The interaction between AI & past scholarly work is going to get weird. Here I gave GPT-5.5 Pro a copy of my first published paper from grad school & asked it to find errors and update it.
It found new data, analyzed it, created reproducible files, extended the key argument...
Users praise GPT-5.5 Pro for updating the 2006 Moore's Law paper with more sophisticated analysis, error checking, new data, and reproducible files, preferring it to typical AI hype.
No Digg Deeper questions have been answered for this story yet.
This is good stuff, including some things that are much more sophisticated than what I wrote in paper long ago. What happens when we turn this sort of AI loose on past academic research at scale? Should we be doing that already?
The interaction between AI & past scholarly work is going to get weird. Here I gave GPT-5.5 Pro a copy of my first published paper from grad school & asked it to find errors and update it.
It found new data, analyzed it, created reproducible files, extended the key argument...
The original paper, by the way: https://www.researchgate.net/publication/3331068_Establishing_Moore's_Law
This is good stuff, including some things that are much more sophisticated than what I wrote in paper long ago. What happens when we turn this sort of AI loose on past academic research at scale? Should we be doing that already?
@emollick This is wonderful for human progress but what we don't want is revisionist history. This is unquestionably fantastic as it does highlight what has changed but it can certainly open the flood gates of undue criticisms or a myriad other issues.
@emollick Do it. Also, does the AI model get a PhD, too, for doing this research?
@emollick Yes. AI can now create an infinite amount of datasets with the same summary statistics and correlations. Rerunning analyses on those datasets illuminates who found the one in a million dataset that produced significant results
@emollick I'd love to see that type of work done, especially with qualitative research where the original content/field notes/interviews are fully available for new analysis. Be interesting to see what findings are reinforced and what new insights are gleaned.
@emollick Yes, and please get on dedicated AI peer review in every discipline, along with chatbot level skills for laypeople to evaluate what they see casually online.
@emollick Yes, but we should converge on some scaffolding for this
@emollick This is where AI gets really interesting. Error checking, new data, reproducible files, updated argument. I’ll take this over another buzzword-heavy AI panel every time.
