So it goes with the animal-spillover theory. The evidence in favor has always been highly esoteric, knotted with data and interpretation. Scientific points are made—a particular run of viral nucleotides is a “smoking gun” for genetic engineering, one famous scholar said in 2021—and then they are re-argued and occasionally walked back. Long-hidden sample data from the market suddenly appear, and their meaning is subjected to vituperative, technical debate. If the evidence for a lab leak tends to come from messy human stuff, the evidence for animal spillover emerges from messy data. Simple-seeming claims are draped across a sprawl of numbers.

John Durham testified before the House Judiciary Committee, and admitted he didn’t know about several elements of the Trump campaign colluding with Russia: releasing stolen email through cutouts, Trump negotiating a tower deal, Manafort and Kilimnik.

So I enlisted an African-American faculty member in my department to argue in favor of eugenics while I argued against; halfway through the debate, we switched sides. We were modeling for the students a fundamental principle of a university education, as well as civil society: When you’re forced to engage a position you strongly disagree with, you learn something about the other perspective as well as your own. The process feels unpleasant, but it’s a good kind of stress — temporary and not harmful to your body — and you reap the longer-term benefits of learning.

Researchers discovered a unique property of psychedelic drugs: their ability to reopen "critical periods" in the brain, times when the brain is highly susceptible to environmental learning signals.

The remarkable zero-shot learning capabilities demonstrated by large foundation models (LFMs) like ChatGPT and GPT-4 have sparked a question: Can these models autonomously supervise their behavior or other models with minimal human intervention? To explore this, a team of Microsoft researchers introduces Orca, a 13-billion parameter model that learns complex explanation traces and step-by-step thought processes from GPT-4. This innovative approach significantly improves the performance of existing state-of-the-art instruction-tuned models, addressing challenges related to task diversity, query complexity, and data scaling. The researchers acknowledge that the query and response pairs from GPT-4 can provide valuable guidance for student models. Therefore,

The three scientists were engaged in “gain-of-function” research on SARS-like coronaviruses when they fell ill

Mojo aims to be as easy to use as Python, but as powerful and fast as Rust. Here's a first look at Python's newest challenger.