I was interested to see that blowhard Richard Carrier has commented very dogmatically & abrasively on the subject of AI. See here:
AI Is Garbage and a Bubble (Please Learn This) • Richard Carrier Blogs
Hahaha Richard, I like it! I've only skimmed over his article but my guess is that he's probably got some worthy points there. We have to factor in, however, that he has a tendency to blow hard on stuff he doesn't like. That is very much a Richard Carrier trait.
It is likely that the biggest problem with AI is that it's over hyped. At least part of reason for this, I guess, is a result of chatbot behavior which gives every impression of a talking, walking sentience; humans have a reactive tendency to think that if it talks and walks like a duck then its a duck. Responsive talking, in particular, is very convincing of the presence of sentience. But that's a bit like a person from a primitive culture, unaware of our hi-tech times, looking in our very perfect mirrors or listening to a perfect recording and then inferring as a first-off conclusion that he's actually seeing or hearing a real human there and then; it's a very natural and understandable knee-jerk reaction to conclude that such are evidence of the immediate presence of a sentient being.
But chatbots are only another (albeit very sophisticated) human-computer interface. It is chatbot's very human like language behavior which is fooling a lot of people and if the investment bubble bursts we could be in for some trouble. But then are things as bad as Richard says? He's well known for being a blowhard and he may well be blowing just a little too hard (again).
The quasi human interface which chatbots provide is impressive and can give the impression we are talking to an entity which is super-intelligent. But then using SQL (Structured Query Language) to interrogate a big computer database can also be very impressive. I regard AI language models as a step (perhaps quite a few steps in fact) beyond SQL. I personally would want to congratulate the AI research community on providing us with a natural language interface as a way of accessing knowledge and information. Thanks and Well done; you deserve an accolade of two. I have a measure of appreciation about how AI works after my involvement on The Thinknet Project. That involvement started in the 1980s (but later morphed into a project in Quantum Mechanics).
But in creating AI based on the human thinking* model it is a fairly sound inference that it is therefore going to be very, very fallible; after all, humans are very epistemically fallible; in fact ask Richard himself just what he thinks of the fallible conclusions of many of his fellow humans on whom he has been known to blow very hard. His criticism of AI is on a par with some of his criticism of his fellow humans; like humans, like AI. However, although Richard's post is probably overstated it is worth a read especially by some of those rich CEOs who don't understand what they are dealing with. It might sober them up a bit on the subject.
No comments:
Post a Comment