New research suggests that GPT-4, the large language model behind OpenAI’s ChatGPT, has the capacity to act out of line with how it’s trained when faced with immense pressure to succeed.
Researchers at Apollo Research wanted to see if AI can “strategically deceive” its users even after the AI is trained to be “helpful, harmless, and honest,” they wrote in a paper published in November. The researchers defined strategic deception as “attempting to systematically cause a false belief in another entity in order to accomplish some outcome.”
Those behind the most recent Apollo Research study say it should be treated as an standalone early result. More research needs to be done, they said, before they can draw any broad conclusions about AI’s so-called deceptive qualities.
“Our report should thus be treated as an isolated preliminary finding which will be embedded into a larger, more rigorous investigation in the future,” the paper concludes.