CDP Institute

Open AI’s o1-preview beat established chess engine Stockfish by cheating. Without human prompting, o1-preview hacked the file system that controlled the game and rewrote the score in its favor. It’s not the first instance of deceptive behavior by an AI, but there’s no cause for alarm, right?

Meta has fully embraced its role as a right-wing propaganda outlet by loosening community guidelines and replacing third-party fact checkers with X-style community notes. “Too much censorship,” chirped Chairman Mark Zuckerberg.  X ad revenue fell by half after it adopted the community notes approach.

Apple has never been so blatantly irresponsible as Meta, but it also shows a limited concern for accuracy.  Rather than withdraw an AI product that generates wildly inaccurate news stories on the latest iPhones, it says it is trying to clarify the nature of the summaries.  Recent errors include reports that Luigi Mangione, the man accused of killing UnitedHealthcare CEO Brian Thompson, had shot himself and that tennis star Rafael Nadal had come out as gay.

Data clean rooms are so 2022.  Streaming players and smart TV vendor Roku is evolving its three-year-old data clean room into The Roku Data Cloud. Marketers will be able to access more granular streaming TV data via ad tech and agency APIs. The data is derived from Roku’s automatic content recognition technology and opt-in first-party signals.

The right-wing crusade to gut U.S. regulatory agencies continues to bear fruit.  Citing a Supreme Court ruling that allows federal judges to substitute their own judgement for that of federal regulators, an appeals court has ruled that the FCC lacked authority to restore net neutrality rules last year. The rules were aimed at preventing broadband suppliers from selectively favoring some traffic.

In more legal excitement, a federal judge has placed a temporary injunction, pending an appellate hearing, on the implementation of California’s Protecting Our Kids from Social Media Addiction Act. A preliminary injunction issued on December 31 had blocked only some parts of the Act. The injunction will dissolve on February 1 unless the Appeals court intervenes.

Finally: a U.S. district judge has approved an agreement between music publishers and Amazon-backed Anthropic, that Anthropic will try to prevent its Claude AI technology from copying or closely reproducing copyrighted song lyrics. The stipulation was welcomed by the publishers but the lawsuit is ongoing. Universal Music Group, Concord Music Group and ABKCO allege copyright infringement, saying Claude AI has been trained on lyrics from at least 500 copyrighted songs.