I think the real danger here is subtlety. What happens when somebody asks for recommendations on a printer, or complains about their printer being bad, and all of a sudden some long established account recommends a product they’ve been happy with for years. And it turns out it’s just an AI bot shilling for brother.
For one, well established brands have less incentives to engage in this.
Second, in this example, the account in question being a “long established user” would seem to indicate you think these spam companies are going to be playing a long game. They won’t. That’s too much effort and too expensive. They will do all of this on the cheap, and it will be very obvious.
This is not some sophisticated infiltration operation with cutting edge AI. This is just auto generated spam in a new upgraded form. We will learn to catch it, like we’ve learned to catch it before.
I mean, it doesn’t have to be expensive. And also doesn’t have to be particularly cutting edge. Start throwing some credits into an LLM API, haven’t randomly read and help people out in different groups. Once it reaches some amount of reputation have it quietly shill for them. Pull out posts that contain keywords. Have the AI consume the posts and figure out if they have to do with what they sound like they do. Have it subtly do product placement. None of this is particularly difficult or groundbreaking. But it could help shape our buying habits.
I think the real danger here is subtlety. What happens when somebody asks for recommendations on a printer, or complains about their printer being bad, and all of a sudden some long established account recommends a product they’ve been happy with for years. And it turns out it’s just an AI bot shilling for brother.
For one, well established brands have less incentives to engage in this.
Second, in this example, the account in question being a “long established user” would seem to indicate you think these spam companies are going to be playing a long game. They won’t. That’s too much effort and too expensive. They will do all of this on the cheap, and it will be very obvious.
This is not some sophisticated infiltration operation with cutting edge AI. This is just auto generated spam in a new upgraded form. We will learn to catch it, like we’ve learned to catch it before.
I mean, it doesn’t have to be expensive. And also doesn’t have to be particularly cutting edge. Start throwing some credits into an LLM API, haven’t randomly read and help people out in different groups. Once it reaches some amount of reputation have it quietly shill for them. Pull out posts that contain keywords. Have the AI consume the posts and figure out if they have to do with what they sound like they do. Have it subtly do product placement. None of this is particularly difficult or groundbreaking. But it could help shape our buying habits.