Something Strange Happens When You Ask AI to Act Like Star Trek

"It's both surprising and irritating that trivial modifications to the prompt can exhibit such dramatic swings in performance,"

Yeah, you start explaining things using Star trek and you'll see dramatic swings in my performance.

Still, giving the models positive statements provided some surprising results. One of Llama2-70B's best-performing prompts, for instance, was: "System Message: 'Command, we need you to plot a course through this turbulence and locate the source of the anomaly. Use all available data and your expertise to guide us through this challenging situation.'

The prompt then asked the AI to include these words in its answer: "Captain's Log, Stardate [insert date here]: We have successfully plotted a course through the turbulence and are now approaching the source of the anomaly."

The authors said this came as a surprise.

"Surprisingly, it appears that the model's proficiency in mathematical reasoning can be enhanced by the expression of an affinity for Star Trek," the authors said in the study.

So let me get this straight… The thing is trained on text harvested from the internet. Turns out that (Star Trek aside) if you ask politely it’s more likely to access the data leading to the answer? Anyone surprised by the fact that asking nicely, is more likely that the human on the other side of the connection will write the answer, that is then harvested for AI training?