 |
|
|
|||||
A baby and his family dog sit across from each other in a podcast studio.“Welcome to the talking baby podcast,” says the infant, wearing headphones and sounding like a deep-voiced radio broadcaster. “On today’s episode, we’ll be talking to the weird-looking person who lives at my house.”So begins a series of humorous interactions between two characters animated by artificial intelligence that’s attracted millions of views on social media. They’re a nod to the 1989 movie “Look Who’s Talking” but produced in a matter of hours and without a multimillion-dollar Hollywood budget.AI helped do all of that, but it didn’t craft the punch lines. It’s a relief to comedian Jon Lajoie, who made the videos, that AI chatbots just aren’t “inherently funny.”“It can’t write comedy,” said Lajoie. “It can’t do any of that.”For now, at least, they won’t take his job.Lajoie’s viral videos have gained him attention as an AI-adopting entertainer that’s he’s not entirely comfortable with as he grapples with what all this means for the future of his very human craft of making people laugh.King Willonius is not feeling so cautious. His first big hit was an AI-generated song called “BBL Drizzy” that made fun of rapper Drake during the height of his feud with Kendrick Lamar. He’s since moved into making AI video parodies like “I’m McLovin It (Popeye’s Diss Song)” and “I Want My Barrel Back (Cracker Barrel song).”“It’s very similar to somebody who’s writing for The Onion or SNL,” Willonius said. “I try to find out, OK, what’s my comedic angle on this particular topic? And then I’ll generate a video from that.”He starts with writing his own notes on an idea, then refines it with a chatbot, and puts that language known as a prompt into AI tools that can generate imagery, video, music and voices. The key, he says, is to keep iterating.But he wouldn’t just ask it for a joke Willonius says most chatbot-generated comedy lacks the “nuances or complexities that it takes for jokes to really land.”A scholar of comedy, Michelle Robinson, said “a lot of the stuff that I’ve seen AI produce is corny as hell.”“It does seem fluent in the basic grammar of jokes, but sometimes they’re slightly off,” said Robinson, a professor of American studies at the University of North Carolina at Chapel Hill. “They may be moderately funny, but I think they’re really missing an important element of what makes us laugh.”What are they missing? She’s not totally sure, except that most good jokes are a little edgy or dangerous and chatbots can’t seem to calibrate “whatever provocation is in the joke to the moment that we’re living in.”Caleb Warren, a professor who studies marketing and consumer psychology at the University of Arizona, said that leaves comedy writers with an opportunity to make use of tools that can’t completely outsource their skills.“The ideas that are driving the humor are coming from the human comedian,” but the AI tools can help them execute and illustrate them, Warren said.Willonius was a struggling comedian and screenwriter who began experimenting with AI during Hollywood’s actor and writer strikes in 2023.“I leaned all the way into AI because I didn’t know what else to do with my free time,” he said. “I was doing everything I could to try to break into Hollywood. And once the writers’ strike happened, that kind of shut that down. I started to learn these AI tools and get really good at them and started to cultivate an audience.”While Willonius saw an opening, the rise of generative AI has stoked division and posed challenges to other professional comedians.Sarah Silverman joined book authors in suing leading chatbot makers, alleging they infringed the copyright of her “The Bedwetter” memoir. The daughter of the late Robin Williams called it “gross” and “maddening” when users of OpenAI’s AI video generator Sora conjured up realistic “deepfakes” of the beloved actor to churn out what she described as “horrible TikTok slop puppeteering.”“You’re not making art, you’re making disgusting, overprocessed hot dogs out of the lives of human beings, out of the history of art and music, and then shoving them down someone else’s throat hoping they’ll give you a little thumbs-up and like it,” Zelda Williams wrote in October.And the estate of legendary comic George Carlin last year settled a lawsuit against podcasters who purportedly cloned his voice to make a fake hourslong comedy special.Comics have also relished mocking AI tools. A recent “South Park” episode called “Sora Not Sorry” had a bumbling police detective investigate a scourge of fake videos.Lajoie, known for his work on the TV series “The League” and comic songs on YouTube, tried to see what would happen if he asked ChatGPT to help craft a bizarre movie script idea. He said it gave him something “super boring” about “grandma’s dentures and a talking raccoon.”“That level of human creativity, it can’t mimic yet or at least maybe I’m not great at prompting,” he said. Instead, he found it useful to cheaply animate ideas he would otherwise never have pursued such as the talking baby, birds wearing jeans, or a podcasting Jesus Christ interviewing an Easter Bunny who’s never heard of him.The prominent venture capital firm Andreessen Horowitz invited Lajoie and Willonius to exhibit their video creations this fall at a new AI gallery space in Manhattan, part of a promotion of AI creativity tool startups that the firm invests in.Willonius obliged. Lajoie ended up bowing out, after an interview with The Associated Press in which he voiced doubts about what he described as AI’s “Napster phase.” The music-sharing website shuttered in the early 2000s after the record industry and rock band Metallica sued over copyright violations.The investment firm’s co-founder, Marc Andreessen, has been bullish about AI’s potential to bring new life into filmmaking and comedy. On a November podcast, he blamed Hollywood opposition to its adoption on “woke activists (who) have picked up AI as the new thing they’re going to agitate about.” He compared it to resistance to computer graphics in movies before they became commonplace.Lajoie said he shared his early AI video experiments with a few friends who are “anti-AI; real, real, anti-AI” and they were surprised by how well the sketches retained Lajoie’s own comedic voice.He insists he’s no AI expert, just “a creative person who can figure out how to make two characters talk to each other.” But even editing the sketches requires understanding comedic timing, and he has no interest in ceding tht part to a machine.“The thing with comedy is it’s so related to performance, delivery and point of view,” Lajoie said. “Do AIs have a point of view? They can grab a few points of view from different people.”“And when it does have a point of view, I think that’s when we all should be afraid for all of the reasons that the Terminator has taught us,” he said. Matt O’Brien, AP Technology Writer
Category:
E-Commerce
More than a decade ago, Malaysia Airlines Flight MH370 vanished without a trace, sparking one of aviation’s most baffling mysteries.Despite years of multinational searches, investigators still do not know exactly what happened to the plane or its 239 passengers and crew.On Wednesday, Malaysia’s government said American marine robotics company Ocean Infinity would resume a seabed hunt for the missing plane on Dec. 30, reigniting hopes that the plane might finally be found.A massive search in the southern Indian Ocean, where the jet is believed to have gone down, turned up almost nothing. Apart from a few small fragments that washed ashore, no bodies or large wreckage have ever been recovered.Here’s what we know about the deadly aviation tragedy. ‘Good night, Malaysian Three Seven Zero’ The Boeing 777 disappeared from air-traffic radar 39 minutes after departing Kuala Lumpur for Beijing on March 8, 2014.The pilot’s last radio call to Kuala Lumpur “Good night, Malaysian Three Seven Zero” was the final communication before the plane crossed into Vietnamese airspace and failed to check in with controllers there.Minutes later, the plane’s transponder, which broadcasts its location, shut down. Military radar showed the jet turn back over the Andaman Sea, and satellite data suggested it continued flying for hours, possibly until fuel exhaustion, before crashing into a remote section of the southern Indian Ocean.Theories about what happened range from hijacking to cabin depressurization or power failure. There was no distress call, ransom demand, evidence of technical failure or severe weather.Malaysian investigators in 2018 cleared the passengers and crew but did not rule out “unlawful interference.” Authorities have said someone deliberately severed communications and diverted the plane. Passengers came from many countries MH370 carried 227 passengers, including five young children, and 12 crew members. Most passengers were Chinese, but there were also citizens from the United States, Indonesia, France, Russia and elsewhere.Among those aboard were two young Iranians traveling on stolen passports, a group of Chinese calligraphy artists, 20 employees of U.S. tech firm Freescale Semiconductor, a stunt double for actor Jet Li and several families with young children. Many families lost multiple members. Largest underwater search in history Search operations began in the South China Sea between Malaysia and Vietnam, then expanded to the Andaman Sea and the southern Indian Ocean.Australia, Malaysia and China coordinated the largest underwater search in history, covering roughly 120,000 square kilometers (46,000 square miles) of seabed off western Australia. Aircraft, vessels equipped with sonar and robotic submarines scoured the ocean for signs of the plane.Signals thought to be from the plane’s black box turned out to be from other sources, and no wreckage was found. The first confirmed debris was a flaperon discovered on Réunion Island in July 2015, with additional fragments later found along the east coast of Africa. The search was suspended in January 2017.In 2018, U.S. marine robotics company Ocean Infinity resumed the hunt under a “no find, no fee” agreement, focusing on areas identified through debris drift studies, but it ended without success. The challenge of locating remains One reason why such an extensive search failed to turn up clues is that no one knows exactly where to look. The Indian Ocean is the world’s third largest, and the search was conducted in a difficult area, where searchers encountered bad weather and average depths of around 4 kilometers (2.5 miles).It’s not common for planes to disappear in the deep sea, but when they do remains can be very hard to locate. Over the past 50 years, dozens of planes have vanished, according to the Aviation Safety Network. US company resumes search Malaysia’s government gave the green light in March for another “no-find, no-fee” contract with Ocean Infinity to resume the seabed search operation at a new 15,000-square-kilometer (5,800-square-mile) site in the ocean. Ocean Infinity will be paid $70 million only if wreckage is discovered.However, the search was suspended in April due to bad weather. The government said Wednesday that Ocean Infinity will resume search intermittently from Dec. 30 for a total of 55 days, in targeted areas believed to have the highest likelihood of finding the missing aircraft.It is unclear if the company has new evidence of the plane’s location. It has said it would utilize new technology and has worked with many experts to analyze data and narrow the search area to the most likely site.
Category:
E-Commerce
Until recently, when you looked at a house for sale on Zillow, you could see property-specific scores for the risk of flooding, wildfires, wind from storms and hurricanes, extreme heat, and air quality. The numbers came from First Street, a nonprofit that uses peer-reviewed methodologies to calculate “climate risk.” But Zillow recently removed those scores after pressure from CRMLS, one of the large real-estate listing services that supplies its data. The reality is these models have been around for over five years, says Matthew Eby, CEO of First Street, which also provides its data to sites like Realtor.com and Redfin. (Zillow started displaying the information in 2024, but Realtor.com incorporated First Street’s “Flood Scores” in 2020.) And what’s happened is the markets gotten very tight. And now they’re looking for ways to try and make it easier to sell homes at the expense of homebuyers. The California Regional MLS, like others across the country, controls the database that feeds real estate listings to sites like Zillow. The organization said in a statement to the New York Times that it was suspicious after seeing predictions of high flood risk in areas that hadnt flooded in the past. When Fast Company asked for an example of a location, they pointed to a neighborhood in Huntington Beachbut that area actually just flooded last week. In a statement, First Street said that it stands behind the accuracy of its scores. “Our models are built on transparent, peer-reviewed science and are continuously validated against real-world outcomes. In the CRMLS coverage area, during the Los Angeles wildfires, our maps identified over 90% of the homes that ultimately burned as being at severe or extreme riskour highest risk ratingand 100% as having some level of risk, significantly outperforming CalFire’s official state hazard maps. So when claims are made that our models are inaccurate, we ask for evidence. To date, all the empirical validation shows our science is working as designed and providing better risk insight than the tools the industry has relied on historically.” Zillow’s trust in the data has not changed, and that data is important to consumers: In one survey, it saw that more than 80% of buyers considered the data when shopping for a house. But the company said in a statement that it updated its climate risk product experience to adhere to varying MLS requirements. Its not clear exactly what happened: In response to questions for this story, CRMLS now says it only asked Zillow to remove “predictive numbers” and flood map layers on listings, while Zillow says the MLS board voted to demand they block all of the data. Its also not clear what would have happened if Zillow hadnt made any changes, though in theory, the MLS could have stopped giving the site access to its listings. Images of Zillows climate risk tools from a 2024 press release [Image: Zillow] Zillow still links to First Street’s website in each listing, so homebuyers can access the information, but it’s less easy to find. The site also still includes a map that consumers can use to view overall neighborhood risk, if they take the extra step to click on checkboxes for flooding, fire, or other hazards. But the main scores are gone. Obviously, seeing that a particular house has a high flood risk or fire risk can hurt sales. Nevertheless, after First Street first launched, the National Association of Realtors put out guidance saying that the information was usefuland that since realtors aren’t experts in things like flood risk, they shouldn’t try to tell buyers themselves that a particular house is safe, even if it hasn’t flooded in the past. First Street’s flood data goes further than that of the Federal Emergency Management Agency, which uses outdated flood maps. It also incorporates more climate predictions, along with the risk of flooding from heavy rainfall and surface runoff, not just flooding from rivers or the coast. And it includes predictions of small amounts of flooding (for example, whether an inch of water is likely to reach the property). Buyers can dig deeper to figure out how much that amount of flooding might affect a particular house. It’s not surprising that some high risk scores have upset home sellers who haven’t experienced flooding or other problems in the past. But as the climate changes, past experiences don’t guarantee what a property will be like for the next 30 years. Take the example of North Carolina, where some residents hadn’t ever experienced flooding until Hurricane Helene dumped unprecedented rainfall on their neighborhoods. Redfin, another site that uses the data, plans to continue providing it, though sellers have the option to ask for it to be removed from a particular home if they believe it’s inaccurate. (First Street also allows homeowners to ask for their data to be revised if there’s a problem, and then reviews the accuracy.) “Redfin will continue to provide the best-possible estimates of the risks of fires, floods, and storms,” Redfin chief economist Daryl Fairweather said in a statement. “Homebuyers want to know, because losing a home in a catastrophe is heartbreaking, and insuring against these risks is getting more and more expensive.” Realtor.com is working with CRMLS and data providers to look into the issues raised by the MLS over the scores. “We aim to balance transparency about the evolving environmental risks to what is often a familys biggest investment, with an understanding that the available data can sometimes be limited,” the company said in a statement. “For this reason we always encourage consumers to consult a local real estate professional for guidance or to learn more. When issues are raised, we work with our data partners to review them and make updates when appropriate.” If more real estate sites take down the scores, it’s likely that some buyers won’t see the information at all. First Street says that while it’s good that Zillow still includes a link to its site, the mpact is real. “Whenever you add friction into something, it just is used less,” Eby says. “And so not having that information at the tip of your fingers is definitely going to have an impact on the millions of people that go to Zillow every day to see it.”
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||