Yet Another Way AI Has Outrun the Law: Gas Pricing

Recent advances in AI that change the price of gas minute-to-minute are, as Sam Schechner writes, “testing a fundamental precept of the market economy”. The precept is that steady price competition pushes prices down. But when algorithms are let loose to predict “what competitors are charging and what consumers are willing to pay”, the opposite apparently comes into effect sometimes. The free (gas) market run by AI sometimes pushes prices up instead.

Rather than older, common AI systems that simply priced gas lower than competitors, newer systems like those used by some gas stations, “crunch mountains of historical and real-time data” to anticipate how both buyers and the market will behave in advance, crucially allowing them to predict what will happen in the near future and price accordingly. For instance, are people more likely to pay higher prices during rush hour? Does a current event on the other side of the world suggest prices will dip overnight?

As is often the case, this technology has outrun the current legal framework in place for regulating markets—specifically when it comes to antitrust law. “Such software could make collusion easier "without any formal agreement or human interaction”, essentially embedding monopolization into an automatic pricing system.

Schechner’s reporting is a healthy introduction to the entanglement of AI and market economics through this specific example of gas pricing. It also covers the basics of neural networks that allow increasingly complex calculations with higher potential for prediction. Many of these advances can also be seen as consumer friendly (anticipating your needs will likely make your shopping faster), but everything comes with, well, a price.