A post you made in 2022 about finding symbols saved me and members of my household from a fairly regular rant this morning about how tired I am of nursemaiding the technology. I recognise your pragmatic and ethical response to the AI issue .I asked Perplexity to help me design a wedding anniversary card for my husband. It’s reply was useful and compliant with IP rules. It showed me photographs, where I wanted them and it hadn’t apparently burrowed around in my photo albums because it delivered pictures which were clearly marked as copyright. Also it listed its sources. And it it did what I needed which was to shortcut my figuring out the Powerpoint options. Saved much time.
Terminology is a major part of the problem. “AI” has become a nearly meaningless marketing buzzword, encompassing everything from route planning tools like this, to software to analyze medical imagery, to LLMs that their creators claim (falsely) will replace most human labor in a matter of years.
3 Likes
Hey?!
Well, I wouldn’t go that far and say that it’s meaningless. If used properly, it can be a very useful tool. One has to remember that it’s just an assistant, a tool…just as a typewriter is a tool, or a word-processor is a tool.
1 Like
I think Kewms meant “AI” as in the name, the appellation. And I agree.
A.I. this, A.I. that. A.I. everyFwhere – lol.
A.I supplemented breakfast cereals.
. . . . . . .
To provide leads as regard to research, I’ll agree.
But not to spit out the research, no.
It is still, after all, just a search engine on steroïds. It has no understanding whatsoever of whatever it replies. (With the added fake enthusiasm.)
2 Likes
No, genAI is not a search engine, and using it as if it were is risky.
It strings together “words that belong together.” Sometimes those words are associated with facts – because they happened to be associated with facts in the training set – sometimes they are not. It routinely references books and papers that do not exist, and quotes individuals saying things they never wrote and don’t believe. Every single “fact” that it purports to provide must be checked against a verifiable source.
4 Likes
Well, it goes along the lines of what I was saying:
Provide leads for research, but not the research itself.
This said, you, @Kewms, most certainly know way more about A.I. than I do. (It is not a topic I feel compelled to learn about that much.) What do we do, then, if you are right?
One can’t even use Google anymore without having it summarize the result. (?)
If it is all weaved with made-up stuff, how can it be good to anything else than the first superficial stage of research, or asking what bus to take…?
Surely the search results that are listed below the summarization, looking as they always did - before the A.I. addition -, are biased by it too.
It is not that useful for novels either, btw. Each time I ask it how to murder someone without getting caught, it refuses to reply. 
Just because Google provides an AI summary, that doesn’t mean you have to use it. The search results listed below the summary still point to actual pages, which (as always) are as reliable (or not) as the page creator wants them to be.
I’m doing most of my searches with DuckDuckGo these days. It offers AI summaries as well, but they’re pretty easy to turn off.
For serious research, Google Scholar seems AI-free, and of course all the subject-specific databases out there still exist.
6 Likes
Yes. But wouldn’t it be fair to think that they are now listed in an order that is influenced by whatever summarization (true or false) it came up with?
How long before it buries what doesn’t fit its model deep at the bottom of page 12…? (If not already.)
I guess what I mean is that as far as researching on the net goes, “relevance” is taking a new meaning. (Or will. Or not. The hell if I know. But I don’t like it.)
For a deeper research, use Google’s NotebookLM (https://notebooklm.google.com/) I have created many notebooks: investment, diet restrictions, health, etc.
I suspect that the list of results comes first, then the summarization is pasted on top. But the fact that it’s even a question is an argument for using a different engine.
3 Likes
Perhaps they’ll start fighting one another and we’ll be left without internet? 
But you verify the content, right? Otherwise… notebooks crammed full of “A.I. tips” regarding your wealth, nutrition and life sounds like a digital kamikaze mission.
1 Like
For years I had an app on my iphone that does that gave me the same results for the London Tube. The developer had done the groundwork visiting every station and noting the interchanges — even to the point of working which carriage to get in from the deparature station to match the exit at .the destination or interchange. THese days I don’t need the app as I no longer an occasion to travel to random Underground station.
Your example makes me wonder if the AI you used plagarised this app or something similar.
Of course. The best thing about having sources are you can have a hundred sources, but you turn ninety-five of them off 
Remind my of 2001: A Space Odyssey when he tries to turn-off HAL and he says, “I’m afraid I can’t let you do that.”
1 Like
https://youtu.be/Wvcu_iieFxw?si=D1XajynaOyQvlsN- long but interesting
I recently built a tool that helps integrate Claude AI with Scrivener.
I call it StoryScrib. storyscrib.ai is the site.
It does use the Claude API which does cost for tokens use, but I’ve optimized the tool to be very inexpensive.
It’s not ment to write novels out in full but it’s been pretty helpful as a tool. I built it just for me to use since there wasn’t a specific tool but I’ve seen some people asking about an AI tool for Scrivener so I decided to package it up and create a basic website for it.
I am a verified Apple Developer as well. The app is signed and it’s notarized by Apple.
The app doesn’t track any data, there isn’t any login or account creation for the app.
Just the Claude API key.
If you want to try it out and run into any bugs or would like to see about adding/editing any features, feel free to reach out.
Thanks!
Kevin