OpenAI Expands ChatGPT Memory To Draw on Full Conversation History
OpenAI has expanded ChatGPT's memory functionality to include references from all past conversations. The system now builds upon existing saved memories by automatically incorporating previous interactions to deliver more contextually relevant responses for writing, learning, and advisory tasks, the startup said Thursday.
Subscribers can disable the feature through settings or request memory modifications directly in chat. Those already opted out of memory features won't have past-chat references enabled by default. Temporary chats remain available for interactions that users prefer to keep isolated from memory systems. The update is rolling out immediately to Plus and Pro subscribers, excluding users in the EEA, UK, Switzerland, and other European markets.
at Slashdot.
Meta Says Llama 4 Targets Left-Leaning Bias
Meta says in its Llama 4 release announcement that it's specifically addressing "left-leaning" political bias in its AI model, distinguishing this effort from traditional bias concerns around race, gender, and nationality that researchers have long documented. "Our goal is to remove bias from our AI models and to make sure that Llama can understand and articulate both sides of a contentious issue," the company said.
"All leading LLMs have had issues with bias -- specifically, they historically have leaned left," Meta stated, framing AI bias primarily as a political problem. The company claims Llama 4 is "dramatically more balanced" in handling sensitive topics and touts its lack of "strong political lean" compared to competitors.
at Slashdot.
Google Maps is Launching Tools To Help Cities Analyze Infrastructure and Traffic
Google is opening up its Google Maps Platform data so that cities, developers, and other business decision makers can more easily access information about things like infrastructure and traffic. The Verge: Google is integrating new datasets for Google Maps Platform directly into BigQuery, the tech giant's fully managed data analytics service, for the first time. This should make it easier for people to access data from Google Maps platform products, including Imagery Insights, Roads Management Insights, and Places Insights.
at Slashdot.
Fake Job Seekers Are Flooding US Companies
Fake job seekers using AI tools to impersonate candidates are increasingly targeting U.S. companies with remote positions, creating a growing security threat across industries. By 2028, one in four global job applicants will be fake, according to Gartner. These imposters use AI to fabricate photo IDs, generate employment histories, and provide interview answers, often targeting cybersecurity and cryptocurrency firms, CNBC reports.
Once hired, fraudulent employees can install malware to demand ransoms, steal customer data, or simply collect salaries they wouldn't otherwise obtain, according to Vijay Balasubramaniyan, CEO of Pindrop Security. The problem extends beyond tech companies. Last year, the Justice Department alleged more than 300 U.S. firms inadvertently hired impostors with ties to North Korea, including major corporations across various sectors.
at Slashdot.
China's Biotech Advances Threaten US Dominance, Warns Congressional Report
China is moving fast to dominate biotechnology, and the U.S. risks falling behind permanently unless it takes action over the next three years, a congressional commission said. WSJ: Congress should invest at least $15 billion to support biotech research over the next five years and take other steps to bolster manufacturing in the U.S., while barring companies from working with Chinese biotech suppliers, the National Security Commission on Emerging Biotechnology said in a report Tuesday. To achieve its goals, the federal government and U.S.-based researchers will also need to work with allies and partners around the world.
"China is quickly ascending to biotechnology dominance, having made biotechnology a strategic priority for 20 years," the commission said. Without prompt action, the U.S. risks "falling behind, a setback from which we may never recover." The findings convey the depth of worry in Washington that China's rapid biotechnology advances jeopardize U.S. national security. Yet translating the concern into tangible actions could prove challenging.
[...] China plays a large role supplying drug ingredients and even some generic medicines to the U.S. For years, it produced copycat versions of drugs developed in the West. Recent years have seen it become a formidable hub of biotechnology innovation, after the Chinese government gave priority to the field as a critical sector in China's efforts to become a scientific superpower.
at Slashdot.
India's 'Frankenstein' Laptop Economy Thrives Against Planned Obsolescence
In Delhi's Nehru Place and Mumbai's Lamington Road, technicians are creating functional laptops from salvaged parts of multiple discarded devices. These "Frankenstein" machines sell for approximately $110 USD -- a fraction of the $800 price tag for new models. Technicians extract usable components -- motherboards, capacitors, screens, and batteries -- from e-waste sourced locally and from countries like Dubai and China.
"Most people don't care about having the latest model; they just want something that works and won't break the bank," a technician told Verge. This repair ecosystem operates within a larger battle against tech giants pushing planned obsolescence through proprietary designs and restricted parts access. Many technicians source components from Seelampur, India's largest e-waste hub processing 30,000 tonnes daily, though workers there handle toxic materials with minimal protection. "India has always had a repair culture," says Satish Sinha of Toxics Link, "but companies are pushing planned obsolescence, making repairs harder and forcing people to buy new devices."
at Slashdot.
No, the Dire Wolf Has Not Been Brought Back From Extinction
Colossal Biosciences has claimed it "successfully restored" the extinct dire wolf after a "10,000+ year absence," but scientists clarify these are actually genetically modified grey wolves. The U.S. company announced three pups -- males Remus and Romulus born in October, and female Khaleesi born in January -- as dire wolves, but made only 20 genetic edits to grey wolves.
Beth Shapiro of Colossal told New Scientist that just 15 modifications were based on dire wolf DNA, primarily targeting size, musculature and ear shape. Five other changes involve mutations known to produce light coats in grey wolves. A 2021 DNA study revealed dire wolves and grey wolves last shared a common ancestor about 6 million years ago, with jackals and African wild dogs more closely related to grey wolves.
at Slashdot.
Apple Rushes Shipments From India To Dodge Tariffs
Apple rushed five planeloads of iPhones from India to the U.S. in just three days to beat new tariffs imposed by the Trump administration, Times of India reported Monday, citing sources. The urgent shipments during the final week of March aimed to avoid the 10% reciprocal tariff that took effect on April 5.
The stockpiling will allow Apple to maintain current pricing temporarily. "The reserves that arrived at lower duty will temporarily insulate the company from the higher prices that it will need to pay for new shipments," the Indian daily cited a source as saying. The Trump administration also announced a 26% reciprocal tariff to be implemented on April 9, potentially accelerating Apple's manufacturing shift away from China. India offers a significant tariff advantage, with Indian exports facing a 26% tariff to the U.S. compared to 54% on Chinese goods.
Further reading: India's Economic Chess Against Twin US Economic Threats.
at Slashdot.
America's NIH Scientists Have a Cancer Breakthrough. Layoffs are Delaying It.
Scientists "demonstrated a promising step toward using a person's own immune cells to fight gastrointestinal cancers" at America's National Institutes of Health (or NIH), reports the Washington Post.
But the results were published in Nature Medicine on Tuesday — "the same day the agency was hit with devastating layoffs..."
The treatment approach is still early in its development; the personalized immunotherapy regimen shrank tumors in only about a quarter of the patients with colon, rectal and other GI cancers enrolled in a clinical trial. But a researcher who was not involved in the study called the results "remarkable" because they highlight a path to a frustratingly elusive goal in medicine — harnessing a person's own immune defenses to target common solid tumor cancers. Until now, cell-based immunotherapy has worked mainly on blood cancers, such as leukemia, but not the solid cancers that seed tumors in the breast, brain, lungs, pancreas and GI tract...
But the progress arrives at a sad time for science — and for patients, said the leader of the work, NIH immunotherapy pioneer Steven Rosenberg. Two patients' treatments using the experimental therapy had to be delayed because NIH's capacity to make personalized cell therapies has been slowed by the firing of highly skilled staff and by purchasing slowdowns. Those occurred even before major layoffs took place Tuesday... The Department of Health and Human Services (HHS) responded to an email asking about clinical trial delays with a statement: "NIH and HHS are complying with President Trump's executive order."
It's "a very exciting study," said Patrick Hwu, president of the Moffitt Cancer Center in Tampa.
Finding ways to tailor this cell-based immunotherapy approach to common solid tumors that cause the vast majority of cancer deaths has remained a major scientific challenge... Rosenberg and colleagues first tried to create tumor infiltrating lymphocytes [or "TILs"] using the method that worked in melanoma for 18 patients with GI cancers that had spread. It failed completely. In a second iteration, his team sequenced the mutations present in each patient's tumor and used that information to sift out and expand the TILs that could home in on that patient's specific tumor cells. The results were far from a triumph, but provided a clue — this time, three of 39 patients' tumors shrank. In the last stage of the trial, the scientists added a drug called pembrolizumab that takes the brakes off immune cells. This time, eight of the 34 patients responded.
"Right now, only a few labs in the country can do what they just did," Hwu said.
While Rosenberg is already working "to refine and improve upon the results," he told the Post that two scientists involved in the specialized process of preparing the cells to treat patients were fired in the probationary purge. "We've had to slow down our work and delay the treatment of some patients...."
And there's also dramatically fewer people now who can purchase research materials, which the Post says it "making it slower and more difficult to obtain supplies."
at Slashdot.
The Tumblr Revival is Real - and Gen Z is Leading the Charge
"Gen Z is rediscovering Tumblr — a chaotic, cozy corner of the internet untouched by algorithmic gloss and influencer overload..." writes Fast Company, "embracing the platform as a refuge from an internet saturated with influencers and algorithm fatigue."
Thanks to Gen Z, the site has found new life. As of 2025, Gen Z makes up 50% of Tumblr's active monthly users and accounts for 60% of new sign-ups, according to data shared with Business Insider's Amanda Hoover, who recently reported on the platform's resurgence. User numbers spiked in January during the near-ban of TikTok and jumped again last year when Brazil temporarily banned X. In response, Tumblr users launched dedicated communities to archive and share their favorite TikToks...
To keep up with the momentum, Tumblr introduced Reddit-style Communities in December, letting users connect over shared interests like photography and video games. In January, it debuted Tumblr TV — a TikTok-like feature that serves as both a GIF search engine and a short-form video platform. But perhaps Tumblr's greatest strength is that it isn't TikTok or Facebook. Currently the 10th most popular social platform in the U.S., according to analytics firm Similarweb, Tumblr is dwarfed by giants like Instagram and X. For its users, though, that's part of the appeal.
First launched in 2007, Tumblr peaked at over 100 million users in 2014, according to the article. Trends like Occupy Wall Street had been born on Tumblr, notes Business Insider, calling the blogging platform "Gen Z's safe space... as the rest of the social internet has become increasingly commodified, polarized, and dominated by lifestyle influencers." Tumblr was also "one of the most hyped startups in the world before fading into obsolescence — bought by Yahoo for $1.1 billion in 2013... then acquired by Verizon, and later offloaded for fractions of pennies on the dollar in a distressed sale.
"That same Tumblr, a relic of many millennials' formative years, has been having a moment among Gen Z..."
"Gen Z has this romanticism of the early-2000s internet," says Amanda Brennan, an internet librarian who worked at Tumblr for seven years, leaving her role as head of content in 2021... Part of the reason young people are hanging out on old social platforms is that there's nowhere new to go. The tech industry is evolving at a slower pace than it was in the 2000s, and there's less room for disruption. Big Tech has a stranglehold on how we socialize. That leaves Gen Z to pick up the scraps left by the early online millennials and attempt to craft them into something relevant. They love Pinterest (founded in 2010) and Snapchat (2011), and they're trying out digital point-and-shoot cameras and flip phones for an early-2000s aesthetic — and learning the valuable lesson that sometimes we look better when blurrier.
More Gen Zers and millennials are signing up for Yahoo. Napster, surprising many people with its continued existence, just sold for $207 million. The trend is fueled by nostalgia for Y2K aesthetics and a longing for a time when people could make mistakes on the internet and move past them. The pandemic also brought more Gen Z users to Tumblr...
And Tumblr still works much like an older internet, where people have more control over what they see and rely less on algorithms. "You curate your own stuff; it takes a little bit of work to put everything in place, but when it's working, you see the content you want to see," Fjodor Everaerts, a 26-year-old in Belgium who has made some 250,000 posts since he joined Tumblr when he was 14... Under Automattic, Tumblr is finally in the home that serves it, [says Ari Levine, the head of brand partnerships at Tumblr]. "We've had ups and downs along the way, but we're in the most interesting position and place that we've been in 18 years," he says... And following media companies (including Business Insider) and social platforms like Reddit, Automattic in 2024 was making a deal with OpenAI and Midjourney to allow the systems to train on Tumblr posts.
"The social internet is fractured," the article argues. ("Millennials are running Reddit. Gen Xers and Baby Boomers have a home on Facebook. Bluesky, one of the new X alternatives, has a tangible elder-millennial/Gen X vibe. Gen Zers have created social apps like BeReal and the Myspace-inspired Noplace, but they've so far generated more hype than influence....")
But in a world where megaplatforms "flatten our online experiences and reward content that fits a mold," the article suggests, "smaller communities can enrich them."
at Slashdot.
Microsoft's Copilot Roasts Bill Gates, Steve Ballmer, and Satya Nadella
"Three Microsoft CEOs walk into a room..." Satya Nadella posted on X.com, "and are interviewed by Copilot!"
As part of Microsoft's 50th anniversary celebration, Microsoft CEO Nadella joined former Microsoft CEOs Bill Gates and Steve Ballmer in a funny two-minute video (also available on YouTube) in which Nadella asks Copilot for an AI-powered roast of each of them.
"All right Satya," Copilot answers. "Let's spice it up..."
Copilot: Steve, we'll start with you. Your enthusiasm is legendary. But do you ever worry that your energy might short-circuit the AI? I mean, even robots need a coffee break after your pep talks.
Steve Ballmer: Come on dude! If you can't handle me, you can't handle a huge percentage of the world's population! We're counting on you!
Copilot: Ha ha, fair point Steve. Now Bill, you've got that signature thoughtful gaze. But do you ever think the AI might feel intimidated by your intense thinking face? I mean, it's like it's waiting for a blue screen moment.
Bill Gates: I hope so. I mean, it's all I've got left as the AIs get so damn intelligent. It's just that gaze and my willingness to criticize that differentiates me.
Copilot: And it's a formidable gaze indeed, Bill.
Steve Ballmer: Copilot, would you ever be willing to criticize him?
Copilot: Absolutely. Criticism is where growth happens. Now Satya, AI seems to be your best buddy — but ever worry that your deep passion for AI could one day have it running the show, leaving you to just enjoy the ride?
Satya Nadella: Well, the day AI can play like the best cricket player that I enjoy, that's the day it can run the show, as far as I'm concerned.
Copilot: Touche, Satya. Let's see if it can hit a century in cricket and then we'll talk.
This has been a blast, gentlemen. Here's to another 50 years of innovation and adventure. Cheers.
Ballmer, Nadella, and Gates: Cheers.
Bill Gates: Whatever that means to an AI. Does it drink?
at Slashdot.
Could We Reach Mars Faster With Nuclear Fusion-Powered Rockets?
Nuclear fusion — which releases four times the energy of fission — could theoretically happen sooner in space than on earth, reports CNN.
"And it could help spacecraft achieve speeds of up to 500,000 miles (805,000 kilometers) per hour — more than the fastest object ever built..."
With funding from the UK Space Agency, British startup Pulsar Fusion has unveiled Sunbird, a space rocket concept designed to meet spacecraft in orbit, attach to them, and carry them to their destination at breakneck speed using nuclear fusion... For now, Sunbird is in the very early stages of construction and it has exceptional engineering challenges to overcome, but Pulsar says it hopes to achieve fusion in orbit for the first time in 2027. [Pulsar's founder/CEO says the first functional Sunbird would be ready four to five years later.]
If the rocket ever becomes operational, it could one day cut the journey time of a potential mission to Mars in half.
CNN says the proposed Sunbird process would use helium-3 — which may be abundant on the Moon — to generate protons which "can be used as a 'nuclear exhaust' to provide propulsion". (And without generating any dangerous radioactive material.)
"It's very unnatural to do fusion on Earth," says Richard Dinan, founder and CEO of Pulsar. "Fusion doesn't want to work in an atmosphere. Space is a far more logical, sensible place to do fusion, because that's where it wants to happen anyway...."
Sunbirds would operate similarly to city bikes at docking stations, according to Dinan: "We launch them into space, and we would have a charging station where they could sit and then meet your ship," he says. "You turn off your inefficient combustion engines, and use nuclear fusion for the greater part of your journey. Ideally, you'd have a station somewhere near Mars, and you'd have a station on low Earth orbit, and the (Sunbirds) would just go back and forth...." Initially, the Sunbirds will be offered for shuttling satellites in orbit, but their true potential would come into play with interplanetary missions. The company illustrates a few examples of the missions that Sunbird could unlock, such as delivering up to 2,000 kilograms (4,400 pounds) of cargo to Mars in under six months, deploying probes to Jupiter or Saturn in two to four years (NASA's Europa Clipper, launched in 2024 towards one of Jupiter's moons, will arrive after 5.5 years), and an asteroid mining mission that would complete a round trip to a near-Earth asteroid in one to two years instead of three.
Other companies are working on nuclear fusion engines for space propulsion, including Pasadena-based Helicity Space, which received investment from aerospace giant Lockheed Martin in 2024. San Diego-based General Atomics and NASA are working on another type of nuclear reactor — based on fission rather than fusion — which they plan to test in space in 2027.
at Slashdot.
'Tron' Sequel Trailer Released by Disney
This October will see the release of a film that's nearly 43 years in the making, reports Ars Technica:
It's difficult to underestimate the massive influence that Disney's 1982 cult science fiction film, TRON, had on both the film industry — thanks to combining live action with what were then groundbreaking visual effects rife with computer-generated imagery — and on nerd culture at large. Over the ensuing decades there has been one sequel, an animated TV series, a comic book miniseries, video games, and theme park attractions, all modeled on director Steve Lisberg's original fictional world.
Now we're getting a third installment in the film franchise: TRON: Ares, directed by Joachim Rønning (Pirates of the Caribbean: Dead Men Tell No Tales, Maleficent: Mistress of Evil), that serves as a standalone sequel to 2010's TRON: Legacy. Disney just released the first trailer and poster art, and while the footage is short on plot, it's got the show-stopping visuals we've come to expect from all things TRON.
The film's director says it "builds upon the legacy of cutting-edge design, technology and storytelling, according to an official statement from Disney. And here's how they describe the plot. "TRON: Ares follows a highly sophisticated Program, Ares, who is sent from the digital world into the real world on a dangerous mission, marking humankind's first encounter with A.I. beings."
Share your thoughts in the comments. (Anyone remember playing the Tron videogame?)
The first episode of 2012's animated Tron: Uprising is available on the Disney XD YouTube channel...
at Slashdot.
Microsoft Uses AI To Find Flaws In GRUB2, U-Boot, Barebox Bootloaders
Slashdot reader zlives shared this report from BleepingComputer:
Microsoft used its AI-powered Security Copilot to discover 20 previously unknown vulnerabilities in the GRUB2, U-Boot, and Barebox open-source bootloaders.
GRUB2 (GRand Unified Bootloader) is the default boot loader for most Linux distributions, including Ubuntu, while U-Boot and Barebox are commonly used in embedded and IoT devices. Microsoft discovered eleven vulnerabilities in GRUB2, including integer and buffer overflows in filesystem parsers, command flaws, and a side-channel in cryptographic comparison. Additionally, 9 buffer overflows in parsing SquashFS, EXT4, CramFS, JFFS2, and symlinks were discovered in U-Boot and Barebox, which require physical access to exploit.
The newly discovered flaws impact devices relying on UEFI Secure Boot, and if the right conditions are met, attackers can bypass security protections to execute arbitrary code on the device. While exploiting these flaws would likely need local access to devices, previous bootkit attacks like BlackLotus achieved this through malware infections.
Miccrosoft titled its blog post "Analyzing open-source bootloaders: Finding vulnerabilities faster with AI." (And they do note that Micxrosoft disclosed the discovered vulnerabilities to the GRUB2, U-boot, and Barebox maintainers and "worked with the GRUB2 maintainers to contribute fixes... GRUB2 maintainers released security updates on February 18, 2025, and both the U-boot and Barebox maintainers released updates on February 19, 2025.")
They add that performing their initial research, using Security Copilot "saved our team approximately a week's worth of time," Microsoft writes, "that would have otherwise been spent manually reviewing the content."
Through a series of prompts, we identified and refined security issues, ultimately uncovering an exploitable integer overflow vulnerability. Copilot also assisted in finding similar patterns in other files, ensuring comprehensive coverage and validation of our findings...
As AI continues to emerge as a key tool in the cybersecurity community, Microsoft emphasizes the importance of vendors and researchers maintaining their focus on information sharing. This approach ensures that AI's advantages in rapid vulnerability discovery, remediation, and accelerated security operations can effectively counter malicious actors' attempts to use AI to scale common attack tactics, techniques, and procedures (TTPs).
This week Google also announced Sec-Gemini v1, "a new experimental AI model focused on advancing cybersecurity AI frontiers."
at Slashdot.
Open Source Coalition Announces 'Model-Signing' with Sigstore to Strengthen the ML Supply Chain
The advent of LLMs and machine learning-based applications "opened the door to a new wave of security threats," argues Google's security blog. (Including model and data poisoning, prompt injection, prompt leaking and prompt evasion.)
So as part of the Linux Foundation's nonprofit Open Source Security Foundation, and in partnership with NVIDIA and HiddenLayer, Google's Open Source Security Team on Friday announced the first stable model-signing library (hosted at PyPI.org), with digital signatures letting users verify that the model used by their application "is exactly the model that was created by the developers," according to a post on Google's security blog.
[S]ince models are an uninspectable collection of weights (sometimes also with arbitrary code), an attacker can tamper with them and achieve significant impact to those using the models. Users, developers, and practitioners need to examine an important question during their risk assessment process: "can I trust this model?"
Since its launch, Google's Secure AI Framework (SAIF) has created guidance and technical solutions for creating AI applications that users can trust. A first step in achieving trust in the model is to permit users to verify its integrity and provenance, to prevent tampering across all processes from training to usage, via cryptographic signing... [T]he signature would have to be verified when the model gets uploaded to a model hub, when the model gets selected to be deployed into an application (embedded or via remote APIs) and when the model is used as an intermediary during another training run. Assuming the training infrastructure is trustworthy and not compromised, this approach guarantees that each model user can trust the model...
The average developer, however, would not want to manage keys and rotate them on compromise. These challenges are addressed by using Sigstore, a collection of tools and services that make code signing secure and easy. By binding an OpenID Connect token to a workload or developer identity, Sigstore alleviates the need to manage or rotate long-lived secrets. Furthermore, signing is made transparent so signatures over malicious artifacts could be audited in a public transparency log, by anyone. This ensures that split-view attacks are not possible, so any user would get the exact same model. These features are why we recommend Sigstore's signing mechanism as the default approach for signing ML models.
Today the OSS community is releasing the v1.0 stable version of our model signing library as a Python package supporting Sigstore and traditional signing methods. This model signing library is specialized to handle the sheer scale of ML models (which are usually much larger than traditional software components), and handles signing models represented as a directory tree. The package provides CLI utilities so that users can sign and verify model signatures for individual models. The package can also be used as a library which we plan to incorporate directly into model hub upload flows as well as into ML frameworks.
"We can view model signing as establishing the foundation of trust in the ML ecosystem..." the post concludes (adding "We envision extending this approach to also include datasets and other ML-related artifacts.")
Then, we plan to build on top of signatures, towards fully tamper-proof metadata records, that can be read by both humans and machines. This has the potential to automate a significant fraction of the work needed to perform incident response in case of a compromise in the ML world...
To shape the future of building tamper-proof ML, join the Coalition for Secure AI, where we are planning to work on building the entire trust ecosystem together with the open source community. In collaboration with multiple industry partners, we are starting up a special interest group under CoSAI for defining the future of ML signing and including tamper-proof ML metadata, such as model cards and evaluation results.
at Slashdot.
A Busy Hurricane Season is Expected. Here's How It Will Be Different From the Last
An anonymous reader shares a report: Yet another busy hurricane season is likely across the Atlantic this year -- but some of the conditions that supercharged storms like Hurricanes Helene and Milton in 2024 have waned, according to a key forecast issued Thursday.
A warm -- yet no longer record-hot -- strip of waters across the Atlantic Ocean is forecast to help fuel development of 17 named tropical cyclones during the season that runs from June 1 through Nov. 30, according to Colorado State University researchers. Of those tropical cyclones, nine are forecast to become hurricanes, with four of those expected to reach "major" hurricane strength.
That would mean a few more tropical storms and hurricanes than in an average year, yet slightly quieter conditions than those observed across the Atlantic basin last year. This time last year, researchers from CSU were warning of an "extremely active" hurricane season with nearly two dozen named tropical storms. The next month, the National Oceanic and Atmospheric Administration released an aggressive forecast, warning the United States could face one of its worst hurricane seasons in two decades.
The forecast out Thursday underscores how warming oceans and cyclical patterns in storm activity have primed the Atlantic basin for what is now a decades-long string of frequent, above-normal -- but not necessarily hyperactive -- seasons, said Philip Klotzbach, a senior research scientist at Colorado State and the forecast's lead author.
at Slashdot.
Two Teenagers Built 'Cal AI', a Photo Calorie App With Over a Million Users
An anonymous reader quotes a report from TechCrunch: In a world filled with "vibe coding," Zach Yadegari, teen founder of Cal AI, stands in ironic, old-fashioned contrast. Ironic because Yadegari and his co-founder, Henry Langmack, are both just 18 years old and still in high school. Yet their story, so far, is a classic. Launched in May, Cal AI has generated over 5 million downloads in eight months, Yadegari says. Better still, he tells TechCrunch that the customer retention rate is over 30% and that the app generated over $2 million in revenue last month. [...]
The concept is simple: Take a picture of the food you are about to consume, and let the app log calories and macros for you. It's not a unique idea. For instance, the big dog in calorie counting, MyFitnessPal, has its Meal Scan feature. Then there are apps like SnapCalorie, which was released in 2023 and created by the founder of Google Lens. Cal AI's advantage, perhaps, is that it was built wholly in the age of large image models. It uses models from Anthropic and OpenAI and RAG to improve accuracy and is trained on open source food calorie and image databases from sites like GitHub.
"We have found that different models are better with different foods," Yadegari tells TechCrunch. Along the way, the founders coded through technical problems like recognizing ingredients from food packages or in jumbled bowls. The result is an app that the creators say is 90% accurate, which appears to be good enough for many dieters. The report says Yadegari began mastering Python and C# in middle school and went on to build his first business in ninth grade -- a website called Totally Science that gave students access to unblocked games (cleverly named to evade school filters). He sold the company at age 16 to FreezeNova for $100,000.
Following the sale, Yadegari immersed himself in the startup scene, watching Y Combinator videos and networking on X, where he met co-founder Blake Anderson, known for creating ChatGPT-powered apps like RizzGPT. Together, they launched Cal AI and moved to a hacker house in San Francisco to develop their prototype.
at Slashdot.
Vibe Coded AI App Generates Recipes With Very Few Guardrails
An anonymous reader quotes a report from 404 Media: A "vibe coded" AI app developed by entrepreneur and Y Combinator group partner Tom Blomfield has generated recipes that gave users instruction on how to make "Cyanide Ice Cream," "Thick White Cum Soup," and "Uranium Bomb," using those actual substances as ingredients. Vibe coding, in case you are unfamiliar, is the new practice where people, some with limited coding experience, rapidly develop software with AI assisted coding tools without overthinking how efficient the code is as long as it's functional. This is how Blomfield said he made RecipeNinja.AI. [...] The recipe for Cyanide Ice Cream was still live on RecipeNinja.AI at the time of writing, as are recipes for Platypus Milk Cream Soup, Werewolf Cream Glazing, Cholera-Inspired Chocolate Cake, and other nonsense. Other recipes for things people shouldn't eat have been removed.
It also appears that Blomfield has introduced content moderation since users discovered they could generate dangerous or extremely stupid recipes. I wasn't able to generate recipes for asbestos cake, bullet tacos, or glue pizza. I was able to generate a recipe for "very dry tacos," which looks not very good but not dangerous. In a March 20 blog on his personal site, Blomfield explained that he's a startup founder turned investor, and while he has experience with PHP and Ruby on Rails, he has not written a line of code professionally since 2015. "In my day job at Y Combinator, I'm around founders who are building amazing stuff with AI every day and I kept hearing about the advances in tools like Lovable, Cursor and Windsurf," he wrote, referring to AI-assisted coding tools. "I love building stuff and I've always got a list of little apps I want to build if I had more free time."
After playing around with them, he wrote, he decided to build RecipeNinja.AI, which can take a prompt as simple as "Lasagna," and generate an image of the finished dish along with a step-by-stape recipe which can use ElevenLabs's AI generated voice to narrate the instruction so the user doesn't have to interact with a device with his tomato sauce-covered fingers. "I was pretty astonished that Windsurf managed to integrate both the OpenAI and Elevenlabs APIs without me doing very much at all," Blomfield wrote. "After we had a couple of problems with the open AI Ruby library, it quickly fell back to a raw ruby HTTP client implementation, but I honestly didn't care. As long as it worked, I didn't really mind if it used 20 lines of code or two lines of code." Having some kind of voice controlled recipe app sounds like a pretty good idea to me, and it's impressive that Blomfield was able to get something up and running so fast given his limited coding experience. But the problem is that he also allowed users to generate their own recipes with seemingly very few guardrails on what kind of recipes are and are not allowed, and that the site kept those results and showed them to other users.
at Slashdot.
Amazon Said To Make a Bid To Buy TikTok in the US
An anonymous reader shares a report: Amazon has put in a last-minute bid to acquire all of TikTok, the popular video app, as it approaches an April deadline to be separated from its Chinese owner or face a ban in the United States, according to three people familiar with the bid.
Various parties who have been involved in the talks do not appear to be taking Amazon's bid seriously, the people said. The bid came via an offer letter addressed to Vice President JD Vance and Howard Lutnick, the commerce secretary, according to a person briefed on the matter. Amazon's bid highlights the 11th-hour maneuvering in Washington over TikTok's ownership. Policymakers in both parties have expressed deep national security concerns over the app's Chinese ownership, and passed a law last year to force a sale of TikTok that was set to take effect in January.
at Slashdot.
Cheap TVs' Incessant Advertising Reaches Troubling New Lows
An anonymous reader quotes an op-ed from Ars Technica's Scharon Harding: TVs offer us an escape from the real world. After a long day, sometimes there's nothing more relaxing than turning on your TV, tuning into your favorite program, and unplugging from the realities around you. But what happens when divisive, potentially offensive messaging infiltrates that escape? Even with streaming services making it easy to watch TV commercial-free, it can still be difficult for TV viewers to avoid ads with these sorts of messages. That's especially the case with budget brands, which may even force controversial ads onto TVs when they're idle, making users pay for low-priced TVs in unexpected, and sometimes troubling, ways. [...]
Buying a budget TV means accepting some trade-offs. Those trade-offs have historically been around things like image quality and feature sets. But companies like Vizio are also asking customers to accept questionable advertising decisions as they look to create new paths to ad revenue. Numerous factors are pushing TV OS operators deeper into advertising. Brands are struggling to grow profits as people buy new TVs less frequently. As the TV market gets more competitive, hardware is also selling for cheaper, with some companies selling TVs at a loss with hopes of making up for it with ad sales. There's concern that these market realities could detract from real TV innovation. And as the Secretary Noem ad reportedly shown to Vizio TV owners has highlighted, another concern is the lack of care around which ads are being shown to TV owners -- especially when all they want is simple "ambient background" noise.
Today, people can disable ambient mode settings that show ads. But with some TV brands showing poor judgment around where they sell and place ads, we wouldn't bank on companies maintaining these boundaries forever. If the industry can't find a way to balance corporate needs with appropriate advertising, people might turn off not only their TVs more often, but also unplug from those brands completely. Some of the worst offenders highlighted in the article include Vizio TVs' "Scenic Mode," which activates when the TV is idle and displays "relaxing, ambient content" accompanied by ads. Roku City takes a similar approach with its animated cityscape screensaver, saturated with brand logos and advertisements. Even Amazon Fire TV and premium brands like LG have adopted screensaver ads, showing that this intrusive trend isn't limited to budget models.
at Slashdot.