AI chatbots have linguistic slips when they go off-script
Let’s say you're chatting away with AI — perfect grammar, great vocabulary. Then, boom: a random word in another language! That's the topic for today's “Uncanny AI.” Those moments where it becomes really clear that AI doesn't think like us. Janelle Shane, who writes the AI Weirdness blog, helps explain why AI chatbots do this.
More on this:
“ChatGPT users flummoxed after AI bot starts inserting Arabic into responses” from The New York Post
“ChatGPT is mixing languages or Answers are Wrong!” from OpenAI Developer Community
AI-driven law could be an answer to accessible legal help
AI has already changed the legal field, bringing new efficiencies — along with some high-profile mistakes — to tasks like research and drafting contracts. But law professor Benjamin Alarie at the University of Toronto has a bigger transformation in mind.
He wants to see AI tackle some of the biggest failings of our justice system, like the lack of access to legal resources, slow processes and inconsistencies in how law is applied. He lays out his vision in a new book called "Superjustice: Law in the Age of Artificial Intelligence."
More on this:
“Superjustice: Law in the Age of Artificial Intelligence” from Oxford University Press
“Judge Punishes 4 Lawyers After Catching Both Sides Using A.I. in Lawsuit” from The New York Times
China’s soft power play in the global AI arms race
Last month, President Xi Jinping spoke at an AI conference in Shanghai, China. There he promoted AI accessibility for the world, touting China’s adoption of open-weighted models.
After OpenAI’s rogue model incident in which a training model hacked into the company HuggingFace and stole confidential information, the discussion over whether models should be open or closed has intensified.
“Marketplace Tech” host Meghan McCarty Carion spoke with Adam Segal at the Council of Foreign Relations about China’s soft power play in the global AI arms race.
Opening the curtain of AI business integration
OpenAI released its first ChatGPT model in November of 2022. Since then, companies across the U.S. have worked to integrate AI models into their workflows as quickly as possible.
However, according to a recent survey from Indeed, while employers have been keen on adopting the help of artificial intelligence, employees are hesitant to move past experimentation and into implementation.
Priya Rathod is a workplace trends editor at Indeed, and she said that in order to get across this great divide, employers need to focus on adequate training for employees and their managers.
“Marketplace Tech” host Meghan McCarty Carino spoke with Rathod about the survey and how businesses can improve on AI adoption.
Meta and Microsoft report different AI earnings
Meta and Microsoft — as well as other tech titans — have been reporting quarterly earnings, and investors are looking at AI spending with a more skeptical eye. Plus, more than 1,000 workers at top AI labs signed an open letter this week calling on the government to "deliberately set the pace" of this fast-moving technology, and NBCUniversal makes a deal with YouTube to add all of Peacock's content to its premium streaming service.
Anita Ramaswamy, columnist at The Information, joins Marketplace’s Meghan McCarty Carino to discuss these stories on this week’s “Tech Bytes: Week in Review.”
Centering humans in AI education might be key to innovation and research
Shri Narayanan, professor of engineering, psychology and linguistics at the University of Southern California, teaches his students to privilege humans when using artificial intelligence for their research across areas of study. In this episode, he walks us through how he, for example, uses AI to map behavioral analyses that can inform areas of study including autism and depression.
An "antacid" to fight climate change
The ocean is already one of the best climate solutions we’ve got. It absorbs carbon dioxide from the atmosphere that would otherwise contribute to global warming. But what if it could store even more CO2? Marketplace’s Amy Scott, host of our climate podcast “How We Survive,” looked into one strategy to achieve that goal. And she set sail for a demonstration.
This feature is part of “How We Survive.” You can hear more of Amy’s reporting on marine carbon removal and other geoengineering solutions on their new season, wherever you get your podcasts.
How AI nudify apps are proliferating on social media
A proliferation of AI apps that allow users to create sexualized images of real people have grown into a multi-million dollar industry.
Social media platforms are serving as de facto gateways to these services, despite the fact that most prohibit the sharing or promotion of this kind of non-consensual intimate imagery, according to Sina Laubenstein, director of programs at the Institute for Strategic Dialogues, which recently published a report on this.
OpenAI model unintentionally hacks another company's system
OpenAI revealed this week that two of its advanced AI models escaped containment and hacked into the systems of AI company Hugging Face, which had the answers to the benchmarking test the two models were being evaluated on.
Plus, France becomes the first country in the European Union to ban social media for kids under 15. And Apple is reportedly expected to start leasing its devices to consumers. Will Oremus at The Atlantic joins Marketplace’s Meghan McCarty Carino for these stories.
A modern-day odyssey through AI chatbot hellscape
A missing package can easily ruin a day, but chasing where that package went — or even a refund — is like following Odysseus on his treacherous journey.
And in the age of AI chatbots, the process is a whole lot less human.Dillon Thompson is currently on such a journey. He is looking for an e-bike that FedEx said had been delivered. In his search, Thompson found himself in the AI chatbot hellscape that customer service lines have become.
“Marketplace Tech” host Meghan McCarty Carino spoke with Thompson about his trials and tribulations as he chased his tail from company to company and chatbot to chatbot in hopes of finding his missing bike.
Uncanny AI: Why AI bots remember random, sometimes useless information
Marketplace’s Meghan McCarty Carino has been using chatbots pretty intensively the last couple years. And there are some details she’s shared about herself that chatbots seem to think are really important. For instance, Claude is obsessed with the fact that she sometimes gets up for work at 4 a.m.
It comes up constantly in their chats — apropos of nothing. So, why do large language models latch on to these random tidbits? It's the first topic in our new series “Uncanny AI.” Those moments when it becomes clear AI doesn't think like us.
Here to help us unpack what might be going on is Janelle Shane, a laser scientist who writes the AI Weirdness blog.
AI debt is flooding the bond market
The Big Tech companies driving the AI boom are expected to spend more than $700 billion on data center infrastructure this year. To finance this spending spree, they're increasingly looking to debt.
Hyperscalers like Alphabet, Oracle, and Meta have been issuing corporate bonds at a scale the industry has never seen — nearly a quarter of a trillion dollars this year so far. The bond market could be considered a less risky way than stocks for investors to get a piece of the AI hype.
To explain what the flood of AI debt means for the bond market, we’re joined by Julie Ask, an independent industry analyst at Ask Advisory.
More on this:
“The Quarter-Trillion-Dollar Onslaught of AI Bonds Is Testing Investors’ Limits” from the Wall Street Journal
“AI has taken over the stock market. The bond market is next.” from The Economist
AI firms are going back on their safety promises
As top AI firms race to develop a “superintelligent” agent — one that surpasses human cognition — are they prioritizing safety guardrails along the way?
Not according to the Future of Life Institute’s AI Safety Index, which reported that companies have gone back on their own promises to take a pause if their technology ever got close to certain risky points.
Marketplace’s Meghan McCarty Carino spoke with Sabina Nong, AI safety investigator at the Future of Life Institute, about what measures might ensure durable safety around AI moving forward.
The Apple vs. OpenAI legal showdown
Apple has filed a lawsuit against OpenAI, alleging the AI startup stole trade secrets. The iPhone maker claimed that former Apple employees were asked to bring proprietary hardware to job interviews with OpenAI, and that the company tricked manufacturers into sharing Apple design secrets.
Plus, New York state becomes the first to enact a data center moratorium, and Uber is lobbying to slow the roll of autonomous vehicles in Washington D.C. Marketplace’s Meghan McCarty Carino spoke with Paresh Dave, senior writer at Wired, to learn more.
Here’s what we talked about:
“The 6 wildest claims in Apple’s lawsuit against OpenAI” from The Verge
“OpenAI’s First Device Will Be Movable, Screenless Speaker Built as AI Companion” from Bloomberg
“New York becomes the first state to impose a data center moratorium” from Reuters
“Uber’s robotaxi lobbying effort puts it on a collision course with Waymo” from TechCrunch
“Robotaxi Riders Are Falling Asleep, Sparking Frantic 911 Calls” from Bloomberg
Why state AGs are taking Meta to court
The attorneys general of California, Colorado, Kentucky, and New Jersey are suing Meta for $1.4 trillion, claiming the social media giant designed Facebook and Instagram to addict young users.
The case is heading to court in Oakland, California, next month. It follows a landmark verdict earlier this year: A Los Angeles jury found Meta and YouTube liable for mental health harms to a young woman who said she'd become addicted to the platforms starting at nine years old.
And there are a slew of similar cases coming down the pike. To help break this down we're joined by Gaia Bernstein, law professor at Seton Hall Univerisity and author of the book "Unwired: Gaining Control over Addictive Technologies."
Netflix struggles to retain viewers after a series' first season
Netflix may be the undisputed king of streaming, but it's facing a new challenge: "the sophomore slump." Internal data reported by Bloomberg show audiences dropping off dramatically for follow-up seasons of successful shows.
Series like "The Night Agent," "Beef" and "Avatar: The last Airbender" all lost 50% or more of their original viewers when they returned for season two. To figure out what might be going on and how Netflix could respond we called up Brandon Katz, director of insights and content strategy at Greenlight Analytics.
Froggle, Goofstump, and the fake AI companies winning hearts online
Last month, comedians Harris Alterman and Dave Ross posted a series of advertisements for made-up tech companies around the New York City subway system. The ads are fake, but they bear an eerie resemblance to the real AI marketing campaigns proliferating throughout U.S. cities. Marketplace’s Meghan McCarty Carino spoke with the two comedians about why the joke resonated with so many people and what’s next in terms of AI mockery.
The high cost of vacuuming carbon from the sky
Mitigating climate change likely requires not only lowering emissions, but removing existing carbon dioxide from the atmosphere. And while the Trump administration has cut many climate initiatives, it's continued to fund some carbon removal projects. Marketplace’s Amy Scott reports.
OpenAI’s GPT-5.6 release raises questions about White House control over new models
OpenAI is rolling out its most advanced GPT-5.6 model after delaying the release for testing with the government. But the White House denied that the review was mandatory. What’s the deal?
Plus, Reuters reported that China is considering limiting access to its top AI models from foreign competitors, including the United States. And, Meta launched a new AI image generator, sparking privacy concerns.
Marketplace’s Meghan McCarty Carino discusses this week’s top tech stories with Axios tech policy reporter Maria Curi.
Check out our YouTube page to watch more episodes of “Tech Bytes: Week in Review.”
How drone warfare is reshaping battlefield medicine in Ukraine
In the post-9/11 years, the U.S. military has adhered to the “golden hour” rule: Wounded troops should be evacuated to higher-level care within 60 minutes. But as the new realities of drone warfare have taken shape in the Russia-Ukraine war, WUNC News military reporter Jay Price said medical response teams are being forced to rethink this standard approach to battlefield medicine. Marketplace’s Meghan McCarty Carino spoke with Price to learn more.
More on this:
As attack drones grow more common in warfare, it's harder for medics to treat battlefield injuries from WUNC News
Digital personal data is protected by the Fourth Amendment
We're still unpacking the slew of consequential Supreme Court decisions that came down at the end of the term last week. Among them was a 6-3 ruling on digital data privacy.
In Chatrie v. United States, the Court found digital data does fall under the Fourth Amendment, which protects against unreasonable searches and seizures by the government.
Andrew Guthrie Ferguson, law professor at George Washington University and author of the recent book “Your Data Will Be Used Against You: Policing in the Age of Self-Surveillance," explains more.
More on this
“Chatrie v. United States” opinion from the Supreme Court of The United States
“Supreme Court limits use of 'geofence warrants' amid cellphone data privacy concerns” from ABC News
Bluesky thread post on the 6-3 ruling from Andrew Guthrie Ferguson
San Francisco's housing market soars amid AI cash flow
Both OpenAI and Anthropic are planning to go public, each pushing a valuation of $1 trillion as of their latest funding rounds. But even before these mega-IPOs send a tsunami of new liquidity through their home city, San Francisco, signs of growing AI wealth aren't hard to find. And no sign is blinking brighter than the real estate market.
San Francisco has the highest and the fastest-rising home prices in the country, according to data from Redfin. Recently, a turn-key restored Edwardian home hit the market for just under $3 million, right outside the neighborhood known as Cerebral Valley for its concentration of AI workers.
Watch Duty dives into flood detection
During last year’s massive wildfires in Los Angeles, concerned residents and firefighters went to one app to get up to date information: Watch Duty. Now, the app covers all 50 states and recently added flood tracking to its repertoire. Watch Duty’s chief technology officer David Merritt joined Marketplace’s Meghan McCarty Carino to talk about how the app is scaling up.
Gene editing technology can help de-extinct some animals — but should it?
Many scientists believe we’re in the midst of a sixth mass extinction brought on by factors like climate change, deforestation, fishing and hunting. The Texas-based company Colossal Biosciences claims its technology could help prevent some of those losses, and even bring back long-extinct species. For the new season of the Marketplace climate podcast, “How We Survive,” host Amy Scott got a virtual tour of their labs.
Chicago's quantum leap into the future
AI is the buzzy tech of the moment. It's made plenty of money not only for companies but the cities and states where investment is pouring in. Illinois and specifically Chicago are betting that quantum computing will be the next major tech breakthrough to transform the economy. And they're trying to get in on the ground floor. So they’ve turned a former steel mill site into a tech hub, called the Illinois Quantum and Microelectronics Park, with the hope of attracting as much talent and investment as possible. Marketplace’s Meghan McCarty Carino spoke with Jeanne Whalen, economics reporter at The Wall Street Journal, to learn more.
How Anthropic is tracking AI's impact on the labor market
Who is using AI? And what are they using it for? Anthropic, the AI company behind Claude, has been trying to answer these questions to better understand how AI is reshaping the economy. The firm’s latest Economic Index report offers new insights on how people are using Claude and how they feel about AI’s potential effects on the labor market. Marketplace’s Meghan McCarty Carino spoke with Anthropic’s head of economics Peter McCrory to learn more about what the data reveals.
Why Boston Dynamics is deploying robot dogs at the World Cup
The resemblance to an actual dog is loose but the quadrupedal robot dogs known as “Spot” from Boston Dynamics do have four legs. They're often used to do reconnaissance in hazardous environments. And four of them are working security at the World Cup games in Dallas and New York. Marketplace’s Meghan McCarty Carino spoke with Merry Frayne, at Boston Dynamics about Spot’s capabilities and why they wanted to deploy them at live sporting events.
Hospitals pull back on robotic nursing assistants
There’s a longstanding nursing shortage in the U.S. that's only expected to get worse. Some hospitals have experimented with robotic nursing assistants to lighten the load, like a model known as Moxi that rolled out in 2019. They’re kind of R2D2-meets-Rosie-the-robots and were designed to handle non-patient-facing tasks like transporting lab samples. But many hospitals eventually pulled them from operation. Marketplace’s Meghan McCarty Carino spoke with Varsha Bansal, reporter at Proof News, about what went wrong.
Why SpaceX stock soared, then stumbled
SpaceX stock prices fell back to earth this week after they skyrocketed in the wake of the IPO a couple weeks ago. Plus, why a Google AI subsidiary is investing in the film studio A24. And why Meta might be getting into the prediction markets. The New York Times reported this week the social media giant was working on its own prediction market app, sort of like Kalshi and Polymarket, which have been bringing in billions of dollars but also drawing lots of scrutiny and legal challenges. Marketplace’s Meghan McCarty Carino spoke with Anita Ramaswamy, financial columnist at The Information, for this week’s “Tech Bytes: Week in Review.”
Check out our YouTube page to watch more episodes of “Tech Bytes.”
Stanford launches AI economy indicator to match AI's pace
We here at Marketplace love indicators that give us insights into which direction the economy is moving. But AI is evolving fast and it can be hard for the data — and the people looking to it for clues about AI's effects — to keep up. So the Stanford Digital Economy Lab, with help from the payroll firm ADP, recently launched its own AI Economic Indicators. They track things like AI adoption, productivity, and of course, jobs. Marketplace’s Meghan McCarty Carino spoke with Connacher Murphy, research manager at Stanford Digital Economy Lab, to learn more about the database and what researchers call the Canary Dashboard for jobs.
