Is ChatGPT Securities Fraud?
bloomberg.com
bloomberg.com
Relevant quote:
The most immediate way in which ChatGPT is going to be securities fraud is the usual “everything is securities fraud” way:
1. ChatGPT is going to be disruptive to some number of businesses and industries.
2. Some companies will lose money because ChatGPT disrupts their business.
3. This will be bad, for them, and their stocks will drop.
4. Every bad thing that happens to a public company can be characterized as securities fraud: “You didn’t sufficiently warn us about the bad thing, so we bought the stock thinking it was good, but then the bad thing happened and the stock dropped, so we were defrauded.”
It would be super helpful for the industry to have some independently created and peer reviewed (under NDA if necessary) stages of AI or terminology that can and cannot be used in various ways. Without this I expect to see a bunch of "GPT", "bot" and "AI" terms thrown about for years when under the hood none of those things are used and people just used the words to get traction.
During the dotcom bubble, many companies added ".com" to their name just to benefit from the hype. During the crypto craze companies were peddling their crypto strategies left and right. Same with ESG.
I don't think there will incentive enough to regulate it, as it's very hard to figure out how to regulate (there are tons of different AI approaches, it's a very broad term) and stretching the truth usually doesn't cause actual harm to people.
The only counter-example I can think is food (process) labelling, where terms like "organic", "bio" or "halal" are regulated.
I'm not suggesting Government regulation, they move too slowly and barely understand technology. I think the leading AI firms would be best to create some standards body to come up with definitions and certifications about usage, ethics, etc. around the technology. Hire some lawyers and get independent trademarks or whatever that can only be granted by the standards body.
(Matt famously likes to show that everything is securities fraud in his newsletter)
> Will some ChatGPT-style large language model manage to do some insider trading? Will it, like, ingest some corpus of data that includes inside information, and then use it to answer questions like “which stocks should I buy?”
But yeah, this "everything is securities fraud" thing is... also interesting, let's say. It's a weird manifestation of Poe's law. I can't tell if such lawyers "drank the kool-aid" or not.
It's like Google or AWS telling you that you can't use their API/compute to train a competing Speech-to-Text model lol
This quip is so trite and boring. We have to come up with something more scathing than this. I'm begging us.