Using AI?

[Nick]

It's a scam and destroying the environment.  The content is scraped without attribution.

You also have to already know the answer to spot when it's returning junk.

Wow! Very sweeping generalisation most of which I profoundly disagree with! First, yes, AI datacentres use a lot of electricity but that situation will improve as algorithms and general technology improves. Some systems can be prone to "hallucination", however it's generally pretty obvious when that happens.

However, to state "It's a scam" is, by any measure, false. There are undoubtedly non-AI systems being passed off as AI for marketing reasons, e.g. systems that are solely based on machine learning, but true generative AI is a very real and valuable beast.

All "intelligent" systems need training - that includes humans and all supposedly intelligent life forms - none of the known systems spontaneously become knowledgeable or sentient. Content isn't simply scraped and regurgitated - just as in humans it's used to formulate the answers to queries. I've asked generative AI systems very complex and finely focussed questions regarding code generation for obscure functionality on unusual microprocessors and have frankly been amazed at not just the quality of the response, but also at the text rationale that is produced along side the code. The decision making seems very clear and makes for instructive reading - really subtle stuff.

Also, AI is used in many many fields, very few of which involve human interaction. I've been privileged to be involved in the funding of some medical applications which have produced astonishing results in the fields of colonoscopy image analysis, specifically the detection of bowel cancer (notoriously tricky to get right) and in the processing of mammograms to enable early detection of breast cancer. In both these cases huge amounts of training data - mammogram images and endoscopy videos together with the known outcomes - have been used, after which these systems are significantly better than human consultants at early stage detection together with far fewer false positives- fewer mistakes overall. They never get tired, never have a heavy night on the town, never get sick.

It goes without saying (but I'll say it anyway), ask a dumb or badly phrased question, and you'll get a dumb answer.

It's very early days in the AI world - there's a long way to go yet.

EDIT: Here's a live ranked table of the various AI chatbots around: https://huggingface.co/spaces/lmarena-ai...eaderboard - I don't use ChatGPT - recently I've mostly been using https://aistudio.google.com/app/prompts/...view-05-06 and am deeply impressed with the quality of responses, mostly concerning complex graphing of large datasets in "R" (that's a language BTW)..