Myth: "Public libraries and book stores have to dispose of books all the time so at least if these companies are scanning the books for AI they will be preserved for history."
-
@futurebird I don't think this is true. Chatbot can search web first, put the stuff in its context window (i.e. read it) and generate answer based on that. There is no reason to do it the other way round.
It's closer to RAG than generation from memmory. Both Chatgpt and Claude, afaik, can do this.
LLM can still misinterpret what it read it or make up nonexistent sources, but I think it's designed better than as you describe it.
@janbogar @futurebird
Chatbot/ LLM /Generative AI is the worst Web search or document search ever invented.Just use a real search engine.
-
I'm really interested in the customization and curation of training sets.
I think there is a lot of potential to make LLMs more useful with bespoke training sets and a totally different kind of output that isn't about chatting but rather about helping to organize and make the document set more useful.
@futurebird @MarkDW
You don't need an LLM to do that better! -
It is likely that these companies *do* keep copies for future training. But they don't want to share this data or make it searchable by the public.
If someone destroys an old rare book to scan it the public should get a copy of the scan. This is about protecting our culture, history, and heritage.
It's about preserving research and science.
5/5
@futurebird Indeed. As I understand it, they specifically avoid keeping copies to avoid copyright issues.
-
J jwcph@helvede.net shared this topic
-
@janbogar @futurebird
Chatbot/ LLM /Generative AI is the worst Web search or document search ever invented.Just use a real search engine.
I was saying this two years ago however I don't think it's true anymore.
LLMs currently do a better job than web search engines filtering spam and returning the most relevant documents. This is in part because search engines have been made worse. But, we can't just tell people "use real search" when their experiences won't make that advice seem effective.
I agree that these systems should not be search engines.
But that is what they are becoming.
-
I was saying this two years ago however I don't think it's true anymore.
LLMs currently do a better job than web search engines filtering spam and returning the most relevant documents. This is in part because search engines have been made worse. But, we can't just tell people "use real search" when their experiences won't make that advice seem effective.
I agree that these systems should not be search engines.
But that is what they are becoming.
It's obvious that the makers of consumer LLMs want their systems to become the new primary interface for the web. When I look at what other people I know are doing in their office work, at colleges, they are using LLMs as search engines. Simply to avoid all of the junk and spam most search engines return. Many of them are bashful about it and don't like "AI" in general.
-
I was saying this two years ago however I don't think it's true anymore.
LLMs currently do a better job than web search engines filtering spam and returning the most relevant documents. This is in part because search engines have been made worse. But, we can't just tell people "use real search" when their experiences won't make that advice seem effective.
I agree that these systems should not be search engines.
But that is what they are becoming.
@futurebird @janbogar
"LLMs currently do a better job than web search engines filtering spam and returning the most relevant documents."I've a nice bridge over the Shannon you might like to buy!
-
It's obvious that the makers of consumer LLMs want their systems to become the new primary interface for the web. When I look at what other people I know are doing in their office work, at colleges, they are using LLMs as search engines. Simply to avoid all of the junk and spam most search engines return. Many of them are bashful about it and don't like "AI" in general.
This is another stage of enclosure and control of information. Obviously some things will not be included in these systems. But, ever since Google started messing with their search algorithm and including ads we were already on this road.
This technology is very much overhyped but I think it's critical to recognize how it is really being used and where it's really experiencing growth and becoming popular.
-
This is another stage of enclosure and control of information. Obviously some things will not be included in these systems. But, ever since Google started messing with their search algorithm and including ads we were already on this road.
This technology is very much overhyped but I think it's critical to recognize how it is really being used and where it's really experiencing growth and becoming popular.
The real success of "AI" has mostly been in:
* spamming
* bad therapy
* searchSearch has been broken and bad for years. Many people have sounded the alarm. Here is the "solution" and it is worse than the problem. It's a disaster.
-
@futurebird @janbogar
"LLMs currently do a better job than web search engines filtering spam and returning the most relevant documents."I've a nice bridge over the Shannon you might like to buy!
This is just true. It was not true a year or two ago but it is now.
I found out because my husband like to use the LLM to search for articles. He's a PR guy and often needs to find out which paper contained a particular article.
Google, yahoo, duckduck go are a maze of paywalls and fake sites. The LLM just gives the answer to the question with a link so you can verify it is correct.
I am horrified.
-
This is another stage of enclosure and control of information. Obviously some things will not be included in these systems. But, ever since Google started messing with their search algorithm and including ads we were already on this road.
This technology is very much overhyped but I think it's critical to recognize how it is really being used and where it's really experiencing growth and becoming popular.
@futurebird @janbogar
Follow the money.Google been going downhill for years with sponsored results.
I used DuckDuckGo for a while.
Now I'm on Qwant.
I remember Altivista before DEC sold it.
Skype before eBay and MS.
There are "AI" successes, but it's a misused label. About zero LLM success. Plausible junk. A toy.
-
This is just true. It was not true a year or two ago but it is now.
I found out because my husband like to use the LLM to search for articles. He's a PR guy and often needs to find out which paper contained a particular article.
Google, yahoo, duckduck go are a maze of paywalls and fake sites. The LLM just gives the answer to the question with a link so you can verify it is correct.
I am horrified.
Fake sites created by using LLM.
-
I was saying this two years ago however I don't think it's true anymore.
LLMs currently do a better job than web search engines filtering spam and returning the most relevant documents. This is in part because search engines have been made worse. But, we can't just tell people "use real search" when their experiences won't make that advice seem effective.
I agree that these systems should not be search engines.
But that is what they are becoming.
@futurebird @raymaccarthy @janbogar the agentic “research” modes are even more effective now, as accurate as having a human assistant do the work but it’s done in minutes instead of hours. Maybe more accurate because it doesn’t get bored and will chase even the tiniest threads with its own assistants taking adverse perspectives. That’s a capability that will be hard to replace with non-LLM alternatives.
-
@futurebird @janbogar
"LLMs currently do a better job than web search engines filtering spam and returning the most relevant documents."I've a nice bridge over the Shannon you might like to buy!
@raymaccarthy @futurebird
It's nice to learn a new english phrase now and then, but I am dissapointed it has nothing to do with Claude Shannon. Also, I found it rude.I personaly see search to chatbot switch just a changing of the guard, and mostly for the better.
It works, and it works well.
Also cnsider that it might finally kill advertisement funded internet. That would be great.
-
Fake sites created by using LLM.
This is all very artificial. There isn't anything about the technology or methods that really make an LLM better at search, in fact, it's less efficient and worse.
But my husband can remember an article that said "a proliferation of bike lanes" and can't find it in duckduck go, but the LLM says "here you go" and finds the content.
This is driving people into using chatbots rather than search. Only strange nerds care about how these things work.
-
@raymaccarthy @futurebird
It's nice to learn a new english phrase now and then, but I am dissapointed it has nothing to do with Claude Shannon. Also, I found it rude.I personaly see search to chatbot switch just a changing of the guard, and mostly for the better.
It works, and it works well.
Also cnsider that it might finally kill advertisement funded internet. That would be great.
OpenAI is going to embed advertising in their chatbots and it will be fully integrated and invisible.
It's a disaster.
-
OpenAI is going to embed advertising in their chatbots and it will be fully integrated and invisible.
It's a disaster.
I suspect they will, but the situation is different now.
With search engines, the user would still go to your website and you would get money for adverts. With LLMs, that's not true. By filtering webpages through LLMs, there is no motivation to host adverts and tracking stuff, because you don't get enough human trafick for companies to pay for it.
Also, unlike search engines of old, current chatbots are already subscription based. So there is slightly less motivation.
-
I suspect they will, but the situation is different now.
With search engines, the user would still go to your website and you would get money for adverts. With LLMs, that's not true. By filtering webpages through LLMs, there is no motivation to host adverts and tracking stuff, because you don't get enough human trafick for companies to pay for it.
Also, unlike search engines of old, current chatbots are already subscription based. So there is slightly less motivation.
Search engines are providing overviews that do the same thing. While LLMs have introduced more "sources" or reference links to outside sites.
I don't think there is much difference between the most popular search engines and LLMs at this point EXCEPT-- when people use an LLM it's an invitation to start "forming a relationship" with the system and I think that is the real goal here.
-
Myth: "Public libraries and book stores have to dispose of books all the time so at least if these companies are scanning the books for AI they will be preserved for history."
I keep hearing people say this and it's based on a huge misunderstanding of what training data is and how it works.
When books are scanned to train AI it's not like with "Internet Archive" they aren't preserved. It's not like you could ask the LLM to tell you what was on page 46 of the book and it could look that up.
1/
@futurebird @Tarnport There was a book and eventually a movie called “Fahrenheit 451” about a dystopian future where books are confiscated and firemen burn books. At the time it seemed like a preposterous premise…
-
It's obvious that the makers of consumer LLMs want their systems to become the new primary interface for the web. When I look at what other people I know are doing in their office work, at colleges, they are using LLMs as search engines. Simply to avoid all of the junk and spam most search engines return. Many of them are bashful about it and don't like "AI" in general.
@futurebird@sauropods.win @raymaccarthy@mastodon.ie @janbogar@mastodonczech.cz a lot of times the search results are generated anyways
-
@futurebird Indeed. As I understand it, they specifically avoid keeping copies to avoid copyright issues.
@cford @futurebird and my sad guess is that they probably dont keep the original scan data around and properly labelled because they just dont care, it probably gets OCRd and _maybe_ the uncorrected text is saved so it can by used for future training runs, but again they dont care about the metadata, or correctness, so it might not be useful as an archive of anything.
