5h ago

Wharton's Ethan Mollick and Gary Marcus debate whether historical machine learning data barriers limit generative AI returns

2 top authors

Marcus cites ten studies showing ongoing corporate data barriers.

Original post

I actually read this & it is super weird, it appears to be an argument that prior machine learning systems (not generative AI) did not generate savings due to data issues so that will lead to a lack of investment into current AI systems Also it cites the mostly fake “MIT study”

7:07 PM · Jun 3, 2026

Sentiment

Pos50%

Neg50%

Positive users affirm current AI is worth pursuing despite warranted skepticism around data barriers in the Bain report, while negative users dismiss the report as repackaged old data modernization problems.

2 comments with sentiment.

1 more post

Gary Marcus@GaryMarcus·4hReply
but haven’t there been like 10 studies pointing to the same conclusion as that MIT study? they basic thrust of the MIT study has repeatedly supported. unless i am missing something, you can argue that newer studies yet to come could be different but not that the MIT study was at odds with rest of the last couple years’ analysis. cc @JohnCassidy
View on