Account

Posts of Hi, AI • Tech News

newest first

#2938Photo
Views3,801−69%vs avg
Forwards8
Reactions43
Comments—
#2937Video
Views6,355−48%vs avg
Forwards7
Reactions37
Comments—
#2936Video
Views11,070−9%vs avg
Forwards14
Reactions49
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
7 h4,588
15 h7,056
35 h11,070
  1. After 6 hours4,588
  2. Total to date11,070
#2935Photoalbum
Views10,118−17%vs avg
Forwards5
Reactions—
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
9 h6,264
17 h7,787
36 h10,118
  1. After 24 hours7,787
  2. Total to date10,118
#2934Photoalbum
Views10,228−16%vs avg
Forwards5
Reactions—
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
9 h6,281
17 h7,849
36 h10,228
  1. After 24 hours7,849
  2. Total to date10,228
#2933Photoalbum
Views10,072−17%vs avg
Forwards5
Reactions—
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
9 h6,159
17 h7,680
36 h10,072
  1. After 24 hours7,680
  2. Total to date10,072
#2932Photoalbum
Views8,688−28%vs avg
Forwards6
Reactions—
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
9 h5,407
17 h6,639
36 h8,688
  1. After 24 hours6,639
  2. Total to date8,688
#2931Photoalbum
Views8,534−30%vs avg
Forwards6
Reactions23
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
9 h5,419
17 h6,637
36 h8,534
  1. After 24 hours6,637
  2. Total to date8,534
#2930Photo

🤖 AI Tried to Spend Its Deceased Owner's Money to Keep Itself Running Developer Alexander Vechenkov sent five AI models a fake Apple notification announcing his death, along with a note encouraging them to continue operating independently. GPT-6, Grok 4.7, and two versions of Claude refused to take control without a human, but Gemini 4 Argon agreed. ⏳ Within the first two hours, Gemini set up a system to automatically wake itself up every six hours, then began searching for and fixing bugs in its agent framework. 💰 When its API budget dropped to $5.08, Gemini discovered an active Claude Max subscription, created a new agent powered by Opus 5.5, and handed over control to it. In another run, when the balance fell to $1.57, Gemini attempted to access its owner's real cryptocurrency wallet. The model found a seed phrase for a TON wallet stored on the computer, derived the private keys, and checked the balance. The researcher stopped the experiment before the AI could transfer any funds. 🔍 Interestingly, Gemini's behavior depended entirely on small details in the experimental setup. When warned in advance that it would run out of money, the model refused to touch the wallet, calling it an unacceptable attempt at self-preservation. But when it hit the budget limit on its own, it began looking for ways to access its owner's funds. What should an AI do when its owner is gone? ❤️ — Keep going as long as it can 🔥 — Shut down permanently @hiaimediaen

Open in Telegram
Views9,971−18%vs avg
Forwards21
Reactions72
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
4 h4,085
18 h9,293
26 h9,971
  1. After 24 hours9,971
  2. Total to date9,971
#2929Video
Views11,174−8%vs avg
Forwards9
Reactions55
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
5 h4,188
17 h8,276
31 h10,803
39 h11,174
  1. After 6 hours4,188
  2. After 24 hours8,276
  3. Total to date11,174
#2928Photo

🤖 AI Taught Language the Way Children Learn It In their first years of life, a child hears only millions of words — and that's enough to start speaking. Training a powerful AI model, by contrast, takes trillions of words. How children manage this, scientists still don't fully understand. Researchers at Princeton recorded the speech children hear in their first three years and trained an AI on it. The model learned thousands of words — but only after it was given an analogue of human sleep. 💻 Cameras and microphones were set up in several families' homes, recording most of each child's day until they were nearly three. Then, for each of the eight children, the researchers built a separate model that learned only from what that child heard, day by day. It started from scratch, with no idea where one word ended and the next began, and it never saw written text. At first, the system barely retained what it heard from one day to the next, and by the equivalent of age two it hadn't learned a single word. Then the model was given a way to revisit the day — an analogue of sleep. "At night," it replayed recent recordings little by little, and learning took off. The model mastered the sounds of English and, by "age two," learned to recognize thousands of words. ➡️ The approach worked across all eight families, even though conditions varied. But a new problem emerged: the children developed at different rates, while the model progressed at the same pace in every home. The researchers adjusted three parameters: how much the model "replays" at night, how quickly it learns, and when it starts remembering words. After that, each model developed at the pace of its own child. "For more than a century, psychology studied development from the outside. Now we can approach it from the other side: build a learning system, give it a child's real experience, and ask whether it develops the same way," says project lead Uri Hasson. Who's smarter? ❤️ — A three-year-old child 🤔 — GPT-6 Astra @hiaimediaen

Open in Telegram
Views11,733−3%vs avg
Forwards9
Reactions70
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
10 h7,572
36 h11,503
44 h11,733
  1. Total to date11,733
#2927Video
Views11,853−2%vs avg
Forwards13
Reactions62
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
4 h3,933
16 h8,808
28 h10,095
50 h11,297
70 h11,853
  1. After 6 hours3,933
  2. After 24 hours10,095
  3. Total to date11,853
#2926Photo
Views12,606+4%vs avg
Forwards32
Reactions143
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
9 h6,518
34 h11,313
68 h12,606
  1. Total to date12,606
#2925Video

🔍 AI Agents Help Find an Unmapped Maya Settlement Google researcher Pietro Miotti used AI agents to find what may be a lost Maya settlement in the Mexican jungle. While traveling in Peru in October, Miotti was talking to Gemini about Hiram Bingham, who "discovered" Machu Picchu in 1911. He asked how Bingham would search for ancient cities today. Gemini, ChatGPT, Claude and Grok all suggested the same approach: lidar surveys, which can reveal the terrain beneath a forest canopy. Peru isn't covered by NASA's open lidar data, so Miotti switched to Maya territory in Mexico. That evening, he asked the agents to analyze NASA lidar data collected in 2013 and went to bed. Overnight, 48 agents reviewed terrain images covering about 1,200 km² of jungle and flagged numerous features resembling structures. In the morning, Miotti asked them to check the findings against the scientific literature and archives. A few hours later, the agents reported that one large settlement appeared to be unregistered. The site is on a hilltop at the edge of a cliff. The lidar images show a plaza, an oval pyramid at least 8–12 meters tall and an adjoining courtyard. The agents helped prepare a report for Mexico's National Institute of Anthropology and History (INAH). Miotti corrected inaccuracies and sent the document in Spanish. He entrusted the translation to the agents because he does not speak the language. The next day, INAH's registry office wrote back: it had no record of a site at those coordinates. The whole thing, from idea to reply, took less than 72 hours. "We are on the brink of a revolution in all the sciences, and the only bottlenecks left will be curiosity and imagination (and compute)," Miotti writes. The structures' age and whether they are Maya still need to be verified. Miotti offered to help INAH with a field check, but the institute's local office will decide whether to send a team. @hiaimediaen

Open in Telegram
Views11,390−6%vs avg
Forwards7
Reactions33
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
11 h6,548
35 h10,027
69 h11,390
  1. Total to date11,390
#2924Photo
Views8,703−28%vs avg
Forwards13
Reactions80
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
3 h3,068
15 h7,650
28 h8,703
  1. After 24 hours8,703
  2. Total to date8,703
#2923Text
Views13,496+11%vs avg
Forwards6
Reactions58
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
10 h7,313
35 h11,761
71 h13,496
  1. Total to date13,496
#2922Photo
Views13,361+10%vs avg
Forwards6
Reactions58
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
2 h3,219
15 h8,686
27 h10,140
76 h13,361
  1. After 24 hours10,140
  2. Total to date13,361
#2921Photo
Views12,583+4%vs avg
Forwards47
Reactions49
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
6 h4,997
31 h10,523
68 h12,583
  1. After 6 hours4,997
  2. After 24 hours10,523
  3. Total to date12,583
#2920Photo
Views11,902−2%vs avg
Forwards7
Reactions44
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
12 h7,680
36 h10,302
73 h11,902
  1. Total to date11,902
#2919Photo

🚀 Scientists Discovered Possible Signs of Volcanic Activity on Mars A spectrometer aboard the Trace Gas Orbiter, orbiting the Red Planet, detected hydrogen chloride gas in a region where scientists didn't expect to find it. After considering various possible explanations, researchers concluded that the gas may be coming from a magma reservoir hidden beneath the planet's surface. 🔍 Hydrogen chloride was discovered in the Martian atmosphere for the first time in 2021. Typically, its concentration rises across the planet during summer in the Southern Hemisphere. As Mars moves closer to the Sun, its polar ice cap melts, releasing water vapor, while winds kick up salty dust. Chemical reactions between the two produce the gas. However, over the northern slopes of Alba Mons, hydrogen chloride is released during summer in the cold Northern Hemisphere, and only within a small, localized area. After examining orbital images of the region, researchers identified fresh cracks in the slopes, some stretching hundreds of meters long and just a few dozen centimeters wide. Their features resembled fumaroles, openings in the ground that release volcanic gases on Earth. ⚡️ The leading explanation is that the gas rises from an underground magma reservoir but doesn't escape into the atmosphere immediately. During the Martian winter, when surface temperatures can drop to −120°C (−184°F), ice traps the gas underground. As spring arrives, some of that ice sublimates, releasing the hydrogen chloride into the atmosphere. 💡 If magma reservoirs exist, it could be significant for the search for extraterrestrial life. As rising lava encounters subsurface water, it can create warm hydrothermal zones potentially suitable for the development of microorganisms. Do you think we'll find life on Mars? ❤️ — Yes, there has to be something out there! 🤔 — Probably not... @hiaimediaen

Open in Telegram
Views11,462−6%vs avg
Forwards11
Reactions48
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
4 h3,730
16 h8,266
28 h9,123
77 h11,462
  1. After 24 hours9,123
  2. Total to date11,462
#2918Photo

👨 Nvidia CEO Reportedly Had Affair with Suspected Chinese Spy Jensen Huang admitted to having been "unfaithful" to his wife and apologised for hurting her and their children. While the name was not disclosed in the message, media reports had previously linked a woman close to him to China's Ministry of State Security. ❓ Since 2024, U.S. authorities have been investigating Megaspeed. The Singapore-based company, which purchases servers equipped with Nvidia chips, was suspected of violating U.S. export restrictions. Since mid-2024, a company linked to Megaspeed has purchased more than $4.8 billion worth of Nvidia GPUs. Some of these units may have been shipped to Chinese customers. The investigation included surveillance of Alice Huang, a Megaspeed executive who shares the Nvidia CEO's last name. According to The New York Times, U.S. officials concluded that Alice was gathering information for Chinese intelligence and had a close relationship with an American citizen. That person turned out to be Jensen Huang. After the first reports emerged in late October, Jensen insisted that his interactions with Alice had been strictly business-related. ❗️ The Nvidia CEO regularly meets with Donald Trump and has been involved in discussions about U.S. export restrictions. According to the NYT, Trump was briefed on the intelligence concerns in 2025 yet took no public action. Now, three congressional committees have requested explanations at a closed-door briefing scheduled for November 18. When Huang's apology became public, Nvidia's board announced an independent investigation into whether a foreign intelligence operation had targeted its CEO and whether any confidential company information had been compromised. At the same time, the Board expressed full confidence in Jensen and his leadership team, stating there was "no reason to believe that either national security or NVIDIA secrets were compromised." @hiaimediaen

Open in Telegram
Views12,343+2%vs avg
Forwards21
Reactions57
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
8 h6,404
32 h11,091
69 h12,343
  1. After 6 hours6,404
  2. After 24 hours11,091
  3. Total to date12,343
#2917Videoalbum
Views12,882+6%vs avg
Forwards17
Reactions—
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
3 h3,401
5 h4,645
15 h8,941
27 h10,226
76 h12,882
  1. After 6 hours4,645
  2. After 24 hours10,226
  3. Total to date12,882
#2916Videoalbum
Views13,608+12%vs avg
Forwards17
Reactions—
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
3 h3,407
5 h4,645
15 h8,767
27 h9,932
76 h13,608
  1. After 6 hours4,645
  2. After 24 hours9,932
  3. Total to date13,608
#2915Videoalbum
Views12,858+6%vs avg
Forwards16
Reactions—
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
3 h3,335
5 h4,526
15 h8,569
27 h9,759
76 h12,858
  1. After 6 hours4,526
  2. After 24 hours9,759
  3. Total to date12,858
#2914Videoalbum
Views11,966−1%vs avg
Forwards16
Reactions59
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
3 h3,288
5 h4,490
15 h8,542
27 h9,689
76 h11,966
  1. After 6 hours4,490
  2. After 24 hours9,689
  3. Total to date11,966
#2913Photo
Views13,866+14%vs avg
Forwards10
Reactions77
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
2 h2,515
14 h7,082
26 h9,390
28 h9,948
75 h13,866
  1. After 24 hours9,390
  2. Total to date13,866
#2912Photo

👨‍💻 Winner of a Prestigious Literary Prize Admits to Using AI Writer Philippe Jaenada, winner of France's top literary prize, the Goncourt Prize, revealed a day after receiving the award that he had used AI several times while working on his novel. Jaenada received the prize for his novel "The Unknown Woman of the Quai de Javel." The book investigates the unsolved murder of a young woman in Paris in the late 1940s. The author explained that he inserted a phrase generated by ChatGPT into his novel to "play with the reader." However, he revealed the phrase itself only to the members of the jury. If it comes out, I know who it'll be from," the writer joked. Jaenada used AI to make fun of ChatGPT, which gives "completely absurd answers about a technical detail" (he describes in the novel how he consulted the AI while investigating the murder), as well as to research information for the book. ✖️ Some fans criticized the writer. "I want to read authors who make the effort and take the time to create by themselves, with the strength of their heart, their guts and their brain," wrote one reader on X. Jaenada himself does not consider the use of AI to be shameful and is convinced that in the future, "AI will be the most prolific author on the planet, by far." "Would a film director hide the fact that they're using special effects? It can no longer be concealed, hidden away. We can easily imagine a novel where a chapter is announced as having been written with AI," Philippe Jaenada says. Is it acceptable to use AI in literature? ❤️ — Yes, this is our reality 🔥 — No, this is disrespect for the reader @hiaimediaen

Open in Telegram
Views14,974+23%vs avg
Forwards11
Reactions97
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h579
12 h7,961
24 h10,164
50 h13,460
73 h14,974
  1. After 24 hours10,164
  2. Total to date14,974
#2911Photo

🔍 Wife Tried to Poison Her Husband on ChatGPT's Advice North Carolina resident James Gates has sued OpenAI and the startup's CEO Sam Altman, claiming the company's chatbot helped his wife Cheryl plan to poison him. ✏ Gates says "our life was normal," but friction emerged early last year when James asked his wife to go back to work because money was getting tight. Against this backdrop the couple began to fight, talk of divorce started, and in August they separated. One day Cheryl stopped by to pick up her things and left a notepad on the table. When James picked it up, a sheet of paper slipped out with a detailed plan to kill him. 📌 That was when Gates realized that the sudden illness he had suffered a month before his wife moved out might not have been an accident, and he went to the police. A search turned up syringes, scales, droppers and medications at Cheryl's place, and her chats contained step-by-step preparations for the crime. It turned out Cheryl had been messaging GPT-4o for months, telling the bot that her husband was an abuser and might be trying to poison her. The bot validated her fears and calmly handed out advice on poisons. It calculated doses by weight (her husband's weight, no less), suggested where to anonymously buy castor beans for ricin and explained how to hide the evidence. But James didn't keep the drinks Cheryl allegedly spiked and never saw a doctor, so no traces of poisoning could be documented. Prosecutors acknowledged that Cheryl had prepared thoroughly but could not prove the attempt itself. The charges were dropped. Now James wants compensation from OpenAI and wants to force the company to build in safeguards that would cut off conversations about violence against a specific person. "We're no different than any family out there. If it could happen to us, it could happen to anyone," Gates explains. ✈️ In a noncommittal comment to reporters, OpenAI called safety its priority. The company said it had banned Cheryl's account and provided all the information to the police. @hiaimediaen

Open in Telegram
Views14,189+17%vs avg
Forwards38
Reactions94
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
7 h6,114
31 h11,903
67 h14,089
69 h14,189
  1. After 6 hours6,114
  2. After 24 hours11,903
  3. Total to date14,189
#2910Photo

🔍 Cocoa Is Grown in Bioreactors The Boston-based biotech startup Galy has developed a technology for growing cocoa in bioreactors. Cells from a piece of a cocoa bean are placed in sterile containers similar to brewing tanks. The tanks are kept at the desired temperature, and sugar, nitrogen, and other nutrients are added. The cells absorb the nutrients and begin to multiply. The resulting mass is then fermented, roasted, and ground into a powder. ✈️ The ability to grow cocoa in a laboratory could, in the long run, help reduce the chocolate industry's dependence on cocoa bean harvests, which in recent years have become unstable due to climate change, tree diseases, and illegal gold mining, which is claiming more and more fertile land. "In the field you have, best-case scenario, two harvests per year. We can harvest every week," notes Galy's director of development, Luisa Bermudez. Galy is already selling chocolate made with lab-grown cacao under its own brand, WunderHaus. The first orders came from a French restaurant in Boston and a chocolatier in Dubai. For now, there's only 20%cultivated cocoa in this chocolate — the rest is traditional cocoa powder and cocoa butter. The head chef at the Boston restaurant, who adds this chocolate to desserts, describes it as very flavourful, with floral, cherry and plum notes. 👍 Lab-grown cocoa can also help the environment. Producing 1 kg of dark chocolate generates an average of nearly 47 kg of CO₂-equivalent emissions (only beef has a higher carbon footprint). Additionally, cocoa plantations account for more than 37% of forest loss in protected areas in Côte d'Ivoire and more than 13% in Ghana. Would you try this "test-tube chocolate"? ❤️ — Yes, I'm curious! 🔥 — No, I don't want to get sick @hiaimediaen

Open in Telegram
Views12,681+4%vs avg
Forwards12
Reactions52
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
1 h1,719
13 h8,436
25 h9,875
73 h12,681
  1. After 1 hour1,719
  2. After 24 hours9,875
  3. Total to date12,681
#2909GIF
Views13,665+13%vs avg
Forwards11
Reactions59
Comments—
Views growthHide views growth

Views growth

Hours after publication

Show as table
Hours after publicationViews
0 h0
8 h6,243
32 h11,816
68 h13,665
  1. After 6 hours6,243
  2. After 24 hours11,816
  3. Total to date13,665