You could order your groceries yourself. Or you could engage in an excruciating Rube Goldberg machine of a conversation by micromanaging an AI agent to describe what food you need from the supermarket — just so it can mess up your order anyway.
What, not excited for this billionaire-ordained future, you Ludditical lunk? Why is that? Is it because you’re too busy laughing at this AI meat proxy who admitted, of his own free will, that “an LLM hallucination resulted in 72 eggs being delivered to my house”? Because that’s not funny at all.
“I put 4 cartons of Vital Farms pasture-raised large eggs (18 count each) in the cart you checked out,” the AI agent said, in a screenshot shared on X by Matt Smith from Rebellionaire, an equity firm led by hardcore Tesla investors.
Smith, in the screenshotted exchange, demanded an explanation. But the AI couldn’t give one.
“I don’t have a good answer,” the agent admitted. “I can’t find any note explaining why I chose 4.”
The AI then suggested that it misread the line for eggs in a photo of the list Smith provided — except, as he immediately pointed out to the AI, “there wasn’t a line for eggs on my list.”
“Then I added them without any reason, and that’s on me,” the agent said, in a classing apologetic AI turn of phrase.
Needless to say, there’s a simple fix to all this: not using an AI agent to automate every part of your life. As impressive the tech can be, it’s still error-prone. Nonetheless, the AI industry is encouraging users to provide their models with more information and control than ever, including intimate financial access like their bank accounts.
The consequences can sometimes be pretty creepy. Shortly after Meta transformed the space with its Muse AI agent, a man said that Muse gave his address to strangers on Facebook Marketplace without his knowledge and sent them to his home.
For his part, Smith said the mishap was the “first time I’ve experienced a real-world consequence for an LLM hallucination.” All things considered, it could’ve been a lot worse.
More on AI: Remember When Bosses Were Forcing Their Workers to Use AI? They’re Now Realizing They Made a Huge Mistake
The post AI Agent Can’t Explain Why It Randomly Ordered its Human Meat Proxy an Absurd Number of Eggs appeared first on Futurism.
