Gregory's Server
@Gregmaletic @gerrymcgovern @gedeonm I *strongly* doubt that significant discoveries will be made by "AI" personally, given the contemporary technology of large language models has notable issues with bias, repetitiveness, and "smoothing out" statistically unlikely expressions of ideas into generica.
Nothing substantial in the contemporary "state of the art text thinker" indicates they're reasoning engines that can be applied to any knowledge base - which is what you would *actually* need.
@Gregmaletic @gerrymcgovern @gedeonm second, I would ask: if we propose to be using that much compute for scientific discoveries, with the afformentioned impact consequences: who decides on what is being researched? Who ensures accessability to results and elevation of *human* research? We already have issues with researchers from the global south lacking in plain *presence* and *accessability* in the wider research community, so what epistemological practice issue are you solving with "AI"?