In reply to a post by zyborg47:

You mean Grok have looked around the net and pinched other peoples designs

Well it's pretty clever to find fan curve designs to suit my situation and stop the bearing rattle.

Get me off ebay. Just bought a Gigabyte GTX 1650 Super Twin as it's supposed to be quieter than a single fan, and the Gigabyte has 3D fan control which stops the fans when under low load.

It just gets info from the net and puts it together. It is clever how it is done as long as the answer is right.

I have not used Ebay for years, my brother do a fair bit.

I just got rid of a load of stuff, I don;t want any more.

I have a load of old memory, DDR and SDR, not sure what to do with them. I would say they are no good for anyone, but some people using old machines may find a use. I also have some old video cards, I mean old, old, AGP and PCI I think. I know there are people who mucks around with retro stuff, so may be useful to someone. Not worth selling.

Been playing with CPU fan curve and Grok today. Nice and quite now and cool on my Gigabyte motherboard.

Its a bit more complex than a store and retrieve; which is why it takes so long to train/lean the models. An LLM transformer model just statistically predicts the next word in the sentence. That is all it is doing, and it turns out when you give it more than the entire works of shakespeare that a computer can convince people it is conscious.

In this case it likely saw thousands of articles in different languages about similar topics of fan wobble, and produced a response that was combination and edited of that learning.

It can often get it wrong; statistics go wrong. In language it ends up with ideas that we laugh at on the news such as glue on pizza.

In reply to a post by jchamier:

Its a bit more complex than a store and retrieve; which is why it takes so long to train/lean the models. An LLM transformer model just statistically predicts the next word in the sentence. That is all it is doing, and it turns out when you give it more than the entire works of shakespeare that a computer can convince people it is conscious.

In this case it likely saw thousands of articles in different languages about similar topics of fan wobble, and produced a response that was combination and edited of that learning.

It can often get it wrong; statistics go wrong. In language it ends up with ideas that we laugh at on the news such as glue on pizza.

It is probably more sophisticated now but I did read that in the early days the AI developers wrote it into the code that the AI engine could not reply with a "sorry, don't know" which was one of the causes of the invented data/facts/reports that has appeared in some AI outputs.

Well Grok is clever enough wanting to fine tune my fan curve based on noise and heat that cant be just searches.

In reply to a post by zyborg47:

You mean Grok have looked around the net and pinched other peoples designs

Interesting piece on the BBC News web-site where they report on an American study that posed questions regarding health issues deliberately phrased to test the capabilities of the AI engines. The most accurate (Gemini) gave advice that was regarded as highly problematic to about 15% of the queries. Grok was the worst performing, producing highly problematic answers to 30% of the queries and nearly 60% with highly or somewhat problematic information. Lots of good in AI but it is still a developing technology so should never be used as an information source without further checks.

In reply to a post by GonePostal:

Lots of good in AI but it is still a developing technology so should never be used as an information source without further checks.

It appears to be another "flawed gem"; perhaps insanely good in computer software development, but likely not a replacement for the original internet search (Google, Yahoo, DDG).

ChatGPT does have at the bottom of every screen... "ChatGPT can make mistakes. Check important info"; now I would argue it is not in bold enough print!!