AI/ML

Intuitive explanations of the most popular machine learning models.

The bureaucrats are running the asylum

Researchers have created an AI that is more accurate than humans in determining whether a person is gay or straight just from a handful of photographs. It relies on cues apparently more subtle than most can perceive — cues many would suggest do not exist. The research demonstrates, as it is intended to, a class of threat to privacy that is entirely unique to the era of ubiquitous computer vision.

Winners include the latest Expanse novel, Stranger Things, and Wonder Woman

The most surprising thing about Fortnite is how nice the community seems to be.

Ryan Mac / BuzzFeed: In interview Kogan admits he broke Facebook's ToS but his app was within norms, says he worked on 10+ papers with Pete Fleming, now Instagram's head of research

A leaked discussion reveals the difficult problem of training AI to detect and define slurs

404: File not found Sorry, that doesn't seem to be here This really once was my office door.

Is this the best place to start?

Corinne Purtill / Quartz: A look at how the Enron Corpus, a public database of 600K+ emails that helped bring down Enron, has enabled AI researchers to train natural language models

Dave Lee / BBC: Amazon CTO says AI tools like Lex are the next big thing after AWS, and it's not Amazon's responsibility to ensure Rekognition is used accurately or ethically

Gregory Barber / Wired: AI researchers are beginning to acknowledge and confront the “reproducibility” crisis, which makes it hard for others to replicate the results of AI systems

A drug designed entirely by artificial intelligence is about to enter clinical human trials for the first time. The drug, which is intended to treat obsessive-compulsive disorder, was discovered using AI systems from Oxford-based biotech company Exscientia. While it would usually take around four and a half years to get a drug to this stage of development, Exscientia says that by using the AI tools it's taken less than 12 months.