Google is reportedly drawing up plans to charge for AI-enhanced search features

submitted by PanzersGhost from (theguardian.com)

sorted by:
new (suggested)
topinsightfulfun
you are viewing a single comment's thread.

view the rest of the comments →

[–]Canbot 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (5 children)

Building a search engine is not as simple as pinging every ip and cataloging it. The software and algorithms used to make it work need a developer. That is exactly what AI does best. Local search would require a massive database.

An open source LLM can already ping random ip's, analyze the page in a very sophisticated way, and do just about anything based on the results. Anyone can set it up with very minimal understanding of computers or software, and run it on easily affordable hardware. That has never been remotely possible before.

The vast majority of people are indeed too lazy and stupid to ever bother doing it when they don't even understand that Google is mind fucking them by controlling everything they get to see. But the tools will be built by the exceptions, then get talked about on tech blogs and spread by influencers. It will be so easy to use even the morons will pick it up if only to feel like they are part of the smart crowd.

[–]binaryblob 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (4 children)

I am looking for an answer based on the economics of the operation. You are not making any sense. A LLM would still need an index for high performance, because otherwise you would have a search engine that is behind all the time.

If I were to stand on the street with a free computer with a Google interface builtin (but not depending on Google), people would still not want to have it, because "they can just use Google".

It was possible at least a decade ago already to run a distributed search engine node locally and even share resources for free. I know, because I did. I would say the key enabling technology for fast big indices is fast and large SSDs.

The thing is that even if it is relatively cheap, the utilization of the hardware would be so low that it would be an economic waste for the 99.5% of the time you aren't using the search engine (Google just serves another customer).

I think you are severely overestimating human intelligence; unless it's built into Windows it's not going to happen at a mass scale (and I am saying that as a Linux user). Surely, perhaps a million people will do it, but we have billions of people.

If there is a killer app, then perhaps it will happen.

[–]Canbot 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (3 children)

A LLM would still need an index for high performance

Forget high performance and think free from manipulation and data collection. As time goes on google's appetite for control grows. As google gets more and more tyrannical people grow more desperate for alternatives. There will be a tipping point when the quality of the alternative and the demand for an alternative cross and then everyone will abandon google for the alternative.

[–]binaryblob 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (2 children)

How would every single individual defend against misinformation? They can't. All Google needs to do is push a main stream narrative, because otherwise they won't have eyeballs anymore. You seem to assume this magical anti-misinformation technology is going to be available open-source .

I think full text regular expression search for the web or some subset of the web would be interesting to have, but even that requires tremendous resources. Searching without any index is just not realistic.

[–]Canbot 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (1 child)

All Google needs to do is push a main stream narrative, because otherwise they won't have eyeballs anymore.

Google does not have to push any narrative. It is morally corrupt of them to push a narrative, even if their intentions are good.

It seems you are incredulous about the potential of implementation because you are coming from a developer's perspective in which you are practiced in fitting ideas into existing architectures because you build software systems that have to work on current hardware. From my perspective I see that hardware changes very rapidly and I see that LLM's built on transformers do not behave like all other software, a fact that I constantly see developers not understand. Maybe I am wrong and you are right, but I am going to have to disagree. I don't think I can change your mind but I think you should try to keep an open mind in the future as you watch the space evolve, and keep an eye out for search agents and potential ways to improve search for everyone.

[–]binaryblob 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (0 children)

You should learn to articulate your ideas better and you should understand that I am not merely a "developer". You still have not provided an equation for what is required to happen before a locally running LLM is going to crush Google Search + a big tech implementation of LLMs. Let's say all uses of Google Search are done by LLM agents, then do you not think Google will simply start charging for API access? So, whoever implements an LLM still needs an API for rank based search, if alone to be able to verify sources.

I think you are literally incapable of providing a multi-agent differential equation describing the economics surrounding LLMs.

If you want to do another attempt, you should just assume that I am much smarter than you will have be or have been in your life and have more experience. Just describe an actual vision of the future (also specify how many years in the future you are describing) based on actual plausible physics.