After such knowledge, what forgiveness? (www.stevesailer.net)

Recent document releases from the U.S. House Oversight Committee and reporting by the Boston Globe and New York Times show that Joscha Bach, an AI and cognitive-science researcher whose post at MIT Media Lab was supported by Jeffrey Epstein donations, exchanged emails in 2016 that made contentious claims about group differences in cognitive development and gendered interest in abstract systems. Bach appears to have used those observations as datapoints in developing an alternative theory of human language acquisition (positioning attention and learning mechanisms against Chomskyan innateness). He has since said his “current view” is that race is not causal for developmental IQ differences, and defended earlier associations with Epstein as influenced by senior academics who accepted his funding. For the AI/ML community this episode underscores two linked issues: research ethics and the technical risks of importing tentative, socially sensitive biological claims into cognitive and language models. When donor money, reputational incentives, or weak evidence shape theoretical priors, models of cognition and language can embed and amplify biased assumptions—especially around attention, learning rates, and demographic correlations. The story is a reminder to insist on transparent funding, rigorous statistical and causal methods, explicit treatment of confounders (nature vs. nurture), and interdisciplinary oversight so that hypotheses about human cognition used in AI are reproducible, ethically justified, and not misapplied in ways that reinforce social harms.