Month: November 2022

Enlarge / Go pieces and a rulebook on a Go board. (credit: Getty Images)

In the world of deep-learning AI, the ancient board game Go looms large. Until 2016, the best human Go player could still defeat the strongest Go-playing AI. That changed with DeepMind's AlphaGo, which used deep-learning neural networks to teach itself the game at a level humans cannot match. More recently, KataGo has become popular as an open source Go-playing AI that can beat top-ranking human Go players.

Last week, a group of AI researchers published a paper outlining a method to defeat KataGo by using adversarial techniques that take advantage of KataGo's blind spots. By playing unexpected moves outside of KataGo's training set, a much weaker adversarial Go-playing program (that amateur humans can defeat) can trick KataGo into losing.

To wrap our minds around this achievement and its implications, we spoke to one of the paper's co-authors, Adam Gleave, a Ph.D. candidate at UC Berkeley. Gleave (along with co-authors Tony Wang, Nora Belrose, Tom Tseng, Joseph Miller, Michael D. Dennis, Yawen Duan, Viktor Pogrebniak, Sergey Levine, and Stuart Russell) developed what AI researchers call an "adversarial policy." In this case, the researchers' policy uses a mixture of a neural network and a tree-search method (called Monte-Carlo Tree Search) to find Go moves.

Enlarge / Go pieces and a rulebook on a Go board. (credit: Getty Images)

In the world of deep-learning AI, the ancient board game Go looms large. Until 2016, the best human Go player could still defeat the strongest Go-playing AI. That changed with DeepMind's AlphaGo, which used deep-learning neural networks to teach itself the game at a level humans cannot match. More recently, KataGo has become popular as an open source Go-playing AI that can beat top-ranking human Go players.

Last week, a group of AI researchers published a paper outlining a method to defeat KataGo by using adversarial techniques that take advantage of KataGo's blind spots. By playing unexpected moves outside of KataGo's training set, a much weaker adversarial Go-playing program (that amateur humans can defeat) can trick KataGo into losing.

To wrap our minds around this achievement and its implications, we spoke to one of the paper's co-authors, Adam Gleave, a Ph.D. candidate at UC Berkeley. Gleave (along with co-authors Tony Wang, Nora Belrose, Tom Tseng, Joseph Miller, Michael D. Dennis, Yawen Duan, Viktor Pogrebniak, Sergey Levine, and Stuart Russell) developed what AI researchers call an "adversarial policy." In this case, the researchers' policy uses a mixture of a neural network and a tree-search method (called Monte-Carlo Tree Search) to find Go moves.

We've been eagerly awaiting director Rian Johnson's follow-up to 2019's delightful Knives Out, and now we have the final trailer for Glass Onion: A Knives Out Mystery. Daniel Craig and his Southern drawl are back and in better form than ever as quirky detective Benoit Blanc, and he has an equally colorful all-star ensemble cast of suspects.

As we've reported previously, Knives Out premiered at the Toronto International Film Festival and was an instant hit, and word of mouth spread quickly. The film ended up grossing $311 million worldwide—the top-grossing film in 2019 not based on existing IP. It was my favorite film of 2019. In my review, I praised Johnson's deft ability to mix dark subject matter with a lighter touch in his old-school classic whodunnit: "Think Clue meets Murder on the Orient Express—or any number of adaptations of novels by the grande dame of murder mysteries, Agatha Christie—both of which the director cited as influences via a 2019 Twitter thread of movie posters."

So naturally there was going to be a sequel... or two. Lionsgate originally took on the project, but in 2021 Netflix won a bidding war for the rights to the film plus a second sequel, beating out both Amazon and Apple with a whopping $469 million bid. Johnson wrote the script for Glass Onion during the pandemic, inspired particularly by the Christie-based "tropical getaway" whodunnit Evil Under the Sun (1982), as well as a 1973 film called The Last of Sheila—an underappreciated gem.

We've been eagerly awaiting director Rian Johnson's follow-up to 2019's delightful Knives Out, and now we have the final trailer for Glass Onion: A Knives Out Mystery. Daniel Craig and his Southern drawl are back and in better form than ever as quirky detective Benoit Blanc, and he has an equally colorful all-star ensemble cast of suspects.

As we've reported previously, Knives Out premiered at the Toronto International Film Festival and was an instant hit, and word of mouth spread quickly. The film ended up grossing $311 million worldwide—the top-grossing film in 2019 not based on existing IP. It was my favorite film of 2019. In my review, I praised Johnson's deft ability to mix dark subject matter with a lighter touch in his old-school classic whodunnit: "Think Clue meets Murder on the Orient Express—or any number of adaptations of novels by the grande dame of murder mysteries, Agatha Christie—both of which the director cited as influences via a 2019 Twitter thread of movie posters."

So naturally there was going to be a sequel... or two. Lionsgate originally took on the project, but in 2021 Netflix won a bidding war for the rights to the film plus a second sequel, beating out both Amazon and Apple with a whopping $469 million bid. Johnson wrote the script for Glass Onion during the pandemic, inspired particularly by the Christie-based "tropical getaway" whodunnit Evil Under the Sun (1982), as well as a 1973 film called The Last of Sheila—an underappreciated gem.

Enlarge (credit: Aurich Lawson | Stable Diffusion)

Twitter owner and CEO Elon Musk yesterday announced a strict ban on unlabeled parody accounts, a step he's taking after numerous Twitter users impersonated Musk. "Going forward, any Twitter handles engaging in impersonation without clearly specifying 'parody' will be permanently suspended," Musk wrote in a tweet thread.

Musk's announcement came as a number of high-profile Twitter users changed their display name and profile picture to match Musk's. Comedian Sarah Silverman posted a screenshot indicating that her account was temporarily locked for violating Twitter rules after she tweeted using Musk's name and profile picture. Silverman's tweet impersonating Musk said, "I am a freedom of speech absolutist and I eat doody for breakfast every day."

Comedian Kathy Griffin's account was suspended under similar circumstances. As Silverman noted, these suspensions happened about a week after Musk wrote in a tweet that "comedy is now legal on Twitter."

Enlarge (credit: Aurich Lawson | Stable Diffusion)

Twitter owner and CEO Elon Musk yesterday announced a strict ban on unlabeled parody accounts, a step he's taking after numerous Twitter users impersonated Musk. "Going forward, any Twitter handles engaging in impersonation without clearly specifying 'parody' will be permanently suspended," Musk wrote in a tweet thread.

Musk's announcement came as a number of high-profile Twitter users changed their display name and profile picture to match Musk's. Comedian Sarah Silverman posted a screenshot indicating that her account was temporarily locked for violating Twitter rules after she tweeted using Musk's name and profile picture. Silverman's tweet impersonating Musk said, "I am a freedom of speech absolutist and I eat doody for breakfast every day."

Comedian Kathy Griffin's account was suspended under similar circumstances. As Silverman noted, these suspensions happened about a week after Musk wrote in a tweet that "comedy is now legal on Twitter."