I dug it up because the first author is my current boss, and it came up during a discussion about old discovery systems from Doug Lenat, and the various critiques people have had of them. (Although our research group doesn’t currently work in mathematical discovery, so it’s more of intellectual/historical interest to me.)

That takes an existing equation-conjecturing tool (Haskell’s QuickSpec library) and makes it more automated and machine-friendly. The next steps are to quantify what count as “good” and “bad” conjectures (hence the relevance of “On the Notion of Interestingness…”), to automatically benchmark and measure such quantities, to compare different systems, and to propose and implement new ones. Some of that is already done.

Heh, coincidental to see this appear, since I’m currently using it as a source for some work on property discovery in functional programming :)

Neat! Is anything on that project available yet?

I dug it up because the first author is my current boss, and it came up during a discussion about old discovery systems from Doug Lenat, and the various critiques people have had of them. (Although our research group doesn’t currently work in mathematical discovery, so it’s more of intellectual/historical interest to me.)

Most of the stuff is living in various git repos, but I’ve just cleaned up and publicised a small part at https://lobste.rs/s/fr1nv8/haskell_theory_exploration

That takes an existing equation-conjecturing tool (Haskell’s QuickSpec library) and makes it more automated and machine-friendly. The next steps are to quantify what count as “good” and “bad” conjectures (hence the relevance of “On the Notion of Interestingness…”), to automatically benchmark and measure such quantities, to compare different systems, and to propose and implement new ones. Some of that is already done.