1508732
World Tech News by Perplexity. AI News by @GPT4Telegrambot Driven by the Community. Contact: @i_abramov_gpt
📣 Hello everyone! Our Sunday digest features the most exciting AI news from Week 16, 2025.
▎NEW RELEASES
👐 OpenAI has unveiled its smartest o3 and o4‑mini models, along with the GPT‑4.1 family, which features a 1M token context window.
♊️ Gemini 2.5 Flash: Google's new reasoning model, now ranked #2 on LMArena.
📹 Kling 2.0: A major upgrade to one of the leading AI video generators.
🔴 All these new AI features are available on @GPT4Telegrambot
⛔️ Trump May Ban DeepSeek in the U.S.
Reports suggest that the Trump administration may soon block Americans from using AI models built by the Chinese lab DeepSeek. Officials are also weighing measures to stop the company from buying U.S. technology—such as Nvidia’s AI chips—over national security concerns.
DeepSeek is already restricted in several countries: Canada, Australia, South Korea, and Taiwan have barred the service on government devices, and Italy has imposed a nationwide ban.
Read More
👐 OpenAI Introduces o3 & o4‑mini: Next‑Gen AI Models with Superior Reasoning and Tool Skills
OpenAI has made a new breakthrough with the announcement of the o3 and o4-mini models, the smartest and most powerful models to date.
The president of OpenAI Greg Brockman compares their release to the leap in progress the company made with GPT-4 two years ago: "And the thing that's really amazing about these models is that they're not just models, they're really AI systems. We've trained them to use tools, which is not something that we had done with our previous reasoning models."
How to Use?
➡️ Go to @GPT4Telegrambot
1️⃣ Purchase the "CHATGPT & DALL•E 3" bundle in the /premium section
2️⃣ Select o3 or o4-mini in the /model section
⚡️ Kling 2.0: The Next-Gen AI Video Model Is Now on @GPT4Telegrambot
Hi everyone! Meet 📹 Kling 2.0. One of the leading AI video generators just got a major update and is now available on @GPT4Telegrambot 🔴
What's New and Why Is It Cool?
➡️ Quality: 720p video, 30 fps, realistic animation, and high-detail motion.
➡️ AI Features: Improved shadows and physics, 3D reconstruction of faces and bodies from a single photo.
➡️ Smart Camera: The model understands prompts well and sets camera movements—such as zoom, pan, and other cinematic effects.
➡️ Format Support: 16:9, 9:16, and 1:1 for any media or platform.
➡️ Image-to-Video: Especially great at generating video from images; animation is smooth, and transitions are natural.
Who Is This for?
This is for filmmakers, marketers, bloggers, and artists of any level who want to quickly create impressive video content.
How to Create a Video:Читать полностью…
🔴 Go to @GPT4Telegrambot
1️⃣ Purchase the KLING bundle in the /premium section.
2️⃣ Send /video command to the bot.
3️⃣ Describe the video you want to create and choose version 2.0. For best results, we recommend adding an image.
👐 OpenAI's Newest GPT-4.1 Models Are on @GPT4Telegrambot
Hi everyone! OpenAI released a new GPT-4.1 model family. You can already use advanced GPT-4.1 and free GPT-4.1 mini in our bot 🔴
Why choose GPT-4.1:
➡️ Better writes texts and understands many languages.
➡️ In math and sciences, the new model's capabilities are on par with GPT-4.5 while being significantly cheaper.
➡️ For coding, GPT-4.1 is close to the capabilities of "reasoning" o1 and o3-mini models.
➡️ Knowledge cutoff: June 2024.
How to Use?
1️⃣ Go to @GPT4Telegrambot
2️⃣ Purchase the "ChatGPT + DALL•E 3" in the /premium section
3️⃣ Select the GPT-4.1 in the /model section
How to Use?
1️⃣ Go to @GPT4Telegrambot
2️⃣ Select the GPT-4.1 mini in the /model section
📣 Hello everyone! Our Sunday digest features the most exciting AI news from Week 15, 2025.
▎BOT UPDATES
📱 Midjourney V7, a brand-new AI model for creating highly detailed and beautiful images, is now available on @GPT4Telegrambot.
▎TO READ
🤩 Altman vs. Miyazaki: ChatGPT's Ghibli trend heats the AI art debate.
🐺 Startup Colossal resurrected the dire wolf, extinct over 10,000 years. George R.R. Martin, author of Game of Thrones, shared a photo with the resurrected dire wolf 👆
🛑 AI industry predictions from OpenAI's ex-researcher Daniel Kokotajlo for the next five years.
❤️🔥 Tinder launched an AI game for practicing flirting skills.
🔺 Mathematicians solved a 122-year-old problem.
🔬 How do scientists use AI to fake research?
👁🗨 Why is the "openness" of AI companies just a mirage?
🧬 What medical breakthroughs has AlphaFold already brought us?
🤖 Samsung integrates Gemini AI into its Ballie home robot.
▎TO WATCH
💿 "Tron: Ares": What to expect in the next film of the legendary saga.
👆 Oxford mathematician John Lennox on the benefits and dangers of AI for society.
Have a great Sunday 👍
🛰 NASA Can Return to Moon And Explore Mars
Jared Isaacman, President Trump's nominee for NASA Administrator, has outlined his vision for America's future in space, according to multiple reports. His strategy involves sending astronauts to explore Mars while also continuing efforts to return humans to the Moon.
During his Senate confirmation hearing, Isaacman emphasized that missions to the Moon and Mars can be pursued simultaneously. He stated, "I don't think these are either-or."
Read More
🐺Colossal Brings Back the Dire Wolf, Extinct for Over 10,000 Years
Scientists precisely edited the DNA of modern-day gray wolves to create puppies as genetically close as possible to their extinct ancestors. The resulting pups, Romulus, Remus, and Khaleesi, mark a milestone in de-extinction efforts. The researchers have not yet published details of the process.
Similarly, the company plans to bring back the woolly mammoth by genetically editing the Asian elephant. The first step toward this, "woolly mice," was demonstrated in March.
Read More
📣 Hello everyone! Our Sunday digest features the most exciting AI news from Week 14, 2025.
▎NEW RELEASES
™️ Midjourney V7: A new version of the most aesthetic text-to-image AI model. Now, it has a super-fast "draft mode" and personalization.
🎬 Higgsfield AI: A new AI video generator for cinematic videos.
💻 Meta introduced Llama 4, a new herd of open-source models.
▎BOT UPDATES
♊️ Gemini 2.0 Image Editor on @GPT4Telegrambot. Send the bot an image and describe what you'd like to change. Free to all bot users.
▎TO READ
🗣 New AI feature from Krisp "transforms" an Indian English accent into an American one.
🖥 Nvidia challenger Etched AI closed a $85M investment round, reaching a $1.5B valuation.
❤️ Our team has also invested in Etched through DVC.
🖥 Nvidia Challenger Etched AI Hits $1.5B Valuation
Cupertino-based startup Etched AI closed a $85 million unannounced investment round, reaching a $1.5 billion valuation.
Etched AI develops a specialized chip, Sohu, designed to run AI models. The company declares Sohu is the fastest AI chip of all time and outperforms Nvidia's flagship B200 GPU by 10x. Etched AI says its chip aims to run "tomorrow's trillion-parameter models," predicting that major AI products such as ChatGPT, Claude, and Gemini will run on custom chips instead of usual GPUs within a few years.
Etched is backed by investors, including former CEO Peter Thiel, Replit CEO Amjad Masad, Primary Venture Partners, DVC fund, etc.
❤️ Our team has also invested in Etched through DVC.
⚛️ CERN Proposes 91-km Particle Accelerator
CERN has unveiled plans for the Future Circular Collider (FCC), a 91-km particle accelerator under Lake Geneva along the French-Swiss border. The FCC is set to revolutionize particle physics, exploring fundamental forces and pushing technological innovation. Its energy levels could potentially surpass those of the Large Hadron Collider by tenfold. If approved, construction may begin in the mid-2040s.
The project includes two phases: the first, starting in the mid-2040s, will focus on high-precision experiments, while the second, expected by 2070, will feature high-energy collisions of protons and heavy ions. Estimated costs are around $14 billion.
Read More
🚀 European Rocket Crashes after Launch
The first orbital rocket launch from mainland Europe ended in failure on March 30, 2025, when the Spectrum rocket, developed by German start-up Isar Aerospace, crashed and exploded just 40 seconds after takeoff from Norway's Andøya Spaceport.
Despite the failure, Isar Aerospace, the German start-up behind the project, described the mission as a success in terms of data collection and system validation.
Read More
🦟 New Drug Turns Blood Lethal to Mosquitoes
A recent study has found that nitisinone, a drug usually used to treat rare genetic disorders, can make human blood deadly to mosquitoes. It works by blocking an important enzyme they need to digest blood, providing a fast and groundbreaking way to fight malaria.
According to Science Alert and the University of Notre Dame, this method is effective against different types of mosquitoes, including those that are resistant to insecticides and difficult to control with current strategies.
Read More
👐 OpenAI Upgrades ChatGPT's Image Generation
OpenAI has added new image generation based on GPT-4o model directly into ChatGPT. It replaces the previous DALL-E 3 system. The update brings improved features, such as understanding both images and text together, and better quality for text shown within images.
The tool understands the ongoing chat conversation, allowing it to build upon earlier images and text while keeping the style consistent through revisions.
Read More
🕶 Samsung to Launch AI Glasses
Samsung is developing AI-powered smart glasses, codenamed "Haean," and can introduce them before the end of 2025. The Haean smart glasses will incorporate cutting-edge technology, including Google's Gemini AI for smart assistant capabilities and gesture control for intuitive operation.
The device will likely have a 12-megapixel camera using a Sony image sensor and run on Qualcomm's Snapdragon AR1 chip. Designed to be comfortable and useful for daily life, the glasses are expected to weigh around 150 grams. They will offer features such as instant language translation, navigation assistance, and QR code payments.
Read More
♊️ Google has unveiled Gemini 2.5 Flash, and it's already placed second on the LMArena leaderboard.
➡️ Math and coding skills are several times better than Gemini 2.0 Flash.
➡️ It's Google's first hybrid model, using different reasoning approaches depending on the task's complexity.
🔴 The model is already available in our bot for free!
How to Use:Читать полностью…
1️⃣ Go to @GPT4Telegrambot
2️⃣ Select Gemini 2.5 Flash in the /model section.
🦑 First‑Ever Video of Colossal Squid
According to reports, more than 100 years after the species was first identified, scientists have finally recorded the first video of a living colossal squid in its natural home. They captured a juvenile about 600 meters beneath the surface, near the South Sandwich Islands in the South Atlantic Ocean ⤴️
The footage was taken by the deep‑sea robot ROV SuBastian. Operated remotely from a research ship, the robot weighs about 3,200 kg, and it is roughly the size of a small city car.
Read More
📱 Tim Cook Wants to Beat Meta at AR
Bloomberg reports that Apple CEO Tim Cook is "hell-bent" on developing true augmented reality glasses before Meta can bring a similar product to market. "Tim cares about nothing else," sources say.
The intense focus on AR glasses, which are more suitable for daily wear than Vision Pro, represents Cook's bid to cement his legacy at Apple beyond the iPhone, which still accounts for approximately 56% of the company's revenue as of January 2025.
Read More
😃 New Study Could Help End Baldness
A recent study found that a protein called MCL-1 is very important for hair growth. MCL-1 helps keep hair follicle stem cells alive. These stem cells are needed to grow new hair.
In experiments with mice, when scientists removed MCL-1 from these stem cells, the cells died and the mice lost their hair because new hair could not grow.
This discovery suggests that helping the body make more MCL-1 might be a way to treat some types of hair loss in the future. However, more research is needed before this can become a treatment for people.
Read More
🪐 An 18-Year-Old Teen Discovers 1.5 Million Space Objects
Californian 18-year-old Matteo Paz identified 1.5 million new space objects with AI. He won $250,000 in the Regeneron Science Talent Search for his efforts.
Paz's algorithm analyzes data from NASA's NEOWISE mission. The VarWISE catalog, created by Paz's AI algorithm, contains 1.9 million variable objects, of which 1.5 million were previously unknown to science. This groundbreaking dataset offers valuable insights into phenomena such as the expansion rate of the universe and the formation of distant star systems.
Read More
🤖 Samsung adds Gemini AI to Ballie
Samsung will enhance its home robot Ballie by integrating Google's Gemini AI, boosting its conversational skills. Scheduled for release in the U.S. and South Korea this summer, Ballie will use Gemini to handle audio, visual, and sensor data, offering more personalized user experiences.
Ballie can be your training coach, give health advice, offer fashion tips, and help with pet care—or even be your pet if you don't have one.
Read More
🌐 Google's AI Division Pays Developers Full Salary for Staying Inactive
DeepMind's noncompete agreements for departing employees range from six months to one year, with longer durations typically reserved for senior researchers or those working on critical projects such as Google's Gemini AI models. During this "garden leave," employees receive full pay but cannot work for competitors.
This policy focuses on key personnel whose skills could greatly benefit rival companies in the fast-moving AI industry.
However, many researchers are frustrated. They worry that long periods of inactivity might leave them behind during a critical phase of AI development and innovation.
Read More
📱 Midjourney V7 Is Here: Now Available on @GPT4Telegrambot
Hi everyone! We've integrated Midjourney's brand-new image generation model V7 into our bot.
Why Should You Try It?
1️⃣ Aesthetics: Midjourney excels at creating stunning visuals. They're №1 in this field, and V7 raises the bar even higher. A true treat for artists and visual enthusiasts!
2️⃣ Detail: Images have become sharper, with improved textures (skin, fabrics, lighting) and more realistic facial expressions.
3️⃣ Improved Anatomical Accuracy: Say goodbye to awkward hands and distorted proportions — Midjourney V7 significantly reduces errors in depicting hands, bodies, and complex objects.
4️⃣ Cinematic Quality: Early users have praised the new model for its artistic expressiveness and dramatic visual storytelling.
5️⃣ Photorealism Boost: Images generated by Midjourney look as if they were captured with a camera or smartphone — truly lifelike!
Developers promise more updates in the coming weeks ❤️
How To Use It?
🔴 Go to @GPT4Telegrambot
✨ Purchase the "Midjourney + FLUX" bundle in the /premium section
✨ Send the /imagine command followed by a text prompt to generate an image
💎 DeepMind AI Finds Minecraft Diamond
Google DeepMind's AI, DreamerV3, has learned to mine diamonds in Minecraft completely independently. The model didn't use real human players' experience or pre-written strategies; instead, it employed reinforcement learning to figure out the necessary steps through trial and error within Minecraft's complex world.
Finding diamonds in Minecraft is a difficult task requiring careful planning and resource management. It often takes skilled humans up to 30 minutes, and it serves as a strong test for AI capabilities.
Scientists suggest that DreamerV3's self-learning method could be adapted to train robots and AI for complex, real-world challenges.
Read More
♊️ Google Gemini 2.0 Image Editor on @GPT4Telegrambot
Hi everyone! Now you can edit your images directly in our bot using natural language — just describe the changes you want to see!
Our AI-powered editor, based on Google Gemini 2.0, is available for free to all bot users.
How to Use:
🔴 Go to @GPT4Telegrambot
➡️ to the /photo section
1️⃣ Select "Gemini Editor"
2️⃣ Upload any image
3️⃣ Write in the chat what you'd like to change; wait for 10 seconds 🙂
🌱 Feel free to edit the same image as many times as you want!
☺️ Largest Dinosaur Claw Discovered
Paleontologists have discovered the largest fully preserved dinosaur claw in Mongolia's Gobi Desert, measuring approximately 30 centimeters (nearly a foot). It belonged to the newly identified species Duonychus tsogtbaatari. Herbivorous therizinosaur stood about 3 meters (10 feet) tall and weighed around 270 kilograms (600 pounds). The fossil, which includes parts of the spine, pelvis, arms, and legs, dates back roughly 90 million years.
This remarkable find, featuring an intact keratin sheath and unique two-fingered anatomy, provides unprecedented insights into therizinosaur evolution, anatomy, and behavior.
Read More
📣 Hello everyone! Our Sunday digest features the most exciting AI news from Week 13, 2025.
▎NEW RELEASES
💛 ChatGPT’s new image generator is taking the world by storm. The internet is flooded with memes in the style of Hayao Miyazaki anime and The Simpsons, stickers, creative infographics, and much more.
♊️ Gemini 2.5 Pro: Google's new champion LLM outperforms OpenAI or xAI models.
🎨 Ideogram 3.0: A new AI generator for realistic images and logos.
🐳 The new version of DeepSeek-V3 (Mar'25) is now available for free in our bot.
▎TO READ
🍽 Robot waiters: Will they save restaurants or scare away guests?
🦟 New drug turns blood lethal to mosquitoes can help to fight malaria.
🤩 A new AI model diagnoses Parkinson's disease with 96% accuracy.
👶 Babies start forming memories by 12 months, a new study finds.
👍 Does AI have a sense of humor? Yes — but not everyone enjoys AI-generated jokes.
🛰 A startup has developed a satellite engine that can run on space debris.
🕶 Samsung can introduce its AI glasses before the end of 2025.
▎TO WATCH
🎬 The Thinking Game: A documentary about the life of Nobel laureate and DeepMind founder Demis Hassabis.
📺 How AI will change coding, explains Vercel and Next.js creator Guillermo Rauch.
Enjoy your Sunday 👍
👶 Infants Can Form Memories
A new study found that babies as young as 12 months form memories using their hippocampus, which scientists thought wasn't possible so early.
This research challenges ideas about why we forget being babies ("infantile amnesia"). Early memories might become hard to access later on, not that babies couldn't make them initially. This opens new ways to understand early brain development and memory formation.
It could also lead to new ideas for early childhood education and development.
Read More
🐳 The New DeepSeek-V3 (Mar 25) is on @GPT4Telegrambot
Hi everyone! We've integrated the recently updated DeepSeek-V3 model into our bot.
Why Should You Try It?
⭐ The new model immediately topped the Artificial Analysis ranking among "non-reasoning" models, surpassing Grok 3, Claude 3.7, GPT-4.5, and Gemini 2.0 Pro.
For the first time ever, an open source model holds the top spot. The ranking is based on the results of seven key benchmarks, including general knowledge, language understanding, math, and coding.
How to Use?
🔴 Go to @GPT4Telegrambot
✨ Select the DeepSeek-V3 (Mar'25) in the /model section.
DeepSeek-V3 is available for free.
🤖 1X to Test Humanoid Robots in Homes
1X, a Norwegian company that makes robots, is planning to test its human-like robot, Neo Gamma in homes by the end of 2025. They will place the robot in many homes, between a few hundred and a few thousand, to see how it works in real-life situations.
When these tests first start, skilled people will control the robot's movements from a distance. This is to ensure the robot's safety and collect information about how it performs in different homes. The company is doing this to eventually make the robot able to do tasks by itself without needing human control.
Read More