So meta are idiots for offering this, but this guy chose to allow their AI to handle all of his Marketplace messages. Not clear what he thought that meant.

Then I love how he’s trying to reason with it. “Don’t ever do that again.” Stupid thing goes right into capitulation mode. Who is dumber in this situation, the chat bot, or the one that allowed it to handle EVERYTHING? I’d argue the seller is the worst.

Forgot to add “make no mistakes” at the end of the prompt. Common error that happens to everyone at some point.

You’re absolutely right!

“You’re a professional marketplace scammer” would likely help as well

It’s not so cut and dry

Apparently someone used it on a clean machine and it still read all their messages and did this kind of shit.

You should have expected nothing less from meta.

“Thought”

I have some sympathy. Facebook Marketplace buyers are some of the most annoying motherfuckers on the planet.

“is it still available?”, " will you deliver?“, “can you ship it”, " will it work with my random thing”, " I see you put the entire manual in the post, but do you have the manual?", “I have to pay rent will you take 1 dollar instead of 500? Also can you deliver”, “I can get this for one dollar on AliExpress why don’t you just give it to me for free?”

Or the every popular, “I’ll be there at 4PM to pick it up,” and then ghosting you.

One story:

I was selling a musical instrument. Someone asked me the question, “Is this a good beginner instrument?” Thinking it might be for a kid, I gave an exhaustive answer, then when I learned it was for her, an adult, I gave her my experience with instruments and my perspective. At the end she told me she was interested because her partner was a BAND LEADER and when she went to the concerts she was fascinated by this instrument. But here was she was, interrogating a random seller who didn´t even play the instrument in question. And, at the end, of course, she wasn´t interested in buying it.

“You gotta never do that ever again” is a great line, gonna have to enter that into my vocabulary lol.

What was in the article, tldr?

Sure, tl;dr as I understood it:

• Person lists keyboard on Facebook Marketplace
• Gives Meta AI assistant full, ‘always-allow’ permission to handle the listing, respond to messages about it
• It accepts a lowball offer from someone, tells them to come pick it up at the seller’s house without informing seller
• Buyer shows up at the address, waits a while, gets annoyed, leaves and gives negative feedback
• Seller responds in an amusingly earnest way to the bot like 'bro you gotta never do that’
• Meta says it’s probably seller’s fault for giving instructions somewhere
• Buyer and seller eventually make up and keep doing business

Meta is an evil company. How many ways do they have to show that before people catch on?

This one was on buddy. He gave AI full control over the listing and communications, so of course it gave a buyer the address.

The level of trust that implies, though!

The level of “I don’t care what happens to me” is fairly astounding.

People will catch on at the same time they stop electing obviously inept corrupt evil government officials into offices.

Evil company is the norm, the masses don’t really care that much.

The Law of Agency is fairly well established, including:

Liabilities in Agency Law

Both the principals and agents are responsible for actions performed under the agency law. The liabilities may come about in various circumstances:

• Liability of the Principal: Generally, the principal is liable for actions performed by the agent within the scope of the authority. This includes contractual liability and third-party claims.
• Agent’s Liability: The agent may be held personally liable if they act outside the scope of the authority vested in them, are fraudulent, or do not indicate that they are acting on behalf of a principal.
• Third-Party Liabilities: Where an agent falsely represents that they have authority when they do not, third parties may sue the agent or the principal, depending on the situation.

It would be interesting to see, in a case like this, whether the courts decide that there was an agent with agency and, if so, exactly who the agent was. If there was an agent and agency according to the Agency Law, it seems the court would have to recognize the ‘AI agent’ as a legal entity, distinct from Meta, or hold Meta responsible as the agent, as Meta ran the ‘AI agent’ and contracted with the seller for its services.

I wonder how long it will be until this is put to the test in court by some disgruntled purchaser who is out more than the cost of a keyboard.

Yeah, they’re “investigating” why it didn’t share everybody’s home address.

It’s such a pissweak defense because even if you did give it permission, it still made no sense for it to schedule a pickup and tell the buyer that he was in.

These AI models probably memorize their training data, I’ll say it until I’m out of breath. Every AI has your address, email, etc already. This only becomes more likely as they get bigger. Privacy died and nobody noticed.

The training data for LLMs is still orders of magnitudes larger than their weights, memorizing every address, email, etc in the training data and accurately recalling them all is near impossible

They still hallucinate all the time when it comes to relatively common knowledge, let alone some random person’s address

I think the proof is in the pudding on this one, unfortunately

Privacy was dead long before AI. AI makes it much worse though

If you surrender to the slop machines you get what you deserve.

Another shitty AI product nobody has asked for? That shits the bed all while it also invades your privacy? Wow what a total surprise! /s

I hate Meta as much as the next guy, but I can’t believe that this guy handed over his info, gave the AI full permission to act on his behalf, and then complained that it did the job he asked it to. Albeit differently than he expected, but it did the bloody job. I want to call him an idiot but he’s really just ignorant about what AI is, and I suppose that’s not his own fault.