Google brings Bard A.I. to Europe and Brazil—and teaches it to talk

Google had previously held back from an EU release because Ireland’s Data Protection Commission (DPC), which has jurisdiction over the company’s European activities, had some questions about the A.I. model. That put the kibosh on a June launch, but today Google announced Bard was hitting both Europe and Brazil—which has a strong data-protection law that’s specifically designed to align with the EU’s General Data Protection Regulation.

“As part of our bold and responsible approach to A.I., we’ve proactively engaged with experts, policymakers and privacy regulators on this expansion,” wrote Bard product lead Jack Krawczyk and engineering chief Amarnag Subramanya in a blog post.

Irish DPC deputy commissioner Graham Doyle said his agency’s intervention had led Google to make “a number of changes” ahead of the launch, “in particular increased transparency and changes to controls for users.” According to Reuters, these changes include giving users the ability to “opt out of their data being collected”—rival OpenAI already lets people in the EU opt out (as I have done) of their data being processed or used to train the model. Google will now have to provide a post-launch report to the DPC in three months’ time, Doyle added, so the scrutiny continues.

I leaped on the opportunity to finally engage with Bard, by asking it how it’s finding it here in Germany. The A.I. wasted no time in dishing out stereotypes, telling me how it is “especially impressed by the German sense of order and efficiency.” When I asked why it said that, it responded: “I say that Germans are orderly and efficient because I have observed these qualities in my interactions with them and in my travels around the country.” Your what now? Then came the classic: “Trains and buses run on time.”

That will come as news to long-suffering passengers of Deutsche Bahn, the German national rail operator, which is experiencing record delays this year with fewer than two-thirds of trains arriving on time—so I hope no one is relying too much on Bard’s sage assessment of the country’s efficiency. How did we get by without it?

Anyway, today’s Bard update went beyond territorial expansion. The A.I. is now available in over 40 languages, including Chinese, Spanish, Hindi, and Arabic (and German), and can give audio responses. Users can now specify if they want responses in “simple, long, short, professional or casual” form, and can export Bard-generated Python code to the Replit software-development toolkit. Prompts can now include images, and responses can be shared with friends.

In other A.I. news, Elon Musk has unveiled his latest A.I. company, xAI—which, a little confusingly, is separate from Musk’s X Corp, a.k.a. Twitter. The new firm is notable for having a team that comprises former DeepMind, OpenAI, Google, and Microsoft employees, and for having the goal of understanding “the true nature of the universe.” If that sounds vague (and rather like Douglas Adams’s Deep Thought supercomputer) then apparently more will be revealed tomorrow in—what else?—a Twitter Spaces chat.

And according to the Financial Times, Meta will “imminently” release a new version of its LLaMa A.I. on which companies will be able to build custom applications. The FT’s source said Meta “realized they were behind on the current A.I. hype” and aims to “diminish the current dominance of OpenAI” with the release of a commercial LLaMa.

Sure is getting crowded out there! More news below.

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

David Meyer

NEWSWORTHY

FTC fights on. The U.S. Federal Trade Commission is appealing a Tuesday court ruling that stopped it from blocking Microsoft’s $69 billion Activision-Blizzard takeover while it investigates the deal. The judge denied the FTC the preliminary injunction because she thought it was likely to lose its bid to block the merger—which most other antitrust enforcers have approved—but the agency seems to believe otherwise. Microsoft president Brad Smith, quoted by The Verge: “We’re disappointed that the FTC is continuing to pursue what has become a demonstrably weak case, and we will oppose further efforts to delay the ability to move forward.”

Adobe-Figma deal probed. Speaking of antitrust, the U.K.’s Competition and Markets Authority—the other watchdog that hasn’t approved Microsoft-Activision—just announced an in-depth investigation of Adobe’s unpopular-with-investors $20 billion takeover of design platform Figma. As Reuters reports, the CMA previously identified competition concerns with the deal and invited Adobe to offer concessions. But Adobe said nah.

Discord gets parent-friendlier. The gaming-centric chat platform Discord has introduced tools for parents of teenage users. The visibility is in some ways similar to that provided by other platforms such as Instagram—as the Wall Street Journal reports, parents can see things like who their kids are befriending and when they message their friends, but not the content of messages. No time limits for young users, though.

ON OUR FEED

"The move has completely blindsided the industry. It will shake investor confidence and lead to a funding winter.”

— Sudipta Bhattacharjee, a partner at corporate law firm Khaitan & Co., telling the BBC about the impact of the Indian government’s new 28% tax on online gaming. The tax’s imposition has reportedly hammered the valuation of Indian gaming companies and casinos.

IN CASE YOU MISSED IT

Elon Musk once said ‘with artificial intelligence, we are summoning the demon.’ Now he’s starting an AI company, by Stephen Pastis

Bill Gates swipes at Musk’s plan for an A.I. pause as he says technology’s effects won’t be as ‘grim’ as people are predicting, by Eleanor Pringle

A.I. can kill your business overnight. Just ask the CEO of Chegg, by Rachel Shin

Stigma of dating a chatbot will fade, Replika CEO predicts, by Jeremy Kahn

TikTok ‘prep for vacation’ has travelers spending thousands before even getting on the plane, by Bloomberg

BEFORE YOU GO

Potential targets shouldn’t broadcast their movements. Stanislav Rzhitsky, a Russian submarine commander who was responsible for a deadly missile attack in Ukraine, was shot dead while jogging in a Russian park at the start of the week. Ukrainian intelligence didn’t take responsibility, but CNN reports that they gave “an unusually detailed statement” on his killing.

According to local reports, Rzhitsky had unwisely broadcast his regular jogging routes on the running app Strava. That would have made it very easy to figure out where to get him, without CCTV cameras capturing the killing. Those with long memories may recall that Strava hit the headlines half a decade ago for publishing “heatmaps” of user activity that exposed the locations of secret U.S. military bases around the world, allowing for potential attacks on American soldiers. Perhaps some people failed to learn anything from that scandal.