Originally published at: https://boingboing.net/2024/07/01/perplexity-ai-has-nearly-stopped-my-google-use.html
…
@frauenfelder is this AI really better than using a “regular” search? I can still get decent results with google or whatever duckduckgo uses. If it’s better that’s interesting.
But for instance, I tried searching for “cafes and bakeries” near where I’ll be staying on a future trip. I got a list. But not a better list than I got last night by searching in google maps.
I did like some suggested similar searches. But I feel too guilty about spending the resources on an AI query.
Am I getting worked up over something that is an ad?
I too have stopped using Google, completely. Perplexity offers a couple of features I love. The answers for even moderately complex queries returned are consistently more helpful and footnoted so I can check to make sure the AI isn’t hallucinating. Additionally, I can ask follow-on questions that keep the response in the context of the current query thread; I don’t have to reformulate a new/better query like I’d do on Google (or any other standard search engine like DuckDuckGo or Brave). If the engine doesn’t understand my query and returns something ‘swing and a miss’, I can tell it ‘not what I meant, can you reframe per this…’ and I get better answers. I can also ask it for things like 'give me a 15 minute yoga flow that focuses on and I get back a lovely flow with poses, what they do, and times for each pose.
I know these systems use way more energy than a standard query engine but, until the others get their shit together and stop giving me ads and listicles, I’m AI bound.
Or you could not burn the planet down.
Tried and a hard pass for me. Almost everything I search for, I want to see what’s available, not get some generic-voiced summary of the first hit. If I want that, wikipedia does a better job.
But nice to know the webpages I spent time making actual content for will never even be stumbled on by anyone again…but at least I might have helped train some of the LLMs that are replacing them. Definitely doesn’t make me think trying to contribute anything is a waste of time.
The RAG approach is good, but I also wonder about delegating our analysis to something that misinterprets 20% of the time. What happens we get get too comfortable with automatically generated explanations that avoid allllll nuance?
I stopped using Google (in favor of DuckDuckGo) a long time ago over privacy concerns, but from what I hear, it’s becoming almost unusable between the results being chaffed with AI-generated, SEO-optimized garbage, and Google itself adding layers of AI to try to answer queries without you leaving their site.
So, I won’t celebrate this AI solution to a problem caused by AI…
@Escobar_Driver thanks for the context.
Perplexity can Fuck Right Off with its rent-seeking and plagiarism. Seriously, I was NOT expecting to see a BB article promoting it.
“AI” is not a solution to search. Full stop.
We are blocking AI as much as possible on some of our servers via Cloudflare but I am lobbying for hard blocks on Apache.
There is no benefit in having your site crawled if your material will be ingested and not credited. Google are shooting themselves in the dick with their insistence on AI. Perplexity are just a poor plagiarism machine.
Unsurprisingly, at least one of the articles I read suggested Perplexity was not honouring robots.txt. Though their weasel answer was that they honoured it, but it wasn’t their fault the third-party services they were using didn’t.
Perplexity seems like everything scummy about “AI” with added “insert ourselves as a paid layer between the user and free (probably ad-supported) content”.
This “AI” is not a solution to anything, and the sooner people realize that, the better.
… at Beschizza’s suggestion I’ve been using https://www.google.com/search?q=%s&udm=14&tbs=li:1
as a default “alternative search engine” in Firefox
I guess “udm=14” selects “Web” results, and “tbs=li:1” selects “Verbatim,” versus whatever the other choices were
AI and Google search, Duck Duck Go, etc are completely different animals. Each has their own strengths and weaknesses. I use Perplexity AI mostly as an unsurpassed How To source. As an example, when I ask it how to build,repair or configure most any computer, operating system or peripheral device, I always get good answers from it, mostly. But if the answer doesn’t work, I can tell it so and it offers other ways to approach the problem immediately. Search engines and You Tube videos can’t do that. I say just use it as a tool. If it works for you, great. If not, your arm is not being twisted to use it.
Oh, yeah, nobody is out there replacing existing tools with AI.
https://www.unesco.org/en/artificial-intelligence/recommendation-ethics/cases
Except of course, every corporation that intersects with technology, cause they they’re afraid of missing out on the next, hot new thing!
The majority of the resources used by modern LLM’s are spent in training the model, well before you or any other user ever interacts with it. Running a query against an existing model might use a bit more juice than a search on 1998-era google, but it’s on par with running a “simple search” on any modern SE.
The real reason to feel guilty about using AI search tools is because they’re trying to turn a profit selling pirated information. I know piracy isn’t theft, and I’m not about to advocate for anyone turning off their adblocker, but there’s a clear line between personal piracy and piracy for profit.
Perplexity works like and is just as shady as that kiosk at the flea market that will download and burn any DVD you want for $10.