Hmm… I see a role for humans producing meta research. Or simply research on a higher level of abstraction, interpreting what the AI generated research really tells us. I doubt we will move into a world where all of us just blindly accept AI outputs.
The results and outcomes are for humans, it definitely makes sense that human effort is spent on verification , interpretation and deriving action items if any.
Many good ideas there. Here's a little primitive anecdote:
It was in the first months of ChatGPT public availability, so the paper I was asked to review (for a dental science journal) was still 100% hand written. As was my review - which had only minor quibbles. For laughs, I asked ChatGPT to review the article as well - and it found a grotesque bias that had escaped human reviewers for two decades.
The study was a sponsored comparison of a manual vs a powered ("electric") toothbrush. The researcher blandly stated that all subjects were required to use their brushes as per manufacturer instructions. I missed the implications.
The sponsor's electric toothbrush comes with detailed instructions including brushing duration and frequency. The manual toothbrush is packaged with none. Subjects in the 'powered' arm were required to adhere to an ideal brushing schedule. Control subjects had no brushing requirements at all.
When confronted with this, the author said they've been doing it like this forever.
Hmm… I see a role for humans producing meta research. Or simply research on a higher level of abstraction, interpreting what the AI generated research really tells us. I doubt we will move into a world where all of us just blindly accept AI outputs.
That makes sense.
The results and outcomes are for humans, it definitely makes sense that human effort is spent on verification , interpretation and deriving action items if any.
Many good ideas there. Here's a little primitive anecdote:
It was in the first months of ChatGPT public availability, so the paper I was asked to review (for a dental science journal) was still 100% hand written. As was my review - which had only minor quibbles. For laughs, I asked ChatGPT to review the article as well - and it found a grotesque bias that had escaped human reviewers for two decades.
The study was a sponsored comparison of a manual vs a powered ("electric") toothbrush. The researcher blandly stated that all subjects were required to use their brushes as per manufacturer instructions. I missed the implications.
The sponsor's electric toothbrush comes with detailed instructions including brushing duration and frequency. The manual toothbrush is packaged with none. Subjects in the 'powered' arm were required to adhere to an ideal brushing schedule. Control subjects had no brushing requirements at all.
When confronted with this, the author said they've been doing it like this forever.
Thank you, baby Heimdall.
This is a thorough breakdown of how AI is disrupting the Academic Paper hustle both today and likely patterns for tomorrow.
Great read!