AI Roundup March 6, 2024
Get ready for $1000 burger YouTubers; Suing superior AIs because yours is worse and your last name is Musk
How long till a YouTuber buys out a Wendy’s to get the price up to $1000?
Last week, Wendy’s, the 12th largest fast food chain in the country by revenue announced that it will be overhauling its menu software to allow for what they are calling “dynamic pricing”. So far, I think a lot of the analysis has been superficial and not understanding exactly how many data points can be fed into this model.
First, Mr. Beast or one of his comrades is totally going to buy the entire stock of a Wendy’s then show the price go up to $1000 and then give out all the burgers they bought for free. Hopefully someone anticipates this and puts in some backstops to prevent this from happening (or at least coordinates it in a way that doesn’t drive up prices in some sort of ripple effect).
But let’s look at some other things that can be used to set prices for burgers. We already know that fast food places are located in areas that are more likely to have a high-prevalence of obesity. So is there anything stopping Wendy’s from feeding in obesity data and lowering the prices of burgers there? Alternatively, you could say that they may be able to incentivize salads in these areas. And while (at least according to the article I linked to) fast food restaurant access isn’t correlated with income levels, you could just as much inflict obesity issues by pricing your food at attractive prices to entice buyers.
The algo could also be hooked up directly to beef prices, which in theory could allow bigger competitors to Wendy’s to drive up the prices of Wendy’s burgers as the bigger companies buy up larger pieces of meat futures - a la Trading Places.
I assume someone thought this through, because I imagine the up-front investment will be quite expensive as well as maintaining the pricing network from hackers. Imagine late-night hackers driving down prices of burgers at their local Wendy’s to 1¢ so they can fuel their hacking sessions.
The future will be interesting!
Lawsuits make everything better
I thought I would have a lot to say about two lawsuits regarding AI that were filed over the weekend. One made big news because it was brought by Elon Musk, who is claiming that OpenAI has deviated from its stated mission of being “open”. He’s probably right. What a lawsuit is going to do about it I’m not so sure about. Also, perhaps it is a small case of the pot calling the kettle black because Musk’s Grok AI is also not open source. It may also be driven by Grok’s inferiority in some areas vs. Chat-GPT4. Seems like a good opportunity, Mr. Musk!
The other lawsuit deals with whether copyrighted information can be consumed and then leveraged in AI models. I don’t know what to say about this. I don’t know how one could prevent copyrighted information from entering into a model, and I also don’t know what sort of compensation model could be proposed. I will say that those who are using AI to generate content for themselves should definitely check or understand if there is any copyrighted material in the responses they receive, and then act accordingly.
Not allowing copyrighted material into models could be a huge setback for model building, but I’m not sure we are far enough down the line in AI usefulness that starting from scratch would be terribly detrimental, after all, it can only successfully complete 14% of tasks a human can at this point.
Out of time folks.