“CriticGPT’s options will not be at all times appropriate, however we discover that they might help trainers to catch many extra issues with model-written solutions than they might with out AI assist,” the corporate wrote. “Moreover, when folks use CriticGPT, the AI augments their expertise, leading to extra complete critiques than when folks work alone, and fewer hallucinated bugs than when the mannequin works alone.”
And therein lies the logic downside right here. One of many criticisms of generative AI is that it’s terrific at mimicking people however fails to really perceive people. I’m reminded of a column I wrote greater than a decade in the past, about engineers making a product that assessments for real love. (It was an precise product: a Bluetooth bra that will unhook solely when it detected real love. Actually. To be clear, I’m not formally suggesting that engineers are as dangerous at perceive human feelings as genAI. Not disputing it, but additionally not formally saying it.)
Getting again to genAI logic, the flawed assumption that OpenAI is making is that people will proceed checking their methods for lies. People are lazy, and human IT staff are overworked and under-resourced. The much more possible end result is that people will belief the AI-watching-AI increasingly. That’s the place the actual hazard exists.
