[falcor84]

>I’m not going to use an LLM to shop for car insurance or look for hotels.

As I understand it, this is a great use-case for AI agents (a-la Custom GPT) and I wouldn't be surprised if the tech for that matures over the next year.

[linkjuice4all]

I guess I could imagine how they might be helpful in initial research - but no way I'm letting some LLM book a hotel that might not exist or get me car insurance from StratesFrarm Unsurance.

[Jensson]

> but no way I'm letting some LLM book a hotel that might not exist or get me car insurance from StratesFrarm Unsurance.

Ah, scammers targeting your LLM assistant will certainly be a thing, this really sounds like the old "I bought the Eiffel tower" scams.

[falcor84]

I'm not clear how this is different from scammers targeting humans? If anything, I'd expect it to be easier to set up the LLM agent to have secure/cryptographical validation of the providers, to prevent against the sorts of traps that humans fall into.

In the simplest case, we could just have trusted authorities provide lists of verified booking partners, same as e.g. Google Flights does now.

[hooloovoo_zoo]

The whole agent idea is questionable anyway. Sites are are already optimized to require the absolute minimum effort for conversion and most of the interesting aggregator sites already exist and are similarly optimized. Who's going to want a clunky natural language interface that breaks half the time?

[falcor84]

I would. I don't know what you're referring to as "absolute minimum", but when I book a flight I'm typically taken through a flow of about 10 pages, each trying to convince me to buy another set of extras. I would love an agent whom I could tell to not buy any extras and just get the basic plane tickets for me in one click, or if it fails to do so, to try again with another partner.

[BosStartup]

That would be the first step of rollout. The next step would be to ask you in ways that make it hard for you to say no. An LLM agent isn't going to stop companies from bad behavior/dark patterns, it's going to make them more effective at deploying dark patterns to segments of the market.

I think only competition and/or regulation helps push down the allure that dark patterns present to companies.

[CamperBob2]

Yes, it's a great use case for using local AI agents to shop for the best deal on the consumer's behalf.

Or did you mean Google's AI agent, or hotels.com, or somebody else with a built-in conflict of interest? Well, in that case, no thanks.

[vineyardmike]

Will local AI agents be the "linux desktop" of the next generation? Because as I understand it, next year is gunna be the year that local AI agents become a thing for the normies.

[CamperBob2]

For the average desktop user, Linux never had anything to offer over Windows. The last nail in the coffin was obvious to me at the time: Linux didn't make any headway when Microsoft started shoving the first ridiculously-bad versions of Windows 10 down everybody's throat by abusing a security update channel. At that point it became clear that there was simply no demand for Linux in the desktop market except at a niche hobbyist level. Stockholm Syndrome was in full effect among Windows users, and it still is, and, well, I guess that's just the way it goes.

The enshittification of the Internet is even worse, though. I think personalized search and content curation through the use of local AI agents may be the only way to buck that trend.

I want something that will do for companies like Google, Amazon, and the news media what SpamBayes did for email: make it useful again. Local filtering by a tireless and incorruptible proxy whose interests are aligned with my own is the only way forward IMHO. The only question is how long it will take. My rather-useless guess is more than a year, less than ten years.

[bamboozled]

And Google will be developing Gemini further so rheu still be in the game