> People needed to be evacuated, but left to its own devices, our AI would have discouraged them with a higher price.
There's two sides to the market. If off-duty drivers saw the higher price, they would be inclined to join. The price conveys information to both parties. Similarly, if I could, but don't have to, take a cab, seeing a very high price would dissuade me. But if I'm in a dangerous situation, I would gladly pay that price. Prices exist whether you want them or not. Charging everyone a low price when there are extraneous circumstances that affect the supply or demand doesn't miraculously solve the mismatch.
Variable pricing in general is a good idea, but you obviously need caps and floors. For instance, I have an adjustable rate mortgage. It makes sense to me, and when it starts adjusting the maximum rate it can go up to is around 10%, still not unreasonable worst case scenario. These contracts apparently didn't have a reasonable maximum rate, but there's no reason they can't be built with one.
We should encourage transparent prices. If energy is constrained, we should discourage people from using energy if they can and encourage supply. And the simplest way to do that is the price system.
There is a limit to the amount of drivers that can be made available, and that limit is so much lower relative to the norm that the limit in demand during an emergency.
> Contrast this with what we have heard about electricity providers in Texas. The surge was allowed to skyrocket up nearly 100-fold, just when people needed power most to survive.
Texas regulators set both the surge maximum and ensured that the wholesale price was set at this maximum for days on end. This was not a market exploitation of surge pricing but rather a regulator’s decision to try to manage a gross imbalance of supply and demand.
I didn’t hear about this change in electric pricing until after the blackouts were over and electricity was restored, mostly because I didn’t have hardly any internet access during the storm. Was my demand supposed to decrease? Because every time the power came on my heat was on max and I would charge as much stuff as possible, boil water, etc. because I had no idea when I’d have power again.
Surge pricing is great within limits (mostly monetary), to help with keeping an actual network effective. If regulations are set correctly (max surge price factor, max highprice interval in a pre-set time interval,...), and surge-pricing-aware devices are available (eg. water heaters that turn off at higher price, and turn back on at lower).
Going from a couple of cents to $9 per kWh, for extended time periods, without users actually having an alternative is well... pretty shitty.
A nice, strong piece. I love the example of people accepting to pay through their teeth to get to the hospital. Makes a nice, clear case for all the research on ethics in ML.
I think ethics is important and it's admirable to help others during a crisis, both as an individual and as a company. But I don't think this kind of stuff belongs in research. If you want to bias or overrule your algorithm to perform charity or social change or whatever, pay some engineers to develop a product that does it. Social goals and priorities are simply not science.
And even the hospital example isn't so simple. People who are willing to pay more to get to the hospital are also customers who should be prioritized higher.. since they're willing to pay more. They would get better service. And how does the A.I. know they don't just work in the cafeteria or just need a checkup or whatever, causing it to lose out on thousands of customers who aren't in emergencies? It would need to offer two tiers of prices, those who want high priority service and those who can wait for normal service. Seems like it could end up being a benefit after all.
Sounds like the author knows a lot about AI and data but almost nothing about economics.
Prices exist for a reason; if power became more expensive that's because it was scarce. People need to stop using it or pay more so that power companies can build more supply.
There are also options like the Tesla Powerwall that will be helped by this event.
"we would never build a system that might “learn” that people whose destination is a hospital tend to be willing to pay higher fares"
I'm sorry, but wouldn't that exactly be the point of "surge pricing"? To allow people who REALLY need a ride to be able to pay a higher price and get where they need to be?
"Surge pricing" would kick in because of scarcity, as in, not enough rides available for everybody.
The person who needs to go to the hospital needs it more urgently than the person who wants to meet her friends at the pub, so she should be able to pay a higher price to get where she needs to be (the hospital).
The alternative is that she doesn't have the option of paying a higher price out of "ethical considerations", and has a high probability of not actually finding a ride to the hospital. How is that the better variant.
If you want to get all ethical about it, maybe you can set up a system where people get their money back if they can prove they needed to go to the hospital (or health insurance could cover it). But it is unlikely to cover all possible cases of urgency ("needed to flee from a terrorist attack", "needed to catch the love of my life before she marries somebody else" and so on)
. Much easier and more effective to regulate by market prices.
It seems to me yet another case of misguided socialist ideology causing harm in the real world.
An even more cynical interpretation would be that companies prioritize "looking good" over saving lives, so they rather implement the virtue signaling "no surges for hospital rides" than actually letting people catch their ride to the hospital.
If you are worried people might not be able to afford the ride to the hospital, you could give them credit.
I think you're missing what the concern is. You're talking about a generic increase in prices due to scarcity.
What they're talking about is a system that identifies that people getting a taxi to the hospital may have no/few other choices and you can just charge them more, regardless of how how many drivers are available.
No, surge pricing would be "lots of people want to travel right now and we don't have the capacity, increase the price to maximise return". The averaging there is "all users in a wide area" rather than "all people that need to go to local place X". Both may cause ethical issues but the latter has a specific kind of problem.
What a smarter but unethical system may discover is "people having a heart attack will pay almost any amount of money to get to hospital". Substitute heart attack for any other medical issue, if that is something you'd say should require an ambulance.
Location based pricing may be fine, charging more for airport travel because there's less competition (due to waiting times, etc) could be fine. Charging more for getting medical treatment is not.
Optimising blindly for profit is not something we want to do, but can easily be what we task a machine to do.
There is a difference between trying to determine the highest price somebody may be willing to pay and auctioning off to the highest bidder (with an AI trying to guess the auction prices also being a kind of auction). The point is exactly that somebody with a heart attack has the option to pay more so that they can a ride.
Yes it would be nice to be able to detect "this person has a heart attack, let's give them a free ride to the hospital". But it is not really viable in practice yet. So giving them an option to pay more is the second best option.
Now they decided to remove that option, so how are people with heart attacks supposed to get to the hospital?
There's two sides to the market. If off-duty drivers saw the higher price, they would be inclined to join. The price conveys information to both parties. Similarly, if I could, but don't have to, take a cab, seeing a very high price would dissuade me. But if I'm in a dangerous situation, I would gladly pay that price. Prices exist whether you want them or not. Charging everyone a low price when there are extraneous circumstances that affect the supply or demand doesn't miraculously solve the mismatch.
Variable pricing in general is a good idea, but you obviously need caps and floors. For instance, I have an adjustable rate mortgage. It makes sense to me, and when it starts adjusting the maximum rate it can go up to is around 10%, still not unreasonable worst case scenario. These contracts apparently didn't have a reasonable maximum rate, but there's no reason they can't be built with one.
We should encourage transparent prices. If energy is constrained, we should discourage people from using energy if they can and encourage supply. And the simplest way to do that is the price system.