Oh, joy.

Google’s I/O conference is next week, and we expect to hear a lot about the company’s AI endeavors. The company says there’s so much to talk about that it’s spilling the Android beans a little early, and yes, a lot of AI is involved. In the coming months, Google will roll out more smartphone AI features under the Gemini Intelligence banner, bringing more automation and customization to your phone.

App automation will be a major element of Android going forward, Google says. Automation for apps is expanding after Google began testing it earlier in 2026 with DoorDash and Uber on Pixel and Samsung phones. It was a very frustrating experience at launch, but Google says it has spent the intervening months fine-tuning the system.

Google promises that Android will be able to handle more complex automations across apps. For example, the robot could find a course syllabus in Gmail and then hop to a shopping app to add the necessary books to your shopping cart. Google also suggests taking a picture of a travel brochure and telling Gemini to book something similar in the Expedia app.

I've yet to find a reason I'd want my apps talking to each other, let alone unsupervised.

But the more basic issue is that agentic "AI" isn't fit for purpose.

So, I have a follow-up appointment for how my dentures are fitting. Would it be cool to just say "appointment at 1 p.m. at this location in two weeks" and have it arrange the rideshares so I don't need to remember anything?

Sure, it would. Kinda creepy, but cool nonetheless. A couple of issues here:

• I generally use Lyft, and outside of trips to the airport, I've learned that booking in advance is always far more expensive than the spot price (YMMV). Atop this, I usually get a beg notification offer by looking into pricing about 15 minutes before my planned need to book. Put it all together, and I can end up paying fully 75% more than I needed to. That's the last thing I want Gemini automating for me.
• It's a fucking medical appointment. Divining when I'll need to be picked up is a fool's errand, no matter what I blocked out for it in my calendar. Unless you plan on closing down a bar, this applies similarly.

So, I'll get the guaranteed highest price with no flexibility for reality, and this is an improvement?

And you want me to use this to book larger travel plans? Highest airfare, highest hotel rate? Reservations at a Michelin restaurant because we happen to have traveled there for our anniversary, but between the outrageous airfare and usurious hotel rate, ain't no one got the money for a $120 steak.

For years, Emrah Bayraktar did just about anything he could to make money: Cleaned cars. Worked a night-shift in a warehouse. Made sandwiches at a Subway during the day.

When he wasn't eking out a living with part-time work, 25-year-old Bayraktar from Antwerp, Belgium would take out his iPhone and edit long interviews of influencers into snippets and post them to Instagram.
Pop star-turned-Arizona State University professor will.i.am teaches “The Agentic Self” at his headquarters in Los Angeles, Calif., on Jan. 14, 2026.

"And then one random night, I saw a notification saying that I earned $12, I was like OK cool," Bayraktar said. "Then two weeks later, I made two-and-a-half thousand dollars, and I thought, 'Maybe I could just quit my jobs and go all-in on this.'"

And he did, earning a cut every time someone bought something from a link he placed in the video clip.

He became so skilled at editing the short videos that he now runs a network of 40,000 freelance clippers and has a YouTube channel where he teaches people how to become clippers, directing them to websites where, instead of getting paid for so-called affiliate link purchases, clippers are paid per view they generate.

Amazon employees are using an internal AI tool to automate non-essential tasks in a bid to show managers they are using the technology more frequently.

The Seattle-based group has started to widely deploy its in-house “MeshClaw” product in recent weeks, allowing employees to create AI agents that can connect to workplace software and carry out tasks on a user’s behalf, according to three people familiar with the matter.

Some employees said colleagues were using the software to automate additional, unnecessary AI activity to increase their consumption of tokens—units of data processed by models.

They said the move reflected pressure to adopt the technology after Amazon introduced targets for more than 80 percent of developers to use AI each week, and earlier this year began tracking AI token consumption on internal leader boards.

“There is just so much pressure to use these tools,” one Amazon employee told the FT. “Some people are just using MeshClaw to maximize their token usage.”

cross-posted from: lemmy.ml/post/47169376

Over the past decade, the AI industry has come to exert an unprecedented economic, political and societal power and influence. It is therefore critical that we comprehend the extent and depth of pervasive and multifaceted capture of AI regulation by corporate actors in order to contend and challenge it. In this paper, we first develop a taxonomy of mechanisms enabling capture to provide a comprehensive understanding of the problem. Grounded in design science research (DSR) methodologies and extensive scoping review of existing literature and media reports, our taxonomy of capture consists of 27 mechanisms across five categories. We then develop an annotation template incorporating our taxonomy, and manually annotate and analyse 100 news articles. The purpose behind this analysis is twofold: validate our taxonomy and provide a novel quantification of capture mechanisms and dominant narratives. Our analysis identifies 249 instances of capture mechanisms, often co-occurring with narratives that rationalise such capture. We find that the most recurring categories of mechanisms are Discourse & Epistemic Influence, concerning narrative framing, and Elusion of law, related to violations and contentious interpretations of antitrust, privacy, copyright and labour laws. We further find that Regulation stifles innovation, Red tape and National Interest are the most frequently invoked narratives used to rationalise capture. We emphasize the extent and breadth of regulatory capture by coalescing forces -- Big AI and governments -- as something policy makers and the public ought to treat as an emergency. Finally, we put forward key lessons learned from other industries along with transferable tactics for uncovering, resisting and challenging Big AI capture as well as in envisioning counter narratives.

Full paper: PDF | HTML | TeX source

Jure Repinc

2026-05-11 11:25:07

Big AI's Regulatory Capture: Mapping Industry Interference and Government Complicity

Over the past decade, the AI industry has come to exert an unprecedented economic, political and societal power and influence. It is therefore critical that we comprehend the extent and depth of pervasive and multifaceted capture of AI regulation by corporate actors in order to contend and challenge it. In this paper, we first develop a taxonomy of mechanisms enabling capture to provide a comprehensive understanding of the problem. Grounded in design science research (DSR) methodologies and extensive scoping review of existing literature and media reports, our taxonomy of capture consists of 27 mechanisms across five categories. We then develop an annotation template incorporating our taxonomy, and manually annotate and analyse 100 news articles. The purpose behind this analysis is twofold: validate our taxonomy and provide a novel quantification of capture mechanisms and dominant narratives. Our analysis identifies 249 instances of capture mechanisms, often co-occurring with narratives that rationalise such capture. We find that the most recurring categories of mechanisms are Discourse & Epistemic Influence, concerning narrative framing, and Elusion of law, related to violations and contentious interpretations of antitrust, privacy, copyright and labour laws. We further find that Regulation stifles innovation, Red tape and National Interest are the most frequently invoked narratives used to rationalise capture. We emphasize the extent and breadth of regulatory capture by coalescing forces -- Big AI and governments -- as something policy makers and the public ought to treat as an emergency. Finally, we put forward key lessons learned from other industries along with transferable tactics for uncovering, resisting and challenging Big AI capture as well as in envisioning counter narratives.

Full paper: PDF | HTML | TeX source

Big AI's Regulatory Capture: Mapping Industry Interference and Government Complicity

^arXiv.org

Over the past decade, the AI industry has come to exert an unprecedented economic, political and societal power and influence. It is therefore critical that we comprehend the extent and depth of pervasive and multifaceted capture of AI regulation by corporate actors in order to contend and challenge it. In this paper, we first develop a taxonomy of mechanisms enabling capture to provide a comprehensive understanding of the problem. Grounded in design science research (DSR) methodologies and extensive scoping review of existing literature and media reports, our taxonomy of capture consists of 27 mechanisms across five categories. We then develop an annotation template incorporating our taxonomy, and manually annotate and analyse 100 news articles. The purpose behind this analysis is twofold: validate our taxonomy and provide a novel quantification of capture mechanisms and dominant narratives. Our analysis identifies 249 instances of capture mechanisms, often co-occurring with narratives that rationalise such capture. We find that the most recurring categories of mechanisms are Discourse & Epistemic Influence, concerning narrative framing, and Elusion of law, related to violations and contentious interpretations of antitrust, privacy, copyright and labour laws. We further find that Regulation stifles innovation, Red tape and National Interest are the most frequently invoked narratives used to rationalise capture. We emphasize the extent and breadth of regulatory capture by coalescing forces -- Big AI and governments -- as something policy makers and the public ought to treat as an emergency. Finally, we put forward key lessons learned from other industries along with transferable tactics for uncovering, resisting and challenging Big AI capture as well as in envisioning counter narratives.

Full paper: PDF | HTML | TeX source

Last fall, we featured an extensive interview with Petter Törnberg of the University of Amsterdam, who studies the underlying mechanisms of social media that give rise to its worst aspects: the partisan echo chambers, the concentration of influence among a small group of elite users (attention inequality), and the amplification of the most extreme divisive voices. He wasn’t optimistic about social media’s future.

Törnberg’s research showed that, while numerous platform-level intervention strategies have been proposed to combat these issues, none are likely to be effective. And it’s not the fault of much-hated algorithms, non-chronological feeds, or our human proclivity for seeking out negativity. Rather, the dynamics that give rise to all those negative outcomes are structurally embedded in the very architecture of social media. So we’re probably doomed to endless toxic feedback loops unless someone hits upon a brilliant fundamental redesign that manages to change those dynamics.

Törnberg has been very busy since then, producing two new papers and one new preprint building on this realization that social media is structured quite differently than the physical world, with unexpected downstream consequences. The first new paper, published in PLoS ONE, specifically focused on the echo chamber effect, using the same combined standard agent-based modeling with large language models (LLMs)—essentially creating little AI personas to simulate online social media behavior.

Real talk: last month I was running a giveaway campaign for a client. The mechanic was simple — comment to enter, tag a friend for a bonus entry. 3,200 comments later, I was staring at a blank Google Sheet wondering how I was going to verify entries, remove duplicates, and pick a winner without losing my mind.
Instagram doesn't give you any export functionality. Zero. You can view comments in the app, you can reply, you can delete — but you cannot export them in any structured way. This is apparently a deliberate product decision, and it's been this way for years.
What I tried first:

Manually copy-pasting — obviously not scalable past ~50 rows
The official Instagram Graph API — requires app review, business account verification, and only returns data from your own posts anyway
Third-party "Instagram data export" services — most of these ask for your password or OAuth credentials, which is a non-starter

What actually worked:
I ended up using a browser extension called Instagram Comments Scraper that runs entirely within your browser session. No password required — it just operates within your existing logged-in session, the same way you're already viewing the comments. The data is processed locally and never sent anywhere external.
The output columns it gives you: comment ID, comment text, username, profile URL, profile pic URL, and timestamp. That's exactly what you need to do any meaningful analysis — filter by date, spot bot accounts, remove duplicates, identify authentic entries.
The rate limiting situation:
The part I didn't expect was how Instagram's rate limits work. There's no published threshold — it varies by IP and activity patterns. When the scraper hits a limit, it enters a cooldown mode automatically (the timer shows you how long), then doubles the cooldown if the limit persists. Once the cooldown clears and a request succeeds, it goes back to normal. This meant I could walk away and come back to a finished export rather than babysitting it.
End result: 3,200 comments exported to Excel in about 40 minutes of unattended processing. Filtered to valid entries (tagged a user + original commenter had 10+ followers) in another 20 minutes using basic Excel formulas.
Caveat I'd add for anyone doing this:
Be reasonable about volume and timing. Don't run 10,000-comment scrapes back-to-back on the same IP. The human-like delay system in the tool helps, but bulk scraping in one long session still carries some account risk. Space it out if you're working with large datasets.
Anyone else found better approaches to this problem? Especially curious if anyone's had success with the official API for use cases beyond your own posts.