OLDaily

Algorithm beats humans for sniffing out fake news
Gabe Cherry-Michigan, Futurity, 2018/08/22

I'll stop harping on this soon. But again, I want to stress, using linguistic (or other) cues in the article itself is a poor way to identify fake news. That's what both the humans and the AI systems do as described in this article, and yet we find they're both wrong a quarter of the time. The way to distinguish between fake and non-fake is not to study the news item more closely, it's to find additional sources, to verify, to confirm. The only thing fake-news-detecting algorithms will produce is more convincing fake news.

Web: [Direct Link] [This Post]

I attended an academic conference and didn’t go to any sessions
Jennifer Polk, University Affairs, 2018/08/22

I've done this. Not on purpose, but I've had conferences get away from me, and I spend the entire time talking to people, doing interviews, working the trade show floor, and doing my own talks. But it's unusual and I really do try to see at least a representative sample of talks. But the point of the article - that there's a lot more to a conference than the presentation - is true.

Web: [Direct Link] [This Post]

“You Can’t Be What You Can’t See”
Larry Cuban, Larry Cuban on School Reform and Classroom Practice, 2018/08/22

The key message in this post is that researchers should look beyond the typical data collected in educational studies. In particular, they should "take a broad view of impact" and "take the long view". In particular, "Research on the effects of an education initiative typically looks only at later educational outcomes, but these investigators examined the impact of preschool on the totality of these children’s lives." Also, "Most education research adopts a short time horizon, rarely looking at the impact of a program beyond two or three years." For example - the early lessons I had in writing and public speaking had a lifelong impact on my character and career, but this impact would be invisible to pretty much every educational research study out there.

Web: [Direct Link] [This Post]

Architecture of the Mouse Brain Synaptome
Fei Zhu, Mélissa Cizeron, Zhen Qiu, Ruth Benavides-Piccione, Maksym V. Kopanitsa, Erik Fransén, Noboru H. Komiyama, Seth G.N. Grant, Neuron, 2018/08/22

This ground-breaking paper makes it pretty clear that knowledge consists of patterns of connectivity in the brain. These patterns now have a name: the synaptome. As Shelly Fan explains, " Like a map into internal thoughts, synaptomes drew a vivid picture of what the mouse was thinking when it made its choice... Like computer code, a synaptome seems to underlie a computational output—a decision or thought." There's a lot more in this paper - read it end-to-end. Keep reading, even if you don't understand bits. The paper is loaded with insights - the connection to small worlds networks, hubs in the hippocampus, modifying the synaptome, distributed representation, and more.

Web: [Direct Link] [This Post]