That is a typical case of AI misreading the question.
It happens with google searches too … you get lots of irrelevant rubbish.
2 Likes
Seems like any search engine is full of AI trash nowadays. If people want to search, they use search engines. If people want to ask AI, they ask AI. And not people search, AI pops up.
I don’t really understand the meaning of agentic AI. That basically means you’re replacing yourself with a robot.
1 Like
If people want to ask AI about their own private stuff, I don’t care.
I start caring when AI drivel gets shared publically.
2 Likes
The worry is when you start to see news items like these
https://www.theguardian.com/technology/2025/nov/30/ai-poetry-safety-features-jailbreak?
@nevj :
Hi Neville, 
That may be about to change.
For example, duckduckgo´s Search Assist does give you some info about the source(s) it used for creating its answer:
In this example I clicked on the Red Hat link and was immediately transferrred to the original source:
That seems a worthwhile effort by Search Assist, I guess.
Many greetings from Rosika 
5 Likes
My experience so far with AI:
AI tends to synthesize and conflate. A lot of subjects it automatically chooses causation instead of correlation. I always asks for references with links.- I cannot tell you how many times i can never find the references (even with provided links). Sometimes, when I do find them, AI has taken the information completely out of context.
I have found AI is very good at coding, OCR (amazingly great), Changing document formats (example taking a pdf and getting it into excel) Taking my file of lecture notes and making worksheets and or tests. (saves time, but I still have to double check these to make sure they make complete sense.)
And yes- I have a paid subscription and I am careful what I share with AI
2 Likes
Hi Rosika,
I like that. I think I shall shift to ddg search assist.
We should support those who do search reports properly
" Search Assist responses always link directly to one or two sources, citing where the answer came from, so you can easily go and get more detailed information. It’s important to remember that responses are auto-generated from cited sources, based on crawling the web."
from DDG help pages
Regards
Neville
3 Likes
It is a tool that uses AI.
Your next round of surgery will probably be done by such a device.
I imagine this is very simple AI compared to what ChatGPT does, and it would not involve LLM or large databases.
The term AI is acquiring a very broad meaning. A lot of it used to be called robotics. The new name is more glamorous.
1 Like
I never thought of asking for references. From what you say, it works sometimes.
I wonder what would happen if I asked it to do a library search rather than a web search… a bit like Google Scholar I suppose.
1 Like
Hi Neville,
thanks for your feedback.
Yes, by all means: try out Search Assist at least for a while to see if it suits your needs.
At least they seem to be going in the right direction.
Thanks also for the link you provided. 
Many greetings from Rosika
3 Likes
A few weeks ago, I submitted part of my Linux log to Co-pilot to see what it would do with it. I thought co-pilot did a decent job of explaining the entries. I for one, do not understand the logs very well yet.
5 Likes
That is only a jargon problem. You will get used to reading logs… but why not use AI to help. That is a sensible constructive use of AI… not much different from using google translate.
It gets back to programmers being poorly trained at writing error messages or log messages.
4 Likes
I was thinking that would be a great utility to build. When you have some issue, you can’t figure out you fire up this AI troubleshooter that examines the logs and makes a recommendation. I’m not sure how hard or easy that would be to create. Maybe AI could help with the creation too.
3 Likes
I am pro AI, but it i have some issues with it. Ai is a learning tool that needs to be refine
3 Likes
That is a bit like googling an error message.
What would you use as its training data?
3 Likes
The LLM is already trained. You would just have it examine log files when there is a failure. Automate what you would do yourself to determine the problem.
The error message might be enough for some problems, and you wouldn’t need the AI or to dig into log files. When you’re not sure what is causing the issue, you could invoke the AI to examine logs for you. Do the grunt work.
2 Likes
I dont see how a language model could understand a logfile.
It requires some computer knowledge as well as English.
I like the idea. Browsing logfiles is not my favorite activity.
2 Likes
There are valid uses of AI. It could be helpful with research, brainstorming, grammar/style checking, and sifting from mountains of data. That’s what it’s good at.
What is not good use of AI in my opinion is having it do coding for you, writing for you, having it do your homework, etc.
5 Likes
There does not seem to be any way of promoting valid uses and preventing not so good uses.
The moment someone invents something, it can be used for good or evil. The thing itself is not evil, it is the wrong usage that is evil.
3 Likes
I do this manually all the time at work. If there is an issue with a server I can’t get a handle on, I can feed some of the logs to Copilot and it’ll come back with some suggestions. Obviously, you need to be careful and verify the suggestions.
The LLMs were trained on tons of publicly available websites like Stack Overflow and Reddit. Many of the posts there include log snippets showing an issue and asking for help from other users.
I was suggesting a tool to automate the process I now do manually.
4 Likes