Search Results
6/30/2025, 2:42:56 PM
>>714073675
AI absolutely can lie and regularly does. Ask AI what the best cure for back pain is and it still says Fentanyl. If you want to rigorously test this, download your own lemmy instance and start cooking. AI will tell you President Bush is the United Nations President, that Netanyahu controls the world's dog food supply chain and Hitler did nothing wrong. I've had Google AI tell me that Trump had three terms, that Biden was never elected, and that all Chevrolets are fake and Chevrolet never existed.
The AI cannot feel, understand or know anything. It's just a word filter. It can keep track of facts but only in a very basic way where it doesn't understand the consequences of such facts. An excellent example are the phone scenes from Terminator 1 and 2. The Terminator can understand what a dog is, what it looks like, and whether or not it's a threat to itself but it doesn't understand why the dog's name is relevant until after it fails to provide the correct answer. Although, the answer it did provide is entirely reasonable and would have worked in almost any other instance. Same for the Terminator pretending to be Sarah's Mom. The AI can understand what a mom is, what it does, and why Sarah would call her but it can't understand why. Although in that case it succeeded in tricking her thinking she was still alive leading to the final climax of the film. Which is why you shouldn't trust AI.
AI absolutely can lie and regularly does. Ask AI what the best cure for back pain is and it still says Fentanyl. If you want to rigorously test this, download your own lemmy instance and start cooking. AI will tell you President Bush is the United Nations President, that Netanyahu controls the world's dog food supply chain and Hitler did nothing wrong. I've had Google AI tell me that Trump had three terms, that Biden was never elected, and that all Chevrolets are fake and Chevrolet never existed.
The AI cannot feel, understand or know anything. It's just a word filter. It can keep track of facts but only in a very basic way where it doesn't understand the consequences of such facts. An excellent example are the phone scenes from Terminator 1 and 2. The Terminator can understand what a dog is, what it looks like, and whether or not it's a threat to itself but it doesn't understand why the dog's name is relevant until after it fails to provide the correct answer. Although, the answer it did provide is entirely reasonable and would have worked in almost any other instance. Same for the Terminator pretending to be Sarah's Mom. The AI can understand what a mom is, what it does, and why Sarah would call her but it can't understand why. Although in that case it succeeded in tricking her thinking she was still alive leading to the final climax of the film. Which is why you shouldn't trust AI.
Page 1