listen.... ... the text / code / etc generation from GPT-3 *is* super impressive but you have you remember its basically just a real good search engine
i also don't know react but i've built plenty of react apps by searching my problems in human language on stack overflow. just like me, GPT-3 doesn't know react (or english) but it knows how to find a pattern learned from someone who does https://twitter.com/sharifshameem/status/1284421499915403264
again, this is in and of itself very impressive! and you could get into the epistemology of what "knowing" actually is, sure. but imo its important as AI gets more complex to continue to frame this process as searching and reorganizing already existent information
you could get into the "but what are humans but reorganizations of existing information" thing, and sure, yes, kind of idk. but regardless, big learning models function beyond our capacity to hold in mind- they gain something like impartiality out of scale bc its too much to know
there are a million big, very real, problems with this, see the well-documented 'ahh bias in the dataset!' stuff over the last decade. but for me, there is also a human tragedy here, one of personal making getting laundered into machine creativity
from @v21 the other day: https://twitter.com/v21/status/1283673776719364096
(idk how i feel about copyright, its complicated)
but seeing a person's words, work, image, and stories gathered without their consent, then hashed through a system that removes their context and agency and returns that human life in small pieces, and calls it invention ..
but seeing a person's words, work, image, and stories gathered without their consent, then hashed through a system that removes their context and agency and returns that human life in small pieces, and calls it invention ..
its such an underselling of the specificity of human knowledge and of human stories. its such a bland way to treat searching through knowledge, which in and of itself can be a joy. (picking up next book over at the library, clicking the linked article on wikipedia)
i use AI in my own work, and i'm moderately positive about its capacity as a tool. but what I'm truly excited about is not the capacity to launder knowledge into new mush, but to organize knowledge in ways that are transparent, traversable, and human readable
GPT-3 is a search engine, and I want to use it like a search engine - with all of the context of its scraped data intact
(just appending this clarification to the end of the thread for folks who thought i was literally confusing search engine technologies with neural network technologies. which are. yes, different systems. but that ultimately function in similar ways) https://twitter.com/everestpipkin/status/1284875924555608065