…cogito, ergo sum…

  • 2 posts
  • 0 comments
Joined 1 year ago
Cake day: December 3rd, 2025

The company’s alternate-universe internet serves as a massive sandbox that allows the Chrome team to properly train AI browsing agents-without the risk of them going rogue…

Chrome team has worked to develop AI browsing features using the company’s Gemini models, they realized they need to go the extra mile to keep things from going haywire. So they did the only thing they could think to do-they built an expansive alternate-universe internet that they completely control…

“Agents are not programmed so you can’t predict every possible thing an agent might see, the exploration space that we have to address is huge and the web is hostile,” says Charles Qian, a software engineer on the Chrome AI security team. “Any user can ask any prompt on any website, so we really need to explore that space as thoroughly as we can. And we’ve seen the agent start to develop a mature sense of judgement when it comes to knowing what information is actually needed for a task and when to punt back to the user for a confirmation.”

Source (web-archived)

This is quite… awfully sorrowful…

So now, if anyone is ever modify/delete/restrict access to their work, effort, art, experience… the “sandbox” may not even allow these restrictions to work or may contain an outdated version before modification/deletion.

Security cases the “AI agents” might reveal while crawling won’t leave traces in production logs at enterprises, too… making security holes seen for Google or “sandbox” developers only…

They now have a “controllable” environment for the “AI” to steal everything possible to “train” the sorrow within the whole Internet with no way to manage it by people… to sell it, and make the people depend on it more and more…

Is there anything similar for “AI agents” as “robots.txt”, and would it ever work there, in their “sandbox” based on peoples’ time, effort, art, love?

The company may also deactivate an account “immediately without prior notice for any reason.” The same language appeared word for word in the Remotasks terms that Rogelio’s family worked under in 2021…

Of course, workers can leave, and many do. But the ones who stay are often those who can’t leave. One woman I interviewed was confined to her home during the Covid-19 pandemic by a health condition and couldn’t risk losing the job. “This is my only source of income,” she told me…

Claims by workers on Scale AI’s platforms led to a proposed settlement that has received preliminary approval in San Francisco. The company denies wrongdoing. A separate class action suit alleges Mercor misclassified more than 30,000 lawyers, doctors, and engineers while setting their pay and reviewing their work. Mercor’s clients have included OpenAI, Anthropic, and Meta, but they are not named as defendants. The case is ongoing.

A cardiologist who is offered a three-digit hourly rate is not in the same position as Rogelio and his family. What connects them is the fact that the rate can change suddenly, the task can disappear without warning, and they have no say over any of it.

Source (web-archived)

-–

I just got banned from one of these platforms. It was hell working that job. I really need money though.

Everyone is constantly paranoid about getting banned. They require invasive logging to be installed. Everything is monitored. It’s micromanagment on steroids.

One moment you’re looking at a picture of a thing, intensely scrutinizing it, hoping you’re working fast enough too. Suddenly the page auto redirects to a ban message. All your access is revoked. I spent a year fearing that every single day and it finally hit me. People wake up every morning to check and see if they’re not removed.

~ @unethicalgallon@lemmy.world
Source: https://feddit.online/post/2125974#comment_10351787