I just wrote this for my students, but thought I may as well let you see it too. See further for the final argument below last summer's Wired article about lambda (that I wrote.)
What went "wrong" with Bing
Here's a twitter thread about it https://twitter.com/MParakhin/status/1629162394764156929 (scroll up to see the thread.)
Which I got from one of my former undergraduates at Bath, Simon Willison, who is a great follow on mastodon and twitter both, and has a blog, I don't know how he is finding time for all this https://twitter.com/simonw
Anyway, in terms of the course material, what you are seeing is how people are trying to hack extra state / knowledge so that the chat is more sensible than a "random walk" through the language model. Unfortunately, a lot of pretty smart, and VERY well-paid people seem to be trying to make people, which they won't this way. They want intelligence to be a trick that just happens to run in our heads too. But from a policy and morality perspective, you can't ignore our physical bodies: the roots of our needs and wants, and therefore the security of our societies.
 |
| Bing: bulbs and books |
I wrote an article about this for Wired during the LaMDA news cycle, attempting to demonstrate from the moral perspective that humans are the centre of human society and security. I didn't there extend this to labour, but I think it's the same. There's no finite number of jobs that AI could "take" (be applied to) all of. Employment is a relationship between people, so the real questions you have are a) how much do people want to have employees [I'd say quite a lot, we enjoy power.] and b) how well do we valuate them (what wages do they get.) This latter is I think what gets disrupted by new technology – in the initial period of confusion we mess with the indices that lead to high wages, essentially commodifying labour. But people are not really exchangeable, so in the longer term (perhaps in response to being outcompeted, cf China's "cultural revolution" and how it ended) we will find knew ways to differentiate wages and encourage both discovery of skill and application of effort.