AI Research Papers - Academia.edu (original) (raw)

A brief exploration of artificial general intelligence and the "singularity," in which I address the problems artificial intelligence theories have had with explaining distributed cognition, consciousness, and various other things, as mediated through the memories of a class assignment writing a program that would respond to random input with natural language responses. I'd call this an inductive commentary on the lacunae of historical AI approaches more than it is a serious exploration of applications of current AI issues. Were I to rewrite it today, I would probably address the differences between a corpus and statistical probabilities of word patterns (n-grams) and the world of superhuman metonymy, as opposed to the kind of communicative creativity that CHAT and Cognitive Linguistics address. I might have wanted to spend more time on supervised learning, unsupervised learning, transfer learning, and so on, and most certainly GPT-3. I might also state more explicitly the difference between my personal effort of generative grammar, and mind in society, rather than working quite so hard to amuse. However, that would have required far more words than I would have been allowed for this little piece, and I'm not qualified to write it, anyway. For a deeper exploration of the research issues, see Kaptelinin and Nardi, Activity Theory in HCI (2012), and George Lakoff, Women, Fire, and Dangerous Things (1987), and Raymond W. Gibbs, Jr., Embodiment and Social Cognition (2005). For a more informed exploration of what my computer science professor was trying to have us think about, see Norvig, Paradigms of Artificial Intelligence Programming (1992), or perhaps Russell and Norvig, Artificial Intelligence: A Modern Approach. And for a recent discussion of some of the interesting fallout of GPT-3, see: https://dailynous.com/2020/07/30/philosophers-gpt-3/