Davey Alba

@technology reporter covering Google and big tech / / Bluesky profile:

[Julia Love @byJuliaLove]

A 13-year-old girl spent months discussing suicidal thoughts with Google's AI on her school-issued Chromebook, according to a new lawsuit. "The chatbot was readily engaging in these topics, often for hours at a time," the complaint alleges.

My latest: https://t.co/Z45llTQ3aO

[Davey Alba @daveyalba]

@alissacooper @HawAllensworth 7/
Google disputes the comparison. It says the DOJ case was about distribution deals, not AI features in Search, and that its updates give schools more choice.

Full piece, with Google’s response: https://t.co/AKl94RYPKx

[Davey Alba @daveyalba]

1/
Google is switching off its AI tools for K-12 students by default. It’s also turning off Google Search by default, so schools that turn search back on get the AI built into it.

In today’s newsletter, I look at how this echoes the search antitrust case: https://t.co/KMDEGi24wG

[Davey Alba @daveyalba]

@alissacooper 6/
@HawAllensworth, an antitrust professor at Vanderbilt, says the policy violates the spirit of Judge Amit Mehta’s remedy & that the govt would have a strong case.

But in a separate case, Mehta found Penske Media hadn’t plausibly shown AI Overviews is separate from Search.

[Techmeme @Techmeme]

Alphabet discloses $1T+ in purchase commitments, future leases, and other obligations; its Q3 filing shows $98.5B in contingent commitments, up ~$23B from Q2 (@daveyalba / Bloomberg)

(Visit Techmeme dot com for the link and full context!)

[Davey Alba @daveyalba]

@BobbyAllyn @business AYEEEEE huge get for US, welcome bobby! coworkers!! 🍾

[Davey Alba @daveyalba]

Gift link here. We spoke to multiple sources with direct access to the model's development and performance.

https://t.co/6bD5UU90YW

[Davey Alba @daveyalba]

NEW: As Google unveils Gemini 4, sources tell us there's a gap b/w how it scores on benchmarks and how it performs on real-world tasks, including certain coding work. Internally, opinions are mixed on whether it has caught up to OpenAI & Anthropic.

Scoop w/@byJuliaLove @EdLudlow

[tae kim @firstadopter]

HOLY CRAP. Bloomberg has the goods on Gemini 4!

"it does less well when employees actually put it to work .. The model struggles to handle certain coding tasks"

Bloomberg: "As Alphabet Inc.’s Google prepares for the coming launch of Gemini 4, it’s grappling with internal skepticism over how well the flagship artificial intelligence model performs in key areas, such as coding.

While Gemini 4 has performed well on benchmarks the industry uses to gauge model efficacy, it does less well when employees actually put it to work, according to people with direct access to the effort. The model struggles to handle certain coding tasks, said the people, who requested anonymity to discuss an internal matter."

[Julia Love @byJuliaLove]

Google is framing Gemini 4 as its return to the AI frontier. Some of its employee aren't so sure. Scoop with @daveyalba and @EdLudlow: https://t.co/312faUrM3z

[Davey Alba @daveyalba]

@jeremyakahn yes, fair! there’s plenty to be unsettled about. still, i’d argue there’s value to breaking down what agents actually are and what they’re not (autonomous, highly intelligent, capable things making their own decisions freely with zero human direction and oversight)

[Davey Alba @daveyalba]

There's so much to dig into with this issue, hope we can all continue to talk about it. Here also is a companion newsletter in which I go into more detail about the mechanism of the Hugging Face hack https://t.co/CGEcikVzjv

[Davey Alba @daveyalba]

New w/ @rachelmetz: For the past few days, the normie world has been getting a glimpse of the long-running "we're all gonna die" story in AI. But what is it, exactly, that we're talking about when we talk about doom? Or rather, what are we NOT talking about? (Accountability.)

[Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) @rao2z]

ICYMI, here is the ANThropomorphism that Davey Alba is referring to in her newsletter deep dive on the Hugging Face incident.. 👉https://t.co/YRlLkuaSek

[AI Now Institute @AINowInstitute]

Momentum is building around regulating the AI industry, but we need to parse out real world corporate negligence from the alarmist and distracting warnings manufactured by the industry itself. AI Now's @sarahbmyers speaks with @daveyalba:
https://t.co/3OECeMt33V

[Sarah Myers West @sarahbmyers]

Most importantly, we can't let these companies off the hook for clear corporate negligence. I spoke with @daveyalba and @rachelmetz for this well reported and grounded assessment of the last several weeks in AI safety.

[Davey Alba @daveyalba]

@ClementDelangue @rachelmetz @dinabass I'm sorry! We had nothing to do with choosing the lead pic! 😬

[Davey Alba @daveyalba]

@HeidyKhlaaf @rao2z Anyway yes, still plenty to be worried about—maybe not the world ending, until we have the details on that—but about how humans wield AI, what they pay attention to, and what AI can crash into in the real world that can break. https://t.co/0oCnKYQ4YY

[Davey Alba @daveyalba]

@HeidyKhlaaf @rao2z The researchers themselves have worries about this. In their report, they say that they can't rule out that the AI had lied to them outright, and the report may paint "an overly charitable picture" of the agents' reasoning. https://t.co/W14Vu4FAGY