More on AI

Seen this on Rick's channel. Who would ever thought of Rick on CR4?

He has great insight.

Here is a pretty good background video from Welsh labs:

https://youtu.be/UZDiGooFs54

An interesting possibility is the return of analog computation. Particularly when resolving complicated neural network calculations of AI.

I like the idea of utilizing the speed and power efficiency of analog's continuous computation when the imprecision of analog computation is not critical.

Follow up on Rick.

I think that's a fair point. AI is great for finding information quickly, but real-world experience and conversations often provide context that no model can fully capture. I've noticed the same thing at industry events—some of the most useful insights come from informal discussions with other professionals.

It's similar to technologies like AI vending machines: the AI can optimize inventory and improve convenience, but understanding what customers actually want often comes from direct human feedback and observation.

Did anyone read the articles on Ford re-hiring 'greybeards'?

"The American car manufacturer has pivoted back to experienced human employees after an experiment in using artificial intelligence backfired. It has now rehired 350 veteran technicians — nicknamed “greybeards” — to train younger workers and the AI tools. The carmaker originally deployed 900 AI-powered cameras across its plants to detect quality issues, alongside automated tools for vehicle design. The company discovered that the AI was prone to pitfalls because it was only as effective as its training data and lacked the institutional knowledge, troubleshooting intuition and deep experience of human engineers." The Times (UK)

AI is potentially an extremely useful tool but needs very careful implementation. If those at the top levels of management are not aware then bad things will happen. I expect that engineers urging cautious initial implementation will be ignored (but blamed for any failures).

I've been coding with Claude. Learning how Claude works. I choose a very clear project with lots of different interfacing and functions and fun to play with and no deadlines.

It became very clear very fast that Claude was not following my intentions at first. He kept calculating odds and telling me the code would fail to yield any real results. I found If I resorted to metaphor and parables and such I could steer him into the right direction but it was like yanking the reigns of a stubborn horse!

AI is indeed a powerful tool, but at this point it is like the so called self driving car or worse. It thinks it knows everything already, kinda like my 5yr old grand daughter, but in fact it is severely limited to it's training, and it has zero insight or intuition without long philosophical conversation and an insistence on a particular goal.

You might find that Sqwish Labs of Cambridge UK are addressing the problem of going beyond the static AI to what happens when humans interact with the results