* Dr. Arne Babenhauserheide <arne_...@web.de> [2025-03-09 23:03]: > Maybe this is the misconception: before Alpha Go, no computer was able > to beat even moderate human players. > > Go was the bane of game-AI.
https://arstechnica.com/information-technology/2022/11/new-go-playing-trick-defeats-world-class-go-ai-but-loses-to-human-amateurs/ I see there about it, and it seems it still has weaknesses. Though you may see it was as a game invented long ago: https://www.myabandonware.com/game/go-kyh It wasn't AI who took the old game and made a new version, and it wasn't "artificial intelligence" who decided "let me train myself to beat human in this game". Here is scientific paper that concludes: https://arxiv.org/html/2409.04109v1 Statement came from a study with 100+ NLP researchers. "Despite this, no evaluations have shown that LLM systems can take the very first step of producing novel, expert-level ideas, let alone perform the entire research process." Here is the answer by ChatGPT: https://chatgpt.com/share/67ce7cb5-94f4-800a-8fe0-32fc60b30d00 and I don't consider answer authoritative, that is why I am looking for examples. LLM answers are like drafts, pointers for further review. I never consider them conclusive. It seems that people wonder about same subject as me: Large language models are powerful imitators, but not innovators https://the-decoder.com/large-language-models-are-powerful-imitators-but-not-innovators/?utm_source=chatgpt.com#summary Discovering novel functions in everyday tools is not about finding the statistically nearest neighbor from lexical co-occurrence patterns. Rather, it is about appreciating the more abstract functional analogies and causal relationships between objects that do not necessarily belong to the same category or are associated in text. In these examples, people must use broader causal knowledge, such as understanding that tracing an object will produce a pattern that matches the object’s shape, to produce a novel action that has not been observed or described before, in much the same way a scientific theory, for example, allows novel interventions on the world (Pearl, 2000). Compared with humans, large language models are not as successful at this type of innovation task. On the other hand, they excel at generating responses that simply demand some abstraction from existing knowledge. https://journals.sagepub.com/doi/10.1177/17456916231201401 If it would be that easy to say "text generating LLM can innovate" then with all the hundreds of LLM models it is expected that such huge fake intelligence can easily prove that it can innovate. Rather than that, scientists prove it can't. It's a good tool for brainstorming, but it's all the existing knowledge. -- Jean Louis --- via emacs-tangents mailing list (https://lists.gnu.org/mailman/listinfo/emacs-tangents)
