Marketplace Tech

Anthropic announced it's adding invisible watermarks to all text generated by its chatbot Claude. The AI startup is rolling out the policy globally as part of efforts to comply with the European Union AI Act.

Plus, there's a patchwork of state laws that regulate AI-generated political messaging — they're being put to the test ahead of the midterm elections. And federal regulators ordered Kalshi to continue to operate in New York, despite a state lawsuit seeking to shut it down.

Marketplace’s Meghan McCarty Carino spoke with Maria Curi, tech policy reporter at Axios, about all this for this week’s “Tech Bytes: Week in Review.”

Generative AI models can produce enormous volumes of text, images and video in seconds. But without expert human instruction, the output is often generic. (You might say slop.)

A number of startups in Silicon Valley are on a mission to change that by giving AI a more human sense of taste.

Marketplace’s Meghan McCarty Carino spoke with critic and journalist Sophie Haigney about what taste is and whether artificial intelligence can truly replicate it.

This story includes discussion of suicide.

Back in 2021, TikTok tweaked its algorithm in an effort to address dangerous filter bubbles — a sort of algorithmic spiral of recommendations for high-risk content like extreme dieting or self-harm. But internal company documents obtained by Bloomberg Businessweek show 10% of U.S. users — 15 million Americans — were purposely excluded from this safety update. This was part of user testing to validate the new feature and track whether it reduced engagement with the famously sticky platform.

One of those users was then-16-year-old Chase Nasca, who subsequently died by suicide. Olivia Carville, Bloomberg Businessweek investigative reporter, has been on the story.

Remember those grainy old baby monitors? So passé. Today, the modern nursery can be equipped with HD cameras and biometric sensors — all analyzed with artificial intelligence.

The technology can detect hazardous position changes or a blanket covering a baby's face, but also tracks less critical metrics that purport to help optimize infant sleep and health. Sapna Maheshwari, business reporter at The New York Times, recently wrote about this. She's been an avid user of the Nanit Smart Baby Monitor System with both of her children.

More on this

“Aw, It’s Baby’s First A.I. Surveillance System” from The New York Times

“AI-powered IoT is creating a new security challenge” from Spiceworks

Substack, the platform popular for online newsletters, recently launched an artificial intelligence detection feature powered by Pangram.

The detector allows users to scan text and get an estimate for how much of it was likely written by AI. But some creators are concerned about the tool’s accuracy and its potential to prompt a “witch hunt” for writers who use AI.

Marketplace’s Meghan McCarty Carino spoke with Substack CEO Chris Best about the new feature.

More on this:

AI Content Is Everywhere on Social Media, Especially LinkedIn from Pangram

I Test-Drove Substack’s New AI Detection Tool, and It Mostly Worked from CNET