Posted by scythe 3 days ago
The idea is to define what “loyalty” means for an AI agent in both technical and legal terms, and then build systems that can prove they’re acting on a user’s behalf (ie not a platform’s or advertiser’s).
It’s early-stage research, but the overlap with many of the questions here is striking. Would be great to get feedback from this crowd as the work evolves.
I’m part of the group working on Loyal Agents and happy to discuss it.
A lot of price gouging is based on you not knowing the details or the process. With LLMs you can know both.
For most anything from kitchen renovations to A/C installation to Car servicing - you can now get an exacat idea on details and process. And you can negotiate on both.
You can also know how much "work" contractors have at this time which gives you more leverage.
For anything above $1000 in spend, learn about it from your LLM first. My usual questions:
1. What are all the steps involved? Break the steps down by cost. 2. What is the demand for this service in my area around this time of the year? 3. using the above details, how can I negotiate a lower price or find a place which will have this at a discount ?
Information asymmetry is only valuable if you can execute on it. All of your examples are actually examples of both asymmetry and market control. HVAC, there's typically only a few legitimate licensed providers in town so they can set the price however they want. Car servicing, indie shops are always better but if you want to maintain your warranty you'll need to use a stealership which goes by a book (and it's mandatory).
I'm not convinced an LLM can help with these situations. I would suspect you're more likely to get a "screw you" price in return rather than winning a negotiation. When I shopped for a new HVAC after mine gave up the ghost after 20 years most providers were within a few hundred dollars of each other. An LLM would've been useful here for warnings ("you probably dont need ducting", "you probably don't need duct cleaning") but as for the bulk of the cost there's a monopoly and there ain't nothin you can do about it. When I got my yard worked on it was a similar story. Despite every landscaper providing offers from cheap to absurd, the ones that I could sue if they hit a gas line were all within the same price range.
These people are also very used to the "know-it-all homeowner". They're more likely to ignore you than help you because if you actually knew-it-all you'd do it yourself.
I think, rather, LLMs will be extremely useful in bill negotiation where the data is absolutely clear, you have a copy of it, and it can be analyzed in full (no asymmetry). For example, an LLM could be trained on medical billing codes and be able to analyze your bills for improperly coded procedures (very common).
Eg: when my shower didn't work I was able to figure out all steps - and also do most of them before getting stuck at one particular point because I couldn't physically pull the unit out of the socket.
I was able to negotiate down $150 for that one.
In another instance with gas pipes I was able to find laborers who were good but just didn't have a branded van yet.
In this case LLMs help me understand that the laborer was damn good at his job and how to cut the cost of the job by breaking into different pieces.
The whole process is very tactical - you will lose quite a few negotiations before figuring it out. Also its not useful to just abstract all the jobs as you've done in your post. You've somehow got to the EMH except for service providers - its just not true.
The way different types of compnaies force you to pay more is very different. Lockpickers are very different from plumbers for example. Also each service provider have their own way of doing things and breaking points.
Also every geography is different. Service providers seem to charge the most with elderly house owners and peopel with nice houses in nice areas. So you can definitely use LLMs in those situations to find areas to put ads to attract better prices.
the best part of all this is how you can apply these negotiation skills to your job search or any other situation. definitely a long game like finance or health.
From books and guides at the library and bookstore, to "This Old House" and "Click and Clack" we have been distributing the knowledge of how to do things for a long time.
The internet just made all of that knowledge much easier to access, with the time/cost/distance dependency being removed.
Have Americans become less capable over time? Or are we just more aware of the portion of the population who simply does not put in the leg work to DIY things?
Maybe a bit of both, with a lean into those who do not know having a larger voice. As an example I saw a video yesterday of someone being a "full on foodie" followed up by someone who was calling an onion "garlic".
Does an LLM really change what COULD have always been done, or just make it more accessible for those of us who do/want to have the tool?
Yes, but I don't know what point this is supposed to make, though. LLMs lowered certain costs in an extreme way.
You could always have become a plumber in order to negotiate with plumbers. The reason you didn't is because the investment to become a plumber was more than you were likely to get the price lowered (or to save by doing the work yourself), and you would have to anticipate your needs before they came up. The people who did become plumbers set up (or joined) a business and marketed themselves so they were negotiating with a lot of people over a lot of jobs, making the investment worth it.
People who invested the time to learn plumbing traded with other people who also concentrated their investments into a few things (but different things), and together, made civilization.
> Does an LLM really change what COULD have always been done, or just make it more accessible for those of us who do/want to have the tool?
I'm trying to figure out if you were arguing with somebody who said that it was IMPOSSIBLE to learn the things that people clearly know how to do. Changing arguments into existence proofs has always made them easy to refute; I'm not willing to say that it's impossible for pigs to fly, it's just not cost effective. AI has clearly made it cheaper to obtain the knowledge negotiate with plumbers about a specific plumbing problem that just came up in your life than watching hundreds of hours of This Old House, buying your own tools, and practicing.
The internet has given anyone/everyone a voice, for better or for worse, both widening and shortening the feedback loop. Now LLMs are shortening the loop even more, while unable to distinguish fact from fiction. Given how many humans will regurgitate whatever they read or heard as facts without applying any critical thought, the parallels are interesting.
I suspect that LLMs will affect society in several ways, assisting both the common consumers with whatever query they have at the moment, as well as DIY types looking for more in-depth information. Both are learning events, but even when requesting in-depth info, the LLM still feels like a shortcut. I think the gap between superficial and deep understanding of subjects is likely to get wider in the post-LLM world.
I do have hope for the garbage in, garbage out aspect though. The early/current LLMs were trained on plenty of garbage, but I think it's inevitable that will be improved.
Thats a meaningful difference to get upto speed.
If you're an engineer you get a hell of a lot of agency in 15 minutes.
A lot providers will mark up prices for things that actually don't matter. You ca definitely figure that out quickly and tell them its not needed.
And hten others will give you a cheaper alternative at a higher price - this is really fucked but I"ve seen it a lot.
Essentially service providers (most) are screwing you on price is terrible ways. LLMs can get you 20% off at the VERY least.
IF you aren't in a hurry you can cut a lot more down and even do most of the things on your own with a LLM.
But regardless, this arms race doesn't happen because the vast majority of people are bad at prompting models, and when you start writing prompts with spelling errors and other grammar issues, your model responds with low quality, wronger outputs just to punish you for your lack of attention to detail.
Out from curiosity I ran though an LLM on it, that pointed out it was full of traps, salary frozen for three years, massive financial penalties on leaving (getting fired with reason, getting fired without reason, leaving on the wrong date, etc), half a week unpaid overwork monthly added back (it was advertised as a 35 hours position and they asked the salary expectation accordingly - then in the contract they added back 5 hours weekly, unpaid), company can deduct money from your salary based on their claims, pre-contractual intellectual property claims, etc.
There were even discrepancies between the German and English text (the English introduced a new condition in a penalty clause on leaving), that could have been nearly impossible to spot without an LLM (or an expensive lawyer).
In hindsight many red flags were obvious, but LLMs are great to balance out the information asymmetry that some companies try to leverage against employers.
I would expect that this will cause certain programs to see more demand than the creators anticipated for (extrapolating previous trends), which might require changes in the programs (i.e. more people apply for benefits than expected, benefits / application might have to be cut, etc).
And in some ways there's a Cantillon effect (though traditionally associated with proximity to the "money printer", but here the proximity is to the LLM-enablement; in that those who use the LLMs first can get the benefit before the rules are changed).
Longer term, there is a real danger that asymmetry will increase. Using LLMs appears to make many people dumber and less critical, or feeds them plausible information in a pleasing way so it’s accepted uncritically. Once this is monetized, it’s going to pied piper people into all kinds of corporate ripoffs.
> Is that a bad thing?
Yes, it is a bad thing to be over-optimistic instead of thinking, to make optimistic assumptions that could lead you to a wrong conclusion.