Hi, AI • Tech News

💻 A "fly brain" is playing Doom and trading crypto. What's going on? On September 3, a team of scientists from Google, HHMI Janelia, and other institutions published MaleCNS, the complete connectome of the male fruit fly. It maps the entire central nervous system, from the brain to the ventral nerve cord, including all 166,700 neurons and 125 million synapses. Just one day later, Georgia Tech graduate student Evan Sinclair Smith hooked MaleCNS up to Minecraft. Events in the game are fed into the fly's sensory neurons, while signals from the network's output neurons are used to control the character. The virtual fly searches for food, responds to light, and dodges other players. The video went viral, and soon people had the "digital fruit fly" playing Doom and Beat Saber, driving cars, and even trading cryptocurrency. But there's a major caveat: almost everything in these demos depends on the developer's choices. They decide which game events map to which sensory neurons and which output neurons correspond to particular controls. The movements themselves are pre-programmed; activity in the neural network simply determines which one is triggered. Meanwhile, Reddit users began debatin

🔥 The Smell Of Chocolate Made People More Resilient In an experiment involving 23 young men, the scent of dark chocolate helped participants complete about 18 more leg extensions than they did without it. At the same time, the participants themselves didn't feel like they were working out any harder. Before the experiment, participants fasted for at least 10 hours. Each person completed three workouts several days apart and, before each set, smelled one of the samples: melted dark chocolate with 90% cocoa, milk chocolate with 60% cocoa, or odorless "control" water. Milk chocolate also had an effect, though a weaker one: on average, it added about nine repetitions. ⚙️ Researchers suggest that familiar food smells may trigger brain responses related to eating in advance. In the case of dark chocolate, participants reported feeling less hungry and more full after smelling it, which may also have affected their motivation during exercise. For now, this is only a hypothesis: the sample size was small, and the study did not measure hormones or brain activity. Does chocolate help you work out? ❤️ — Yes, I've known that for a long time! 🔥 — I'll have to try it @hiaimediaen

⏳ AI Trained to Predict Death in ICU Patients Researchers have proposed an unusual way to combine medical data to predict when a patient may be nearing death. 📌 The authors collected data from 54,000 adult intensive care unit patients, including vital signs, lab results, ECGs, chest X-rays, and doctors' notes. The authors converted all information, including text, into images and then compressed it into a single "visual memory" of the patient—a vector of 1,024 numbers. Based on this representation, the AI model attempted to predict how the patient's condition would change over the next 3 to 48 hours. The system performed best at predicting death within 48 hours, achieving a score of 0.852 (where 0.5 represents random guessing and 1 represents a perfect prediction). The AI could also identify the risk of acute kidney injury and organ dysfunction in advance, as well as predict ICU discharge. 🔍 However, the value of the study is not simply that AI can accurately predict death—a conventional statistical model delivers similar accuracy. Instead, the researchers demonstrated that a single AI model can process radically different types of data at the same time. In theory, you could add ev

🤖 Teenager with a Terminal Illness Creates an AI Twin 17-year-old Chen Zi-kin from Hong Kong lives with Duchenne muscular dystrophy—a rare genetic disorder that leads to a gradual loss of muscle strength. On average, people with this diagnosis have a life expectancy of about 25 years. Over the past year, Chen Tsz-kin's condition has deteriorated significantly. A year and a half ago, the boy began working with the Hong Kong-based company Digital Domain to develop his digital twin—AI Tsz-kin. The avatar was created using 400 photos of his face and voice recordings over six weeks. "Life may come to an end, but through different means, AI can allow my message to continue. Even if one day I am no longer here, I just hope people remember the spirit behind it," Chen said at the project's presentation. The boy got the diagnosis when he was just a year and a half old. He now uses a wheelchair and sleeps with a breathing machine. Nevertheless, Chen has led an active lifestyle all these years: he paraglides and dives, climbs mountains, and has completed a 100-kilometer expedition across the desert. 💡 The creators of Chen Zi-kin's AI avatar insist that they are not trying to replace the boy, b

👍 Robotic Hand Learns to Write in 18 Seconds Researchers at ETH Zurich taught the ORCA robotic hand to write with a pen without thousands of training examples or human demonstrations. At first, the robot spends about 18 seconds moving its fingers and observing how those movements affect the pen tip. After that, it can write letters and draw shapes while continuously adjusting its movements in real time. The system does not try to calculate the complex physics of contact between the pen and the fingers in advance. A camera tracks the pen tip, while an algorithm determines in real time how the hand's joint movements affect the pen. ⚙️ Writing is a demonstration of the technology. The authors present their method as an alternative to reinforcement learning and imitation learning, which require hundreds of hours of data collection and vastly greater computing power. The entire algorithm, including computer vision, ran on a standard laptop CPU. This approach could therefore be well suited for applications such as controlling prosthetic devices or small robots. @hiaimediaen

💰 "What Better Way to Create Demand Than to Create a Problem?" The AI industry itself is partly fueling the hype around AI cybersecurity threats as it prepares to launch new products, Nvidia CEO Jensen Huang says. "And what better way to create demand than to create a problem? Who doesn't want their market to be hysterical about their product and line up around the corner for it?" Huang said at a Goldman Sachs conference. In July, Huang called both the idea that AI could bring about the end of humanity and predictions that it could wipe out half of U.S. jobs "complete nonsense." ✈️ But his comments are particularly interesting now. Over the weekend, Anthropic CEO Dario Amodei called on AI companies to slow the development of their most powerful models and allow independent auditors to evaluate even closed models. OpenAI CEO Sam Altman and Elon Musk backed the proposal. ⚙️ Markets fell amid the debate. Investors fear that slowing the AI race could weaken demand for chips and reduce investment in data centers. Nvidia shares fell more than 3%, AMD dropped 4.6%, and Intel lost nearly 5%. Should we slow down the AI race? ❤️ — Yes, we're losing control of AI! 🔥 — No, it's just marketing.

👨‍💻 AI Can Detect Its Mistakes From Your Brain Signals Mutual understanding is crucial for human-robot collaboration, yet smart systems mostly rely on visual cues and direct commands and often get things wrong. The challenge is that if a robot sees its human partner reaching for a cup, the AI must first grasp the underlying intent: does the person want a drink, clear the cup from the table, or pass it to someone else? Once the AI identifies thirst as the motive, it must then determine how to fulfill the request: what to fill the cup with and where to get it. 💡 Researchers from KAIST and Microsoft Research Asia recorded EEG signals from 12 volunteers while they watched an AI agent perform different tasks. The analysis revealed distinct brain responses depending on whether things went according to plan, the AI misunderstood the goal, or it chose the wrong way to accomplish it. The researchers then used these brain signals to guide a virtual robot. The system was able to adapt even when the EEG readings weren't perfect or the person's goal suddenly changed. ❗️ For now, it's only an experiment. But the researchers believe EEG-based systems for interpreting human intentions could eventu

🔄 Russian Mathematicians Cut AI Computing Costs 20x Startup Mostik AI has developed a way for AI models to communicate directly by sharing their internal states rather than passing text back and forth. The approach lets a large model do the computationally intensive work of processing a task, while a smaller, cheaper model generates the response. ⚙️ Mostik AI CEO Sasha Malysheva told @hiaimediaen how it works: Multi-model systems are already used in AI agents to distribute work between models with different capabilities and costs. One model might plan a series of actions, for example, while another executes them. But they still typically communicate through text—one model generates an output, and the next one reads it. When processing each token, a model builds numerical vectors that make up its hidden state. This internal state contains the context, information about how the model understands the task, and what a possible answer might look like. According to Mostik, producing a single token involves more than 100 hidden vectors, or about 1 million numbers—roughly 2 MB of internal state. Yet only one selected token leaves the model, equivalent to about 17 bits. "Systems built to th

❤️ One in Two People Consider Their Partner's "Relationship" with AI to Be Cheating These are the findings revealed by AI Girlfriend Coach research project, which surveyed more than 1,700 American adults. One-third of respondents consider any intimate interactions with AI to be cheating. Roughly one in six believes that being unfaithful specifically refers to attachment (emotional or sexual) to a bot. One in five does not consider such interactions to be cheating, but would still be uncomfortable with it. 💡 Men and women enter into "relationships" with AI in roughly equal proportions. 44.7% of Millennials, 31.6% of Generation X, and just 12.5% of Gen Z are willing to have an AI partner. ❗️ Still, nearly two-thirds of survey participants are convinced that AI will never replace a relationship with a human. Is a romance with AI cheating? ❤️ — No, it's just a bot 🔥 — Yes, definitely! @hiaimediaen

💰 Micron Is Paying Staff Up to 68 Months of Salary Micron will pay its Taiwan production workers annual rewards worth 35 to 68 months of salary. An entry-level engineer will average about $108K for the year, including equity. ℹ️ Micron holds roughly a quarter of the global DRAM market and about 15% of NAND. In Q3 of fiscal 2026, revenue jumped 4.5x year over year to $41.5B and net income rose almost 15x, to $28.2B. The payout comes amid a standoff with the unions, which had threatened to strike and demanded a one-time bonus equal to about 83 months of salary. ⏫ The unions also want Micron to channel 15% of operating profit into bonuses starting next fiscal year and to pay them quarterly. No agreement has been reached on those demands. What would you take? ❤️ — Take the 68 months 🔥 — Push all the way, we need 83 🤣 — One month is plenty for me @hiaimediaen

❤️ Gen Z Don't Want to Date AI Builders? These are the findings of a recent survey by the American dating app Hily, in which about 3,000 Gen Z and Millennials participated. One in three Gen Z and one in five Millennial daters said their interest in someone would decrease if that person was building AI tools that replace human jobs. The opposite effect is significantly less common: such a job would make a partner more attractive to 10% of Gen Zers and 12% of Millennials. "Working in AI isn't a neutral data point for many daters. Which side of the AI line someone sits on is being noticed. For Gen Z especially, siding with AI has a cost," the study's authors say. 💼 Roughly one in eight respondents would not be too eager to date someone whose job could be replaced by AI. About half consider job stability to be an extremely important factor when choosing a partner, and one in five women would think twice before dating someone whose job could be replaced by AI. Among men, this number is less than 10%. 💡 At the same time, nearly half of respondents would be more interested in a partner who is learning new skills to stay ahead of AI. Would you date an AI developer? ❤️ — Yes, no problem 🔥 —

❤️ In India, Dogs and AI Are Learning to Detect Cancer Together Startup Dognosis is training dogs to detect cancer by human breath. Each dog wears a 3D-printed helmet that reads the animal's brain activity, along with a harness fitted with sensors that track its reaction to a sample. A tumor changes the body's metabolism, and with it the smell of exhaled breath — and dogs can pick up on that. They have around 300 million scent receptors, compared to about 5 million in humans. The patient only needs to breathe into a cotton mask for 10 minutes, which is then sealed and taken to the lab, where the sample is "examined" by the dog. 💡 There are 15 animals in the project in total, mostly beagles and shepherds. They are trained through positive reinforcement: find the right scent, get a treat. In an hour, a dog can do thousands of sniffs, each one evaluating a separate sample. "Some of the dogs run directly to the feeder expecting rewards, some dogs freeze, some dogs will scratch the sample with excitement," said Dognosis co-founder Itamar Bitan. The AI model analyzes the dog's behavior and cross-references it with patient data — medical history, harmful habits, sex, age, and other parame

👨 What Can DNA Tell You About Your Personality? An international team of scientists analyzed data from 46 cohorts, including up to 1.14 million people, and compared genetic differences with the "Big Five" personality traits: extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience. In total, researchers identified 1,260 genetic variants linked to personality traits—824 of them for the first time. 👍 It turns out that a person's overall genetic profile can partly predict the first impression they make on others. People with a lower tendency toward neuroticism and higher tendencies toward extraversion, agreeableness, and openness were more often perceived as pleasant and attractive in personality. 📌 The study also found a link with where people eventually live. Those with a genetic tendency toward openness were more likely to move to cosmopolitan areas by middle age, while more conscientious people were more likely to live in affluent suburbs. ❤️ Romantic partners are barely similar when it comes to genetic predispositions related to personality. However, their similarity in education level was much stronger. ⚡️ The researchers also found some darker cor

🎧 Robot Can Play Almost Any Song on Request At IFA 2026, Europe's largest consumer electronics show in Berlin, China's TermiTech demonstrated a robot that plays an electronic piano with both hands. Each hand has 21 degrees of freedom. The developers claim TermiPianist can handle more than 80% of popular music—just ask, and it performs the piece on the spot. And these aren't pre-learned movements. 💻 The robot's "AI brain" analyses each composition and matches it against the robot's own physical limits—working out where to place the mechanical fingers, at what speed and range of motion, and in what order to press the keys. @hiaimediaen

📱 Foldable iPhone Duo Officially Unveiled Folded, the 5.4" screen has the aspect ratio of a passport (like the Samsung Z Fold 8 or Xiaomi's new foldable); unfolded, it's an iPad mini with a 7.6" screen. Most foldable phones feel "like two phones stuck together," says Apple CEO John Ternus. iPhone Duo is meant to feel like a single, whole product that opens up new possibilities for the user. Key points: ✅ Unfolded, it's Apple's thinnest smartphone. ✅ A20 Pro chip, the same one as in the iPhone 18 Pro. ✅ 4 cameras: the main Fusion camera and the 48 MP ultra-wide camera from the iPhone 18 Pro, one selfie camera on the small screen, and a "FaceTime camera" hidden under the inner display. ✅ For convenience, the controls have been moved to the right side of the screen when unfolded. Touch ID is in the lock button. ✅ The batteries will last all day: up to 44 hours with the outer display, up to 31 hours unfolded. Just 5 minutes of charging gives up to 5 hours of video on the outer screen. ✅ Survives water thanks to IP68 protection. ✅ There's a split-screen mode for two apps, the option to prop it up "tent-style" without a stand, and instant interface adaptation to the open/closed position,

🔍 Global Demand for AI Slop Cleanup Is Growing On freelance platforms, employers increasingly want not content creators, but people who can clean up traces of slop in AI-generated content. According to Freelancer.​com, which has over 88 million users worldwide, job listings tagged with "correct AI errors" have grown 87% over the past year. On Upwork, such postings have increased by 70%, and on Fiverr, demand for these services has grown 20-fold over the past three years. At the same time, researchers note a decline in demand for traditional creative services susceptible to AI automation. 💡 "AI Cleaners" are especially in demand for graphic design, video editing, proofreading and writing texts, including editing books written by AI. "AI output needs a human to turn it into something polished that can actually be used," says Freelancer.​com CEO Matt Barrie. 🔄 Such jobs are typically posted by small companies or entrepreneurs who can't polish AI-generated content on their own but believe it can be done quickly and cheaply. At the same time, many freelancers complain that editing AI slop takes just as much effort as creating original content. As a result, they burn out from this kind o

🔴 GPT Image 2.5 Is Now Available for Free in @GPT4Telegrambot Now you can try OpenAI's newest "AI Photoshop," GPT Image 2.5, via @GPT4Telegrambot. The model generates even more realistic, detailed images; preserves details more carefully across multiple rounds of edits; and renders long text without typos. How to Try: 1⃣ Go to @GPT4Telegrambot 2⃣ In the menu select Image Generation → GPT Image 2.5 3️⃣ Send your prompt. ✅ 25 images a week for free, 100 a day with Premium. Can you see the progress? ❤️ — Yes, a big leap! 🔥 — Can't really tell anymore... @hiaimediaen

⚡️ OpenAI's AI Has Solved One of the Millennium Prize Problems OpenAI says its internal AI system has solved the existence and smoothness problem for the Navier–Stokes equations. This is one of the seven Millennium Prize Problems, for which the Clay Institute has offered $1 million each. So far, only one of them has been solved—in 2010, the prize for proving the Poincaré conjecture was awarded to Grigori Perelman. ✏️ What is the problem? The Navier–Stokes equations describe Newton's second law as applied to continuous media—liquids and gases. No general solution to them has ever been found, so engineers and mathematicians rely on numerical computation or on formulas for particular cases. The main open question is this: if a fluid is moving smoothly at the initial moment and has finite energy, will its flow stay smooth forever, or will a point appear at some later time where the velocity becomes infinite? Such a point would mean that the equations stop describing a real fluid and that the continuum model breaks down at that spot. OpenAI's model constructed an example of a flow that starts from rest under a smooth and everywhere bounded external force, yet still accelerates to infini

👨‍💻 AI Is Helping a Blind Artist Create Again Jon Paul Ferrara is one of the best-known illustrators in the publishing world. Over the course of his career, he created more than 1,000 book covers, primarily for romance novels. But that career came to an end in 2020, when he underwent surgery to remove a brain tumor. The operation was successful, but Ferrara lost nearly all of his vision. 📱 Now, however, Ferrara is experimenting with AI image generation in ChatGPT and Gemini. He writes prompts the way an artist would, carefully describing the lighting, colors, perspective, composition, and the placement of every person in the scene in minute detail. "It starts with a thought, and then it's all up here. Everything's imagination," Ferrara says. His assistants describe the generated image to him in detail, allowing Ferrara to make adjustments when the AI gets something wrong. After a few rounds of revisions, the final image emerges. "I feel like I have my eyes, even though I can't see what I'm creating," Ferrara explains. 👨 Religious themes have now replaced the romantic couples and glamorous characters that once filled his book covers. Ferrara shares his new work on social media and i