6 ms·

If there's excess demand at night, you increase the price at night. If there's excess supply in the morning/afternoon, you lower the price. It's not complicated

by alexb_ 4y ago

If there's excess demand at night, you increase the price at night. If there's excess supply in the morning/afternoon, you lower the price. It's not complicated, and the problem fixes itself. You don't need AI to simulate what groups of people making rational, independent decisions do themselves. That's how you get cars that won't charge because "AI says so", similar to how all of these dogshit smart thermostats can be completely shut off from heating if "AI says so".

gpm 4y ago
The car needs to charge for 4 hours overnight (say, before 7am). It's currently 9pm and the price is currently 7c/kwh, should it charge or not? That question is obviously underspecified, you can only answer it by knowing whether or not the price will be below 7c/kwh for 4 of the remaining hours in the night. You answer the question with a predictive model of the price, and you build that predictive model with "AI" because that's how we build all our best predicitve models. AI doesn't get to say "you can't charge", you get to say "and I'll delegate the question of which 4 hours to charge during to the AI". > similar to how all of these dogshit smart thermostats can be completely shut off from heating if "AI says so". I assume you're referring to the instance where texas power companies remotely increased the temperature on peoples thermostats? That's not AI, or even remotely related to AI, that's humans controlling other peoples stuff without permission. That risk exists whether or not you use machine learning models to predict electricity prices. https://www.khou.com/article/news/local/texas/remote-thermostat-adjustment-texas-energy-shortage/285-5acf2bc5-54b7-4160-bffe-1f9a5ef4362a https://www.khou.com/article/news/local/texas/remote-thermos...
3pt14159 4y ago
You don't get what I'm saying. I'm saying market dynamics (ie, setting the price lower when the grid needs less power) plus the inevitable AI that will game it to everyones advantage will solve this problem automatically. There is no reason to worry about this. This is trivial AI.

by alexb_ 4y ago

gpm 4y ago

The car needs to charge for 4 hours overnight (say, before 7am). It's currently 9pm and the price is currently 7c/kwh, should it charge or not? That question is obviously underspecified, you can only answer it by knowing whether or not the price will be below 7c/kwh for 4 of the remaining hours in the night. You answer the question with a predictive model of the price, and you build that predictive model with "AI" because that's how we build all our best predicitve models. AI doesn't get to say "you can't charge", you get to say "and I'll delegate the question of which 4 hours to charge during to the AI". > similar to how all of these dogshit smart thermostats can be completely shut off from heating if "AI says so". I assume you're referring to the instance where texas power companies remotely increased the temperature on peoples thermostats? That's not AI, or even remotely related to AI, that's humans controlling other peoples stuff without permission. That risk exists whether or not you use machine learning models to predict electricity prices. https://www.khou.com/article/news/local/texas/remote-thermostat-adjustment-texas-energy-shortage/285-5acf2bc5-54b7-4160-bffe-1f9a5ef4362a https://www.khou.com/article/news/local/texas/remote-thermos...

3pt14159 4y ago

You don't get what I'm saying. I'm saying market dynamics (ie, setting the price lower when the grid needs less power) plus the inevitable AI that will game it to everyones advantage will solve this problem automatically. There is no reason to worry about this. This is trivial AI.