Reuters: Why Tesla’s AI trainers don’t trust its self-driving tech – or its safety stats. “Tesla says its Full Self-Driving software is up to 10 times safer than human drivers. But the figures the company uses to support its claims don’t withstand scrutiny – and staffers who trained the technology say it isn’t close to safely delivering autonomous vehicles at scale.”

https://rbfirehose.com/2026/05/28/reuters-why-teslas-ai-trainers-dont-trust-its-self-driving-tech-or-its-safety-stats/
Reuters: Why Tesla’s AI trainers don’t trust its self-driving tech – or its safety stats

Reuters: Why Tesla’s AI trainers don’t trust its self-driving tech – or its safety stats. “Tesla says its Full Self-Driving software is up to 10 times safer than human drivers. But the figure…

ResearchBuzz: Firehose

TechCrunch: This startup is betting India’s gig economy can train the world’s robots. “In the last few years, India’s online food delivery market has grown significantly, with both Zomato and Swiggy going public and the number of cloud kitchens increasing. … Silicon Valley-based startup Human Archive is tapping into this trend, partnering with these companies to have workers wear […]

https://rbfirehose.com/2026/05/27/techcrunch-this-startup-is-betting-indias-gig-economy-can-train-the-worlds-robots/
TechCrunch: This startup is betting India’s gig economy can train the world’s robots

TechCrunch: This startup is betting India’s gig economy can train the world’s robots. “In the last few years, India’s online food delivery market has grown significantly, with both Zomato and…

ResearchBuzz: Firehose

"I no longer knew what the Golden Task of my own life might look like. When given the chance to respond to details from this article, Mercor said that it strives to give workers “as much notice as possible when these projects change”—a sentiment roughly echoed by other companies. Between February and April 2026, I was hired and fired on seven different projects over four different platforms. The dismissals were always abrupt, shocking. One moment I would be typing rubrics into an Airtable, waiting in line on a 24-hour Zoom to talk through a task with a reviewer. The next, the UI would vanish. The Slack channel would disappear. The Google docs would lock me out. No message. No warning. No explanation.

I never intended to write about this industry. I came to it not as a journalist but as a disgruntled, broke TV writer determined to make a dent in student loans and keep paying LA rent while my industry withered in front of me. But working with and for AI had proven even more cruel than I could have ever imagined. Mercor says it employs about 300 full-time staffers. Meanwhile, each week it keeps some 30,000 independent contractors caught up in a fever dream of aimless, directionless urgency, corralled across Slack channels by achingly young adults, sending messages at 3 am to “push on” and “finish strong” and “lock in” and “Go Team GO!” All in service of the grandest purpose in history: to successfully remove a scuba diver from a picture with one click of a mouse, transport him to the moon without any glaring artifacts—and bring him back again.

The next generations of team leaders won’t know our specific talents or our unique skills, but they will know the Average Time it takes us to annotate a grainy video uploaded without the owner’s consent into a vast catalog of other possibly stolen videos (...) To make the machine more human, they will make us more like the machine."

https://www.wired.com/story/i-work-in-hollywood-everyone-who-used-to-make-tv-now-training-ai/

#AI #GenerativeAI #AITraining #LLMs #Mercor #GigEconomy

I Work in Hollywood. Everyone Who Used to Make TV Is Now Secretly Training AI

For screenwriters like me—and job seekers all over—AI gig work is the new waiting tables. In eight months, I’ve done 20 of these soul-crushing contracts for five different platforms. It’s bad.

WIRED

https://winbuzzer.com/2026/05/19/andrej-karpathy-joins-anthropic-claude-pre-training-push-xcxwbn/

Andrej Karpathy has joined Anthropic, with the move tied to pre-training work and a reported Claude-assisted research team.

#AI #Claude #Anthropic #AndrejKarpathy #AIResearch #AITraining #AIModelDevelopment

Plan For Teachers to Record Children For AI Training Purposes Scrapped After Parent Backlash

It was opt-out.

PetaPixel

Bloomberg: Musk’s xAI promised staff $420 for their tax returns, hasn’t paid. “Elon Musk’s xAI asked employees earlier this year to offer up their own tax returns as training data for the company’s Grok chatbot, promising a $420 payment as incentive for doing so. Two months later, those payments still haven’t materialized.”

https://rbfirehose.com/2026/05/19/bloomberg-musks-xai-promised-staff-420-for-their-tax-returns-hasnt-paid/
Bloomberg: Musk’s xAI promised staff $420 for their tax returns, hasn’t paid

Bloomberg: Musk’s xAI promised staff $420 for their tax returns, hasn’t paid. “Elon Musk’s xAI asked employees earlier this year to offer up their own tax returns as training data for the com…

ResearchBuzz: Firehose

Reuters: US judge considers Anthropic’s $1.5 billion settlement of authors’ lawsuit. “A federal judge on Thursday pressed lawyers for more information on several points of artificial intelligence company Anthropic’s proposed $1.5 ‌billion settlement with authors who accused it of misusing their books to train ‌its AI chatbot Claude.”

https://rbfirehose.com/2026/05/17/reuters-us-judge-considers-anthropics-1-5-billion-settlement-of-authors-lawsuit/

Self-Distillation

Discover how self-distillation enables continual learning with minimal overhead

https://airanked.dev/posts/self-distillation-in-ai

#SelfDistillation #ContinualLearning #AiTraining