Davey Alba
@daveyalba
Followers
31K
Following
19K
Media
304
Statuses
2K
@technology reporter covering Google and big tech / https://t.co/jHd4a2emnj / NO LONGER HERE, FIND ME ON BLUESKY https://t.co/PZSYGPbxkg
Joined April 2008
I have a dummy account that follows one Twitter account -- @taylorswift13. Unprompted, I started getting notifications from that account every time Andrew Tate tweets. I don't understand what's going on with Twitter's algorithm, to be honest.
28
180
3K
NEW: Spent the day immersed in viral falsehoods about the Israel-Hamas conflict that spread over the weekend and into Monday. Here's what we're bringing back from the misinformation trenches. (gift link). w/ @dzuidijk.
73
806
2K
don't know what to say. vile accusations without basis have consequences.
Twitter's former head of trust and safety has fled his home due to an escalation in threats resulting from Elon Musk's campaign of criticism against him, a person familiar with the matter told CNN.
9
176
2K
Summing the reactions of ex-Twitter workers to "The Twitter Files Pt 2" . Ex-employees seemingly saying three main things: 1) This has long been public; 2) The platform features called out are innocuous; 3) The new T&S head handed everything over, making this little more than PR
15
219
710
NEW: We analyzed hundreds of posts with misinformation about the Israel-Gaza conflict carrying Community Notes. We found that X has transformed from a flawed but policy-driven system to a free-for-all where falsehoods are prominently served up to users.
35
278
573
you guys. mastodon is FINE, or pretty good even. some starter tips (most of these c/o @mmasnick):. 1. is the easiest tool so far to find and follow all your friends from Twitter who are on Mastodon. 2. lets you pin the main timeline.
11
134
488
NEW: We documented an experiment run by @accountabletech that shows, tangibly, that Elon Musk's popular Twitter polls can be gamed by bot-for-hire manipulation services originating in Russia, India and Turkey
28
204
450
NEW: Ella Irwin, Twitter’s current head of trust and safety, has become the chief executor of Musk’s whims—even when it goes against established protocols for social media content work that Twitter and its peers have refined for the past decade.
19
208
482
im exhausted by this meticulous account of Elon's Twitter by @KurtWagner8 @sarahfrier @BradStone, charted by the day. but—excuse my bias—it's one of the smartest & most engaging versions of the story i've followed so closely from the start that's out there
16
99
470
Google asked ~80k employees to "dogfood" and test generative AI tools like Bard. Some employees concluded: Bard was "a pathological liar," "cringe-worthy" and gave dangerous advice, according to internal messages we reviewed. Google launched Bard anyway.
11
196
486
According to researcher @rothschildmd, Twitter is failing spectacularly in this moment. “It's now almost impossible to tell what's a fact, what's a rumor, what's a conspiracy theory, and what's trolling."
21
168
417
First (!) @BW cover: Investigators are increasingly using warrants to obtain location and search data from Google, even for nonviolent cases—and can end up looking through the user data of people who had nothing to do with a crime. With @byJuliaLove.
22
223
473
Our reporting found that posts abt the attack in Israel have led to confusion, misinformation and conflict on Musk's X (yes, this website), exposing how his policy changes have transformed the platform into an unreliable resource during a time of crisis.
6
114
346
In this post, which Community Notes also fact-checked, the author claims that the video is of Palestinian militants going door to door & killing Israelis—12.7M views on X. (It's actually of Israeli police.). Others who posted this misinformation deleted, but not this author
5
80
333
NEW: Twitter has made deeper cuts into its already radically diminished Trust and Safety team handling global content moderation + its Site Policy team that handles hate speech and harassment. w/ @KurtWagner8 .
24
208
298
New from us late last night: Twitter Spaces is down. To be clear, the fact that folks with suspended accounts could still be on Spaces IS a bug. But cutting off the feature as the Space was still happening—in a way that cut off debate—was also a choice.
7
75
274
I think I love this dog more than the fact that it’s Friday, as you can see in my face
2
13
219
NEW: We analyzed more than 53,000 posts over more than a decade sent by Elon Musk on X, and found that immigration and voter fraud has become — by far — his favorite and most popular policy topic online.
19
115
249
Here's Twitter's former head of product Kayvon Beykpour, called out in Weiss' thread, responding to her assertions. "You are characterizing any de-amplification as equating to shadow banning which is either a lazy interpretation or deliberately misleading."
5
52
216
These Twitter ex-employees are troubled by this. "The real story here is third parties getting uncontrolled access to tooling which reveals private information of Twitter users. This is bad, and seems to be escalating."
2
74
209
We don't include any links to the posts themselves that contain misinformation in our story, in order to guard against the content spreading much farther. But for the purposes of this thread, I'll share screenshots (with watermarks) of some of the viral falsehoods.
1
25
208
NEW: Internal documents shared with @technology show the complex instructions that Google contractors are asked to complete in minutes in order to train Google’s generative AI products, like Bard. AI experts say the workers have an "impossible job.".
4
98
211
New—Google will soon require that all election advertisers disclose when their messages have been altered or created by AI tools. Afaik, that's the first company from big tech requiring such disclosures on generative AI content in political ads.
9
84
212
New: Wrote about the AI photo that sent stocks down on Monday AM. The first image appears to have been posted by a QAnon account on Facebook, which saw little engagement. Then it was picked up by Twitter Blue verified accounts and RT, and exploded online:
9
114
206
NEW: Wrote about a semi-secret Discord that Google runs for heavy users of Bard. In it, Google PMs, designers & engineers tasked w/ developing Bard say they don't know what LLMs are truly useful for, don't trust LLM output before verification, and more.
6
87
218
That was on Sat AM. A few hours later, this verified anon account shared a manipulated image that purported to prove the US was sending $8B worth of military aid to Israel. But the dateline and details in the screenshot were manipulated, per @Info_Rosalie. 402.7K views on X.
1
29
193
@MichaelKugelman And another verified X account spread the falsehood that Ukraine sold weapons to Hamas—7M views. The author quickly followed up with a post saying, "For the record, we don't know if this is true or not.". That's a tactic for misinformers to "escape culpability," per @jbruce218
1
43
176
NEW: For weeks, @byJuliaLove and I talked to current and ex-Googlers to understand the trade-offs between speed of deploying generative AI products—which Google has been doing under code red—and carefully considering the ethical consequences of such tech.
7
76
175
Here is Melissa Ingle, a former Twitter contractor working on civic integrity and political misinformation, making the same point -- a lot of this was already public, not a secret, published in blog posts, etc.
2
29
151
on the precariousness of journalism, aging bodies, and our broken health system. this is harrowing, from a writer and editor i've long admired
3
45
176
@Info_Rosalie @NewsGuardRating This post, which is meant to look like it's from the Taliban, claims that the group was asking Iran, Iraq & Jordan for passage to join up with Hamas. 2.5M views on X; no community note. It was also picked up by The Gateway Pundit, and reached as many as 440K people on Facebook.
4
25
145
NEW: Maybe you think of deepfake pornography as something that happens on the internet's edges. Not so. It's facilitated by Google, Microsoft, the app stores, hosting services and payment processors—all on the open web. @cecianasta and I document how:
6
72
158
We spoke to more than a dozen current & former employees for this story. They told us that since Musk took the helm of Twitter, Irwin has helped him break conventions in how Twitter manages user account policy. Other experts agree, like @LauraEdelson2.
4
68
156
my favorite few paragraphs. "Longtime colleagues say he struggles with self-reflection and has an inability to take constructive feedback or tolerate criticism.".
2
30
148
@Info_Rosalie The above is the most popular X post pushing the falsehood. According to @NewsGuardRating, it's one of 1400 similar posts on X—604k collective views. The false claim also spread on Tiktok, collecting 17,600 views. It also spread on Telegram channels and QAnon forum posts.
1
22
140
But I spoke to @MichaelKugelman, who has studied the Taliban since 2007. He said there's no reason to believe this claim. The Taliban have always focused on Afghanistan alone. Previous posts by the account also criticized Qatar, which the group wouldn't do.
2
15
129
If this is about “real-time posting,” why does Elonjet violate this brand-new Twitter policy? (The flight data is delayed) . And as Sweeney notes in this NBC story, the last time his bot tweeted anything was Dec 12—a whole day before Musk alleges this “stalking” incident occurred
6
31
129
NEW with @LeonYin @Leonardonclt: Ever since the advent of ChatGPT in late 2022, there’s been so much excitement for generative AI across industries. We wanted to ask—is there a risk of racial bias in generative AI being used for hiring and recruitment?.
11
66
138
Twitter used to have a group called the Global Escalations Team, acting as a check on power at the top of the company. It could even overrule execs. But the group has folded; employees told us Irwin & Musk can no longer be challenged thru formal process .
1
52
125
Meanwhile, @Imi_Ahmed said that X's statements—saying Community Notes will help, and that "while difficult, it's in the public's interest to understand what's happening in real time"—pushes the burden for a solution onto its users
3
24
114
Lots of outlets are carrying the news of these Facebook studies today. Here's my and @aishacounts' take. We focus on one study that we thought was most revealing: . Facebook Misinformation in US Election Reached More Conservatives, Study Says
4
52
120
@MichaelKugelman .@MichaelKugelman also told me that at least one account that's been around a long time and is known to be associated with the Taliban has rejected the allegation
1
12
113
@Imi_Ahmed And the risk of the viral falsehoods isn't just giving people an inaccurate picture of the conflict, @Imi_Ahmed added. "Lies underpin the hate. some of these images and videos are designed to invoke the most extreme reactions possible.”
7
28
106
There has been a lot of rhetoric about “independent journalism” lately and what it entails. What is independent journalism if not backing your reporters’ work, even when it makes powerful people angry? What is journalism FOR?.
Axel Springer said it's conducting a review of the "motivation and the process leading up to the reporting" around Business Insider stories alleging plagiarism by Bill Ackman's wife Neri Oxman, following backlash from Ackman and others.
3
35
110
New/breaking from me and @byjulialove: Google is laying off several hundreds of people, with folks finding out right now. Workers affected include Google's digital assistant, hardware and engineering teams. More here:
9
71
118
Scoop with @josheidelson: Contract staffers who worked on Google's AI announced their intention to unionize in June. Less than a month later, over 70% of the bargaining unit was fired. Now @AlphabetWorkers has filed an unfair labor practice with the NLRB.
7
63
110
Being the head of trust & safety at Twitter has long been a closely scrutinized job. Twitter has had to defend its content moderation moves publicly—even to Congress. Now, Twitter makes decisions on the unilateral directives issued by its new owner.
5
35
100
New: Here come the websites populated by ChatGPT. and potentially Bard.
9
71
106
🚨 NEW: Google Search has ruled the web for >20 yrs. But recently, a chorus of discontent has been building among the web-going public. Google is drastically remaking search with gen AI just as people are complaining that the service is slowly degrading—and degrading the web.
4
48
114
"After five years of pioneering research into the abuse of social platforms, the Stanford Internet Observatory is winding down," @CaseyNewton & @ZoeSchiffer report. :/ .
10
52
107
NEW: Wrote about Gemini generating ahistorical images of people and the right-wing backlash it sparked. The company tried a Band-Aid fix to reduce bias in its outputs, but Google didn’t anticipate the ways the tool could misfire & wasn’t fully transparent
127
31
108
NEW: It is trivially easy to get Bard, Google's answer to ChatGPT, to generate blog posts, op-eds, letters, paragraphs and more promoting known misinformation narratives, according to new research from @NewsGuardRating shared exclusively with us.
6
42
99
Really important detail that I hope doesn’t get lost in the story: Dozens of unpublished—but completed—reports detailing information operations on the platform will likely never become public, said four ex-employees who worked on the reports. Irwin said she is unaware of them.
1
26
99
It's an honor to be recognized for our reporting on bias in generative AI systems, which have already made their way into countless industries around the world. Grateful to be working with such a dream team at @technology ❣️.
WINNER! Excellence in Technology Reporting, Large Newsroom: Uncovering bias in two leading AI systems by @business #OJA24
7
18
104
New from me and @rachelmetz: LAION-5B, the world’s largest openly accessible image-text dataset, includes at least 1,008 instances of child sexual abuse material and could include thousands more, per a new report from @stanfordio .
12
64
96
Eric Frohnhoefer, who politely disagreed with Musk then was fired in a tweet, also says Bari's description of VF is "not true". "Lets say your account was suspended, or Tweet was deleted, or a reply from a protected account you don't follow. Twitter would 'filter' it."
2
13
85
"The tech industry, w/ Google right at the front, has again put themselves in a terrible bind of their own making," @LauraEdelson2 told us. “The industry desperately needs to portray AI as magic and not stochastic parrots. but parrots are what they have".
2
33
100
Eye-opening story about the candy sellers in New York, a common sight on city subways. Deeply reported and well worth the read—I hope the story gets the attention it deserves
1
52
96
reporting using large randomized survey data > reporting with preconceived story framing
0
23
79
With less than $100, one can buy tens of thousands of votes for Twitter polls—and two Twitter ex-employees in Trust & Safety tell me the company has lacked proper safeguards to identify and eradicate inauthentic behavior and manipulation of Twitter polls.
2
23
83
We finally wrote about the trial exhibits being held back in #USvGoogle. Also, we published the documents that disappeared from the DOJ’s website. Read them all here. Story w/ the inimitable @leah_nylen .
2
54
86
New: We wrote about the leaked AI document making the rounds in Silicon Valley today. We confirmed that a senior software engineer working at Google has argued: “We have no secret sauce”. w/ @byJuliaLove @rachelmetz
2
21
88
In one listserv, an employee pleaded w/ Google: "bard is worse than useless: please do not launch." It was seen by nearly 7000 workers; many employees agreed that Bard's answers were contradictory or could even go off the rails on simple, factual queries.
2
20
83
Here's my very last day at the @BuzzFeedNews office, the place where I did some of my best accountability work, including reporting on sexual harassment and on the erosion of democracy in the Philippines because of Facebook. My heart goes out to all my ex-colleagues—hire them.
0
6
79
New from me: ChatGPT is awe-inspiring tech. It wrote a wonderful poem about my cats! But it's plagued by the same issues of bias as AI chatbots that came before it. (Since I assume y'all aren't reading @nytimes today ✊, feel free to read this instead).
2
25
82
A small scoop from me and @shiringhaffary: Google is trimming jobs in its trust and safety group, just as the company is also asking those employees to be on standby over the weekend in order to troubleshoot problematic Gemini outputs
6
48
83
Just the most amazing sentence in @CaseyNewton’s newsletter today re Twitter. I literally lol’ed
0
9
70
!! Researchers disagree with Meta's framing of the papers published in Nature and Science today via @JeffHorwitz
2
46
65
New: The Alphabet Workers Union said that thousands of Google "raters"—contract workers assessing the quality of Google search and ads—won a $15/hr pay hike following a Bloomberg report documenting allegedly inadequate pay and benefits.
1
28
70
Our story on the disinformation swirling on social media tonight after shots were fired at the Trump campaign rally Saturday.
18
32
75
The cuts happened last night. At least a dozen more people were slashed from Twitter Trust & Safety, including the relatively new head of its APAC region + workers in misinfo policy, global appeals and state media.
5
45
65
News: After being placed on a brief leave, 28 Googlers have been fired for being involved in Project Nimbus protests. Google says the employees physically impeded other employees' work; the workers dispute this, saying some of them only rallied outside.
6
35
73
Bullet points from internal Google presentation about antitrust.-Avoid references to “markets,” or “market shares” or “dominance”.-Avoid discussions of “scale” and “network effects”.-We do not “leverage” anything.-We don’t “lock up” or “lock in” our users/partners. (1/2).
5
18
70
Another one from Lara Cohen: "It's not 'reporting' when you're handed a document all-you-can-eat buffet by the new billionaire owner. You're doing his bidding. It's a weirdo PR campaign."
1
13
53
New: Three months after @byJuliaLove and I published our investigation into reverse warrants served to Google by police, the company says it is ending its ability to respond to geofence warrants.
6
36
69
Accountable Tech director @nicolelgill says the experiment shows "Twitter is unsafe and open to manipulation by bad actors—including foreign govts." Manipulation to the tune of 500K-1M votes would cost about $2,600 to $3,600 to throw, researchers found.
2
5
59