Is AI Limiting The Knowledge Base On 914 World? Options

[Osnabruck914]

Might just be me, but it sure seems like since the ramp up of Artificial Intelligence, especially in the last six months or so, the number of new threads on 914World Garage Forum has dropped off. If true, is it because Google searches on questions about our cars are now routinely answered by the the Google AI component at the beginning of the search results presented, often negating the need to start a new thread on the World? I'm not saying the Google AI answers are foolproof by any means, but quite often they are spot on, and use the exact verbiage of what posters have previously written on the form.

This is concerning, because if true, it is essentially putting a cap on the online knowledge base the administrators and contributors to the site have created. If we all begin to rely solely on AI availability, then new perspectives, methods, repair tricks, etc. will never be generated and added to the knowledge base.

I guess Andy and Co. can confirm or debunk my suspicions, by looking at statistics on new posts. I certainly hope I am wrong about all of this.

Osnabruck914

[VaccaRabite]

I work for the federal govt, and the current administration is pushing AI HARD. So now I have accounts with three different AI tools, and I need to know how to proficiently use them.

Which is fine. For the past 6 months or so I've been using it more for almost anything I can think of.

Right now, what I've found is that for general information, AI is pretty good. For helping wrap your head around a problem, AI is pretty good. For conceptualizing ideas, AI is pretty good. It helps focus your research, and it make mundane tasks faster.

HOWEVER.

Once you get past generalities, you need to be careful. I've asked ChatGPT to develop EFI systems for air-cooled engines, including detailed breakdowns of wiring and detailed lists of all the parts needed for the work as well as where to get them. The list included radiators and coolant plumbing fittings.

Is that an automatic fail? No. Most of the list was pretty good, and as a thought experiment it was pretty impressive. But you have to be willing to use critical thinking when using AI, and not just accept the output as your answer.

Back in college, I had this incredible professor that let us know a secret. We were not in college to learn everything - that is impossible. We were in college to learn how to find anything. How to gather knowledge and use it. AI makes a good tool.

IMO, AI should be being used at the college level as a tool for research. But turning in papers written by AI is plagiarism, and should be treated as such. I know this an issue that many colleges are currently facing.

And here is another thing - that makes me glad to be middle age now. By about the time I'm looking at retirement in a decade or so, AI is going to be gutting most white collar jobs. There will be new types of jobs that open up - there always are - but its going to be rough for younger Millennials I bet as they get into their 50s.

Zach

[DennisV]

If I see Sarah Connor running down the street in my time on this planet, I'll know that AI has gone too far.

I've asked ChatGPT to develop EFI systems for air-cooled engines, including detailed breakdowns of wiring and detailed lists of all the parts needed for the work as well as where to get them. The list included radiators and coolant plumbing fittings.

It still frequently fails for me on even simple things. Here's a ChatGPT drawing for how to wire an inline fuse. Forget the misspellings. One lead no terminal. One lead wrong terminal. Definitely potential for bad outcomes if you accept its responses as fact.



I've had Claude double down on a wrong answer. Tell me to check for myself, and provide a resource that contradicts it's own wrong answer.