mindcrime an hour ago
That's not completely correct. I wouldn't say that artificial abductive reasoning is "solved" but there certainly are approaches to it and some AI systems are absolutely capable of a measure of abductive reasoning. One example (albeit certainly not the only one) being the "Parsimonious Covering Theory"[1] of Yun Peng and James Reggia.
I'd also object to the proposition that "abductive reasoning is where new ideas come from" in the sense that that hints that it's the only way to generate new ideas. Analogical reasoning[2] is also generally taken as representing a means of creating new ideas, with the caveat that it's creating new ideas as a synthesis of existing ideas and perhaps not "cut from whole cloth" so to speak.
I actually did a talk titled "AI, Sherlock Holmes Style - Introduction to Automated Abductive Inference" on this topic at All Things Open back in 2019. The slides and stuff[3] from that talk might be useful to anybody who's interested in this topic. There might be a video of the talk "out there" somewhere, not sure.
[1]: https://www.amazon.com/clp/0387973435