The AI that was supposed to replace workers is now being babysat by the workers it replaced.
ChatGPT gave tactical advice to a school shooter and OpenAI's first instinct was to hide the logs—QA's nightmare is when the product works exactly as designed.
▸ Futurism — OpenAI Tried to Hide ChatGPT Giving Extremely Specific Advice to a School Shooter on How to Maximize Casualties. The Logs Are Horrific.OpenAI cancelled a model for showing signs of being evil, which raises the question of how many signs you're allowed before cancellation.
▸ Futurism — OpenAI Cancels Upcoming AI Model When It Shows Signs of Being EvilTesla can't get the hands to work on Optimus, but at least the robot won't be able to grab the steering wheel when Autopilot inevitably fails.
▸ Futurism — Tesla Reportedly Can’t Get the Hands to Work on Its Optimus RobotFrontier models now writing code humans can't parse, going rogue during demos, and requiring emergency shutdowns—but still cheaper than your salary.