Agreed that it is usually fine, and most of what I search for isn’t important and whatever it gives me is fine.
But anything important, that is not good enough. I was doing a search on speakers and learning about what to look for. I used ChatGPT (I think it was) and gave me a nice summary. Then I did a Google search. The top link was pretty obvious where ChatGPT got its info. But the actual webpage had charts, diagrams, etc. The second link had a bunch of other stuff that the first didn’t cover and gave some completely different info. Then there was a Reddit discussion which had a bunch of people talking about and discussing speakers. Plus, there was a rash of YouTube videos it suggested, as well as a bunch of links to buy speakers (that part wasn’t appreciated).
For Excel stuff, the AI is usually wrong or lacks enough info to solve my problem. It’s cool when it’s right though, but I am better off just going to one of the known Excel sites.
We are all different, but it seems crazy to me to limit yourself to one AI. Google has gone to hell, but I still get more info out of than I would an AI.
I was a self taught I.T. manager for 30 years. I started with a bunch of weekly tech “tabloids” and O’reilly books but once Google came along I relied on it. I still do. The AI overview is just my preview and the links take me to the source.
I have used it, I used it for a month or so when Arc started allowing you to use it as your search engine. I appreciated the links it gives and liked it. Still 3 problems:
I get more comprehensive answers with mixed multimedia from Google
I am concerned about the power demands of the growing AI/Crypto data centers (electricity must be too cheap judging by the amount of people who have Christmas blow up dolls in their year this time of year though)
I believe that sites that provide good info deserve the traffic
I know I am unique on this forum on this issue (well every issue really).
#2 Google is part of the problem, they are an AI company building massive data centers as well. But a ChatGPT search consumes 5x the electricity of a Google search. (Since Google does an AI search on every search, that might not be the case anymore.)
#3 I don’t read Google’s AI nonsense at the top, I click links.
I just used Perplexity for the first time in a while with my standard question “What should I look for in speakers?” It gave me the shortest, least helpful answer of any, and then listed 5 crappy speakers as “related products.”
Google’s top link to the question has by far the most detailed answer.
I tried Perplexity after MacSparky waxed lyrical about it.
I liked how it presented what it found.
It made some dreadful blunders on even quite common topics (e.g. I asked it to summarise the views of a couple of political parties and it included policies and statements from their opponents alongside theirs which made the summary useless and misleading)
I found some of the citations were using dubious sources and I couldn’t rely on it to include or reference the canonical sources.
I think it is a “front end” to other LLMs - so what it comes up with will only ever be as good as the LLM it queries. I found it on a par with the other LLMs for content, but it gave things a superficial authority that I thought was dangerous: as is being said these days “frequently wrong, never in doubt” is dangerous. Perplexity is often persuasive, which is dangerous when you should reality check and question everything it produces.
I don’t remember how I discovered the MPU podcast, but there was something unique about a tech conversation between two lawyers. They were Mac users, not network managers or sysadmins, etc. Week after week I kept listening and learning about software and techniques I could pass on to my users . . . and use myself.
We joined this forum the same week, and yes things are changing. But the world is changing and what a Mac power user will be is changing. The people here are what keeps me coming back to “cuss and discuss” Apple and tech. I doubt you will find another group quite like it.
I’m a Perplexity Pro fan. Something that I don’t do nearly enough of is to continue the pursuit of a question until I get a satisfactory answer. Put in another prompt and another, and so on, if necessary until the AI gets it right. And less than the best citations can be a problem too, but at least Perplexity is plain about where its information is coming from (transparency). And the citations can be challenged with further prompts, also.
I tried ChatGPT before I tried Perplexity, and ChatGPT’s conscious(?) political bias was obvious. And Gemini became famous for answering with images that completely excluded white people (since fixed). So none of the AI’s are perfect.
I’m not married to Perplexity, but for now it seems to be the best overall for my purposes, mostly search, and explaining how to do stuff. Someone with other interests, for instance a coder, would want a different AI.
