The author says answers take about 0.25 seconds per question on one core.
- Blind tests were written by people who never saw the code, the post says.
- The compared language model is not named.
- Historical systems cited include Logic Theorist, MYCIN, XCON, Prolog, and Cyc.
From the original article
16 min read at fleetingswallow.comIn the last post we introduced Minotaur, an AI that reasons with no language model and no GPU. This post looks at why language models think the way people do, at what was tried before them and where each attempt stopped, and at how Minotaur’s brain does the same kinds of reasoning differently: not by imitating how people write their thinking down, but by working on the structure of the system it’s reasoning about.
By edwardchalk · fleetingswallow.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.