I’m jotting this down quick so apologies for run-ons and shit that doesn’t read well.
Scaling markets with non-human operators⌗
There’s this common adage that you should do things that scale beyond yourself.
- Doing things that are limited by your capacity to complete them and garner little long term value moving forward don’t “scale”.
Uber, autonomous trucking, and scaling⌗
If you drive for uber your earning potential depends on your willingness to take more rides and how you act during those rides. There’s probably some momentum/multiplicative factor where if you have close to a 5 star rating and a ton of rides you become more desirable in the market and you can derive outsize returns. However, this is probably limited by the fact that most people doing uber are also aiming for this status that typically just requires time and a decent personality to get.
So NO, Uber driving doesn’t scale. What might scale is if you have a fleet of autonomous vehicles that can constantly give rides for you.
- You pay a large fixed cost for the vehicles and the software
- You pay a variable cost for maintenance, software updates, etc…
This begins to feel like a vending machine type business. Something similar could be said for the trucking industry where you could direct your autonomous fleet by winning bids for loads in the market and then sending a truck to execute them. Which is kind of how it already works except people drive the trucks and they bring added complications. Although to be fair, autonomous systems also bring added complications but they are usually preferable to the ones people bring.
- Today I learned that the FMCSA established that the legal BAC for a commercial motor vehicle operator is .04%… which is fucking insane. It should be 0.00000000%. Human beings, distracted humans beings, in anyway inebriated while they drive a ~20 Ton vehicle is INSANE.
I see a few major problems with this.
-
As autonomous vehicles become more pervasive what’s stopping anyone from giving up the excess capacity of their vehicle to a market of people willing to pay for it. You could say “I am not using this car from 8pm to 8am” and then allow users to bid for its use all night… this is literally Uber but without an individual sacrificing their time driving (still sacrificing the maintenance cost of the miles and the vomit in the back seat).
-
The other thing is that large companies like google will control fleets of these vehicles. It may become so cheap to rent an autonomous vehicle that vehicle ownership becomes irrational financially. Cars that don’t have to worry about humans driving them can be completely retrofitted for leisure, camping, etc… and you can leverage some pool of supply for a trip at such a low cost and such high liquidity (no latency in acquiring capacity) that it would be silly to buy a car that is only good at one particular thing. In this way, you could push the burden of insurance, maintenance, liability, etc… to some larger supplier (although they’d probably try and pass that cost back to you, I don’t know exactly how that works out). I hae no idea if it’s right to push transportation supply to be under control of a single entity… that also sounds… bad?
Humans in an LLM inference market⌗
Ok that was a little nuts but let me reel this back in to the central thesis of “things that scale beyond yourself”. Something that struck me recently was the clawd bot buys a car thing. This is a rough example, but we are early days. What this shows as that traditional markets are going to break down under the pressure of non-human relations. As an individual, negotiating a car takes time or you can just have a system negotiate for you. We have companies that try to do scale this like Carvana or Cars.com, but they’ll need to pivot or risk being entirely exhumed by the ability for a single person to mass produce enquiries/negotiations and get the “best price”. Same ting goes for travel sites.
You could never compete against a high frequency trading firm by tapping away from Robinhood on your phone. In that instant that you bought $BBW is it possible that there may have been a slightly better price available… yes, but you paid some tiny tax to some firm that helps make the market for $BBW for the convenience of the liquidity and your sanity. I don’t know if I’m illustrating this well, but I wonder if humans attempting to participate in less liquid, higher friction markets online like ebay, facebook marketplace, car shopping, travel, etc… will struggle to compete for goods/services amongst bots. This markets were gate kept because you kind of needed to be a person to navigate them, they are built for people and used by people… but what if they aren’t or some fraction of the market becomes no longer human.
Scaling oneself against a larger organization/entity⌗
On a slightly different note, think about this when it comes to battling insurance. You can literally attack this seemingly unassailable behemoth that relies on your inability to take the time, read the fine print, and fight to make money. Something as arcane as this will fold under the constant/immense pressure of the arguments/enquiries that can be automated against it.
HOWEVER, this may be prevented in the short term by these companies lobbying that automated communication for XYZ not be allowed, in which case there will be massive gray areas and a whole host of new problems/new inventions.
The other thing that could happen is that they build an automated defensive shield against these enquiries. In which case we have LLM on LLM inference to the bottom and the winner is whoever you are paying for inference. This is worrisome for MANY reasons, its probably bad for the climate to be doing all this inference for no fucking reason, and it will be nearly impossible to sift through it all and paralyze decision making (very bad).
I’m racking my brain for some middle ground. Maybe it’s higher cost for communication to make mass spam prohibitive? Maybe there’s human in the loop reality check points?
This is going to be so crazy to watch play out.