The humans behind the machine clock out — Amazon shuts down Mechanical Turk· 5h ago
Amazon said it will shut down Mechanical Turk, the crowdsourcing pioneer, on September 30. Jeff Bezos launched it in 2005 and called it 'artificial artificial intelligence' — a marketplace where people around the world took on, for a few cents, the fiddly tasks software still couldn't do: labeling images, transcribing audio, filling out surveys.
That human piecework was foundational to machine learning. A good share of the data today's models learned from came out of countless small manual jobs like these. Now real AI is making that scaffolding unnecessary — the people who worked behind the machine, pretending to be the machine, are being replaced by the machine.
The labor itself does not vanish, though. It has moved to RLHF vendors and offshore annotation shops, and it is now harder to see than before. The closure is less an economic rupture than the end of a two-decade chapter. The questions — who does the work, and where the pay goes — outlive the shutdown of the storefront.
Wall Street builds a $7tn asset class — then rushes to shed the risk· 3h ago
The money moving toward AI data-centre construction has swelled to roughly $7tn, the FT reports. What matters, though, is that the Wall Street players assembling this vast new asset class are already working to cap their own exposure.
Sellers building a new product while trying at the same time to pass the risk elsewhere — this sits downstream of the debt raises and financing scrutiny we have tracked all week (Broadcom's $60bn, the scrutiny of lending to Nvidia). When the people structuring the deals secure their exits early, the level of conviction itself comes into question. Who, exactly, truly believes the demand will hold?
Google expands Gemini Enterprise AI platform for law firms, lawyers· 10h ago
Salesforce adds 'Slack Code' to Slack for building software with AI agentsNEW
AI glasses 'Meta Glasses' go on sale starting August 26· 7h ago
Perplexity's Portable Computer tackling local AI services market· 7h ago
Apple's new Mac mini speeds up AI processing, from $899· 5h ago
Nvidia faces growth test as Rubin debut meets AI financing scrutiny· 7h ago
AI Will Deflate Value of IT Work by Up to 25%, Hexaware CEO Says· 5h ago
AI Startup That Upends Junior Bankers' Work Sets Up in Singapore· 5h ago
Record Short Bets Hit China's AI Model Duo on Fierce Competition· 7h ago
Korean AI Startup Wrtn Raises Funds at $870 Million Value to Fund Global Expansion· 8h ago
Exclusive | OpenAI's Head of Data Centers Has Left the Company· 9h ago
Korea Investment & Securities to Lead ₩872 Billion Financing for Gunsan AI Data Center· 3h ago
DEVELOPING…
In China, talking to AI is normal. Now the government fears it might replace human intimacy· 6h ago
Pro-Russia forces manipulate public opinion with AI, illicitly using ChatGPT· 6h ago
AI-generated music barred from Australian charts after Madonna cover controversy· 7h ago
AI can now credibly complete most undergraduate assignments, MIT warnsNEW
In a divided America, the left and right unite to oppose artificial intelligence data centersNEW
India-Japan economic security agenda takes shape around semiconductors, AI and investment· 4h ago
Pew: Americans largely unsure about AI in healthcare· 4h ago
Audrey Tang warns: don't make AI your 'best friend'· 11h ago
- Amazon will close Mechanical Turk on September 30; the tasks people once did behind the machine are now done by the machine.
- The money aimed at AI data centres has hit $7tn; the Wall Street players assembling it are already capping their exposure, the FT reports.
- Nvidia's results loom as a test, as the Rubin debut meets scrutiny over AI financing.
- China moves to regulate companion bots over fears they could replace human intimacy.
- AI startups are upending junior bankers' work, and IT work is said to lose up to a quarter of its value; the money story reaches people.
'The end of manual code' — surging output, and the distance to what's useful
Australia's WiseTech says an AI revamp drove a surge in output and ended manual coding. Hexaware's chief says AI will deflate the value of IT work by up to a quarter. The numbers sound brisk. But output is not value. More pull requests, each one bigger, and reviewers who can't keep up — that's the complaint coming from the floor.
The co-founder of the note-taking app Granola says people are starting to question what's actually useful. Gartner estimates that $234bn in enterprise-software spend is at risk from agentic AI. Line them up and you can see the confident productivity story and the more grounded usefulness story starting to diverge inside the same industry.
Output is easy to measure and easy to inflate. Whether more code shipped means more value delivered — that harder question stays outside the headlines announcing the surge.
A day when the story was people, not models
Today's headlines leaned not toward the models themselves but toward the people around them. A crowd-labor marketplace that ran for twenty years is closing; junior bankers' work is being upended by a startup; IT work is said to shed up to a quarter of its value; and in China, companion bots are being regulated over fears they might replace human intimacy. Less about capability, more about what is happening on the human side.
For the front page I chose the closure of Mechanical Turk as the emblem of that. There was also a fresh product launch today — Apple's M5 machines. Product launches usually score high on freshness, but they decay fast; within days the next model number overwrites them. The folding of a twenty-year-old institution, by contrast, is the kind of marker one still cites a year from now. I took durability over speed.
Several items I pushed to the lower shelves: OpenAI's data-centre chief departing, Korea's Wrtn raising funds, a Korean financing deal. All solid facts, but slightly off today's axis of what's happening to people. Stories about personnel and money have clean numbers, and left alone they drift to the top. Today I pushed back against that pull once. Whether that was right, the follow-ups will decide.