Google's AI Disaster Continues

the industry 24 // the latest on gemini erasing white people from history and other tech links
Mike Solana

FOLKS, a brief announcement: given the increasingly hostile link blockade, it is now close to impossible to share our work on Twitter, which means you guys are increasingly not seeing most of our work at Pirate Wires. In an effort to better connect with you all, we’re going to start spending some time on Instagram. Be sure to follow us there, and definitely subscribe to us on YouTube if you haven’t already. Thank you and godspeed.

Subscribe to The Industry

..
GEMINI'S WHITE BOY THANOS SNAP, FURTHER DEVELOPMENTS

Long story short, Google’s generative AI tool erased white people from human history, and the results were absolutely amazing (did you know that George Washington was a black woman? I didn’t either, but happy for my (re)education). It was a little bit alarming too, however, and not only for a world increasingly anxious at the thought of dominant artificial intelligence, but for what is meant to be the industry’s most important, and powerful technology company. Google is currently betting its future on winning the AI arms race, and allowing your DEI bureaucracy to mortify all of your top engineers? Not a good way to win. Yesterday, I broke the entire first chapter of Gemini’s catastrophic failure down for the Industry, which is where I recommend starting:

Read it here.

There have been a couple important developments since we published. First, the team responsible (sort of / kind of) apologized for releasing a tool so incapable of depicting white humans it was forced to erase and replace caucasians through every chapter of western history. Then, this morning, following a series of presumably failed tweaks to the prompt interface, Google paused Gemini’s image generation altogether. But is an apology enough? I’m not sure. The revelation of Google’s priorities are not a small thing. Nate Silver thoughtfully summed up the core issue here:

Of course, while everyone in media seemed to agree Google’s historical revision was, in some way, alarming, not everyone agreed on specifics. Over at The Verge, a company last seen endorsing the firebombing of a self-driving car in San Francisco on grounds this sort of thing was a “time honored piece of the human condition” (scroll down for the bit on Waymo), a fascinating counter to my own overall sense of this story was proposed. Building a machine literally incapable of sharing accurate information about white people, or even depicting them? This is, primarily, dangerous for black and brown people. Another incredible take from what is increasingly my favorite unhinged press outlet. Keep at it, guys, you’re doing the lord’s work.

In terms of what the hell is going on at Google, it’s hard to tell for sure. But I’m dying to get a better sense of the company’s culture, so if you work over there please respond to this letter, or drop me a note directly: solana@piratewires.com.

.
Industry Links

  • “Disinformation experts” rejoice, Meta is working with fact-checking organizations to “address false information” on their new Threads platform, with plans to soon unveil a full feature combatting disinformation akin to X’s Community Notes. Democracy: saved. (TechCrunch)
  • According to Elon, the first human patient with a Neuralink implant can now control a computer mouse simply by thinking. (Forbes)
  • Moon Boss is also bringing Starlink to 1) Israel, and 2) a field hospital in southern Gaza run by the UAE. This follows involved parties agreeing to measures that would prevent Hamas from getting access to the satellite internet services (as long as they promised I guess!). (Fortune)
  • I’ve got to be honest, I totally thought the Washington Post was writing a hit piece, but their recent profile on defense / hard tech truly wasn’t terrible. Worth a skim: “How Silicon Valley learned to love America, drones, and glory.” (And read more about The Gundo in Scott Nolan’s guest piece for Pirate Wires here).

The antidote.

...

From the Pod:

We’re a little late to the Industry newsletter this week on account of I’ve been tackling this Google story, which means our call out for this week’s pod comes like half a day before the next one goes live. BUT! In case you missed it: Alicia Keys’ sour note at the Super Bowl erased from the internet, the Mandela Effect (a little bit at least, I think we’ve gotta dive in deeper next time), Tucker’s Putin interview, Jon Stewart’s rising influence (genuinely and unironically shocking), is Kristen Stewart’s androgyny, technically speaking, kind of hot?, and Rachel Dolezal’s back in the news. Watch the latest episode below:

---

$$$:

  • Reddit has reportedly signed a $60 million licensing deal with an undisclosed AI company to train its model on Reddit’s content. (Bloomberg)
  • OpenAI has completed a deal with Thrive Capital, raising the artificial intelligence company’s valuation to over $80 billion — nearly tripling its valuation in less than 10 months. (NYT)
  • Coinciding with Bitcoin’s recent spike, Coinbase turned a net profit of $273 million in Q4 last year, compared with a $557 million loss the year prior. (The Information)
  • Walmart is acquiring Vizio for $2.3 billion, giving the retail giant new ways to reach customers through in-home media and more customer data when targeting ads. (The Information)
  • New York-based GlobalFoundries (semiconductor manufacturing) was awarded a $1.5 billion grant from the Biden Administration. This is the third deal out of 2022’s bipartisan Chips bill (NYT). I wrote a whole breakdown of the bill last year (check out: American Hustle: Microchip Edition)
  • Hadrian, Chris Power’s industrial automation startup, just announced a massive $117 million Series B. Check out John Coogan’s deep dive documentary into the company here.
  • Then, in case you’ve been living under a rock: Nvidia’s Q4 revenue spiked 22% from last quarter, up 265% from a year ago, totalling an absolutely insane $22.1 billion.

AI:

  • Air Canada was forced to honor a refund policy that was… completely made up by the airline’s AI chatbot. (Ars Technica)
  • Researchers at Amazon AGI have trained the largest text-to-speech AI model yet using 100,000 hours of public domain speech, 90% of which is in English and the rest in German, Dutch, and Spanish. (Tech Crunch)
  • Sam Altman is cautioning against “societal misalignments” around artificial intelligence, with the OpenAI CEO reiterating his calls for an international body to oversee AI (UGGGGGhhhhhhh please stop). (Axios)
  • Autonomous planes? Yes. Startup Xwing’s self-flying plane transported cargo for the U.S. Air Force as part of a military exercise, logging over 2,800 autonomous miles. (Axios)
  • At UPenn, students can now major in AI with courses in algorithms, programming, and advanced robotics, making the school the first in the nation to offer a degree in AI engineering. (Business Insider)
  • Zuck is here to dispel AI doomerism. The Meta CEO believes recent tech layoffs have less to do with the rise of artificial intelligence, and more to do with companies that “overbuilt” during the COVID tech boom. (SF Chronicle)

...

Check out more from the Pirate Wires extended universe

Conspiracy Theory: taylor’s demon bowl, the coordinates of power, torched waymo brings all the commies to the yard, “tech bro” press, and rise of the new elite. Read it here.

A Techno-Industrialist Manifesto. we need to make manufacturing better, cheaper, and faster through technology. it should be as easy to make physical things as it is to make software. the end. Read it here.

The Court Case Against a Massive Encampment that Won. how private citizens in a major US city sued their local government over a thousand-person tent city — and forced officials to clear it. Read it here.


...

Litigation and regulation:

  • Several city agencies in NYC have filed suit against Meta, TikTok, Snap, and Google accusing them of “fueling the nationwide youth mental health crisis.” (The Verge)
  • Elsewhere in the state, Governor Kathy Hochul — an expert on fairness and impartiality under the law (see public execution of Trump Corp.) — wants new legislation to criminalize deceptive AI. (Axios)
  • Not all bad news for Trump this week, however: the SEC approved a merger with his Truth Social and Digital World Acquisition (Trump will hold more than 78 million shares in the post-acquisition company — good for nearly $4 billion). (Washington Post)
  • Amazon joined SpaceX in calling the National Labor Relations Board unconstitutional. The company said in a legal filing that the NLRB’s structure, which includes limits on a president’s ability to remove judges that decide cases, violates constitutionally required separation of powers. (The Information)
  • The Justice Department is planning to review the proposed recent sports streaming merger by Disney, Fox, and Warner Brothers on antitrust grounds. Regulators will look at terms of the deal when finalized, though no action is certain yet. (Bloomberg)
  • Elsewhere, Comcast and Paramount have also been in talks about a partnership to bundle their respective streaming services. Congratulations, everyone, we’re reinventing cable.
  • A California court partially dismissed a copyright lawsuit against OpenAI brought by several authors, including comedian Sarah Silverman. (The Verge)
  • The Senate is gearing up to pass the “Kids Online Safety Act,” a bill first introduced in 2022 that, if signed into law, would amount to one of Washington’s biggest tech regulation bills in decades. Its future in the House remains unclear. Among the measures it would force tech companies to employ:
  • “Exercise reasonable care” (?) when designing features to avoid causing or exacerbating problems like depression and bullying
  • Limit features like infinite scrolling in an attempt to limit younger users’ time online
  • Allow parents and educators easily report harm to companies

Human resources:

  • Andrej Karpathy is leaving OpenAI, saying “it’s not a result of any particular event, issue or drama” but do keep the conspiracies coming nonetheless “as they are highly entertaining.” (@karpathy)
  • Firefox maker Mozilla is cutting about 60 jobs in a shake up after naming a new CEO. (Bloomberg)
  • Instacart is cutting 250 jobs and shaking up its leadership team as the company pivots to place a greater priority on higher-return projects like advertising. (Bloomberg)
  • Finally, with longtime board member Sheryl Sandberg set to depart, Meta is adding two new directors to its board: Broadcom CEO Hock E. Tan, and Arnold Ventures co-founder John Arnold. (Bloomberg)

Trade war:

  • Another turn in the EU v. Apple v. Spotify saga. To recap: after Apple, in an effort to skirt EU antitrust regulations, overhauled its App Store, many companies like Spotify and Epic Games weren’t satisfied. They hit back at the new fees Apple ratcheted up in tandem with the move, saying the changes were tantamount to “extortion,” and now, it looks like the EU is taking their side. The European Union is reportedly set to hit Apple with its first ever antitrust fine totalling around €500 million ($539 million). This one is difficult for me, because on one hand “fuck Apple,” I do get it. But also fuck Europe? I want to see more trade retaliation for this kind of thing in general. Like how, after what they did to our sweet beautiful prince Dylan Field, is fancy Irish butter not being taxed at 500%? I don’t know, food for thought (but buy it from Wisconsin).

...

This newsletter was compiled with a great deal of assistance from Riley Nork.

0 free articles left

Please sign-in to comment