Baldur Bjarnason argues that AI companies are offering pseudosciece, not real science, as evidence for the effectiveness of their tools. First, cue the list of AI failures *yawn). Then, the real argument: "They make grand claims, that this is the first step towards a new kind of conscious life, but don't back it up with the data and access needed to verify those claims... They make claims about something working—a new feat accomplished—and then nobody else can get that thing to work as well. It's a pattern." Right. If we focus on the corporate research - which is where all the attention is being paid today - then AI is more a world of illusion and trade secrets than it is genuine science. Much of the actual work (beyond simply scaling up to a billion processors) is being done by real scientists, with open data, testable algorithms, and reproducible results. I know; they work down the hall from me.
Today: 4 Total: 103 [Direct link] [Share]
