技术
Hi, AI | Tech News
技术
25.04.202511:59
🔫 Microsoft's AI Generates Quake II in Real-Time, and You Can Even Play It
Microsoft has entirely recreated the iconic 1997 shooter Quake II within an AI model. The Muse AI processes user actions and generates the environment, enemies, gunfire, and effects in real-time. The developers have trained the model on week's worth of human gameplay.
Despite the AI version of Quake resembling the original, playing it is challenging. For instance, if the player turns away, they might end up in a locked room or teleport to another part of the map. Health and damage counters are often wrong, and objects may disappear. These issues stem from the model's limited context: the game "remembers" only the last nine frames at any given moment.
Nevertheless, enemies react as they should when hit by the player, and shots can still cause explosive barrels to detonate. Players can even stumble upon secret locations from the original game thanks to the variety of training data. Developers emphasize that AI-generated Quake should be seen primarily as a demonstration of the model's capabilities rather than a fully functional game.
In the future, however, this technology could make it possible to run classic games on any modern platform, according to Microsoft Gaming CEO Phil Spencer.
🔫 You can try AI-powered Quake II here—it works directly in your browser.
More on this topic:
👉 Top 7 Most Anticipated Games of 2025
👉 Games with Infinite Worlds: Can You Reach the End?
#news #games @hiaimediaen
Microsoft has entirely recreated the iconic 1997 shooter Quake II within an AI model. The Muse AI processes user actions and generates the environment, enemies, gunfire, and effects in real-time. The developers have trained the model on week's worth of human gameplay.
Despite the AI version of Quake resembling the original, playing it is challenging. For instance, if the player turns away, they might end up in a locked room or teleport to another part of the map. Health and damage counters are often wrong, and objects may disappear. These issues stem from the model's limited context: the game "remembers" only the last nine frames at any given moment.
Nevertheless, enemies react as they should when hit by the player, and shots can still cause explosive barrels to detonate. Players can even stumble upon secret locations from the original game thanks to the variety of training data. Developers emphasize that AI-generated Quake should be seen primarily as a demonstration of the model's capabilities rather than a fully functional game.
In the future, however, this technology could make it possible to run classic games on any modern platform, according to Microsoft Gaming CEO Phil Spencer.
🔫 You can try AI-powered Quake II here—it works directly in your browser.
👉 Top 7 Most Anticipated Games of 2025
👉 Games with Infinite Worlds: Can You Reach the End?
#news #games @hiaimediaen
23.04.202513:00
🙌 Trust in AI Companies is Falling, Stanford Researchers Have Found
According to Stanford researchers, the world is growing cautiously optimistic about AI every year. Still, public trust in AI companies is falling.
For the new 2025 AI Index report, researchers surveyed more than 24,000 people from 32 countries. It's one of the largest and most authoritative studies on the topic.
Highlights from the report:
⚫️ A large majority believe AI-powered products and services offer more benefits than drawbacks in countries like China (where 83% welcome AI adoption), Indonesia (80%), Thailand (77%), and Mexico (70%). However, less than half share this view in Australia (44%), Canada (40%), the US (39%), and the Netherlands (36%).
⚫️ Even so, optimism about AI is growing, especially in Europe. In Germany and France, 10% have changed their opinion of AI for the better since 2022, and in the US, — 4%.
⚫️ At the same time, people are becoming disillusioned with AI leaders. Globally, confidence in AI companies' ability to protect personal data has fallen from 50% to 47% within a year.
⚫️ Globally, 66% believe AI will significantly change their lives in three to five years. However, only 55% see AI as doing more good than harm (52% thought so in 2022).
⚫️ Just over half of those surveyed believe AI helps save time and improves the quality of entertainment. Only one in three thinks AI will improve healthcare, the national economy, and working conditions.
⚫️ Only 36% of respondents expect AI to replace their jobs in five years. Still, 60% agree that AI will change their profession in one way or another.
⚫️ Generational differences are obvious. While nearly two-thirds of Generation Z anticipate AI to significantly impact their jobs, fewer than half of Baby Boomers share this belief.
How do you think AI will affect your job?
🔥— I don't think it will affect
👍 — it might change my job, but it won't take it away
👾 — perhaps AI will replace me...
#news #science @hiaimediaen
According to Stanford researchers, the world is growing cautiously optimistic about AI every year. Still, public trust in AI companies is falling.
For the new 2025 AI Index report, researchers surveyed more than 24,000 people from 32 countries. It's one of the largest and most authoritative studies on the topic.
Highlights from the report:
⚫️ A large majority believe AI-powered products and services offer more benefits than drawbacks in countries like China (where 83% welcome AI adoption), Indonesia (80%), Thailand (77%), and Mexico (70%). However, less than half share this view in Australia (44%), Canada (40%), the US (39%), and the Netherlands (36%).
⚫️ Even so, optimism about AI is growing, especially in Europe. In Germany and France, 10% have changed their opinion of AI for the better since 2022, and in the US, — 4%.
⚫️ At the same time, people are becoming disillusioned with AI leaders. Globally, confidence in AI companies' ability to protect personal data has fallen from 50% to 47% within a year.
⚫️ Globally, 66% believe AI will significantly change their lives in three to five years. However, only 55% see AI as doing more good than harm (52% thought so in 2022).
⚫️ Just over half of those surveyed believe AI helps save time and improves the quality of entertainment. Only one in three thinks AI will improve healthcare, the national economy, and working conditions.
⚫️ Only 36% of respondents expect AI to replace their jobs in five years. Still, 60% agree that AI will change their profession in one way or another.
⚫️ Generational differences are obvious. While nearly two-thirds of Generation Z anticipate AI to significantly impact their jobs, fewer than half of Baby Boomers share this belief.
How do you think AI will affect your job?
🔥— I don't think it will affect
👍 — it might change my job, but it won't take it away
👾 — perhaps AI will replace me...
#news #science @hiaimediaen
21.04.202507:31
🔄 Gamma: Create Presentations in Just a Few Clicks with AI
The recent update of the AI-powered presentation creation service, Gamma, has expanded its capabilities: now it generates not only text and images but also tables with charts, converts slides into social media cards, and allows you to choose from original illustrations in addition to AI-generated images.
How to create a presentation in Gamma?
1️⃣ Go to the Gamma website, click "Start for Free," and sign up.
2️⃣ Click "Create New AI" and choose one of the options for generating your presentation: based on your outline, fully AI-generated, or by uploading an existing presentation for editing.
3️⃣ Select the type of project (presentation, website, document, or social media post), the number of slides, the language, and click "Create Outline."
4️⃣ Review the outline content. Choose a design, image creation method, add style preferences, and click "Generate!"
✒️ The AI will immediately begin creating your presentation. You can manually edit slides, make changes via the chatbot, animate illustrations, and create infographics.
Gamma is an excellent tool for studying, creating social media content, or quickly prototyping websites.
More on the topic:
🔆 Our productivity guides
🔆 Free AI tools for studying
#news #Gamma @hiaimediaen
The recent update of the AI-powered presentation creation service, Gamma, has expanded its capabilities: now it generates not only text and images but also tables with charts, converts slides into social media cards, and allows you to choose from original illustrations in addition to AI-generated images.
How to create a presentation in Gamma?
1️⃣ Go to the Gamma website, click "Start for Free," and sign up.
2️⃣ Click "Create New AI" and choose one of the options for generating your presentation: based on your outline, fully AI-generated, or by uploading an existing presentation for editing.
3️⃣ Select the type of project (presentation, website, document, or social media post), the number of slides, the language, and click "Create Outline."
4️⃣ Review the outline content. Choose a design, image creation method, add style preferences, and click "Generate!"
✒️ The AI will immediately begin creating your presentation. You can manually edit slides, make changes via the chatbot, animate illustrations, and create infographics.
💲 Upon registration, you receive 400 credits—enough to create 10 presentations. Remember that all AI edits consume additional credits (starting at 10 per edit).
Gamma is an excellent tool for studying, creating social media content, or quickly prototyping websites.
More on the topic:
🔆 Our productivity guides
🔆 Free AI tools for studying
#news #Gamma @hiaimediaen
18.04.202506:59
⚪️ How Pokémon Helps Benchmark AI Progress
Pokémon Red was released on the Game Boy nearly 30 years ago. However, the game still has a devoted fanbase—including some of the researchers at AI startup Anthropic. In June 2024, the team behind Claude decided to test how well their AI model could play the classic game. What started as a lighthearted experiment quickly became something of a cult phenomenon inside the team.
When Anthropic introduced Claude 3.7 Sonnet, they highlighted its performance in Pokémon. Researcher Diane Penn explained that watching a model play the game says more about AI progress than most standardized benchmarks. "We're at a point where evaluations don't tell the full story of how much more capable each version of these models are," she said.
🎮 Pokémon Is Harder Than Chess for AI
AI has long surpassed humans in chess, Go, and even in complex games like StarCraft. But open-world RPGs like Pokémon Red, with random events and dynamic choices, are a better proxy for real-life tasks. They require more than just knowledge—they demand agentic skills: decision-making, goal-tracking, and interaction with characters. That's what makes them a powerful testbed for real-world AI applications.
So far, Claude 3.7 Sonnet has only defeated a few gym leaders, but that's already a leap forward. The previous version struggled to leave the starting house in Pallet Town.
Even more impressive is Google's Gemini 2.5 Pro. Since early April, an AI enthusiast has been livestreaming the model as it plays Pokémon Blue — a slightly modified version of Red. According to Reddit users, Gemini has already made it much farther than Claude did over the same timeframe. Then again, Gemini does get a little help: a custom mini-map helps it avoid getting stuck.
More on the topic:
🔴 Pokémon Go data is being used to train AI
💎 Google DeepMind's AI plays Minecraft
#news #games @hiaimediaen
Pokémon Red was released on the Game Boy nearly 30 years ago. However, the game still has a devoted fanbase—including some of the researchers at AI startup Anthropic. In June 2024, the team behind Claude decided to test how well their AI model could play the classic game. What started as a lighthearted experiment quickly became something of a cult phenomenon inside the team.
When Anthropic introduced Claude 3.7 Sonnet, they highlighted its performance in Pokémon. Researcher Diane Penn explained that watching a model play the game says more about AI progress than most standardized benchmarks. "We're at a point where evaluations don't tell the full story of how much more capable each version of these models are," she said.
🎮 Pokémon Is Harder Than Chess for AI
AI has long surpassed humans in chess, Go, and even in complex games like StarCraft. But open-world RPGs like Pokémon Red, with random events and dynamic choices, are a better proxy for real-life tasks. They require more than just knowledge—they demand agentic skills: decision-making, goal-tracking, and interaction with characters. That's what makes them a powerful testbed for real-world AI applications.
So far, Claude 3.7 Sonnet has only defeated a few gym leaders, but that's already a leap forward. The previous version struggled to leave the starting house in Pallet Town.
Even more impressive is Google's Gemini 2.5 Pro. Since early April, an AI enthusiast has been livestreaming the model as it plays Pokémon Blue — a slightly modified version of Red. According to Reddit users, Gemini has already made it much farther than Claude did over the same timeframe. Then again, Gemini does get a little help: a custom mini-map helps it avoid getting stuck.
More on the topic:
🔴 Pokémon Go data is being used to train AI
💎 Google DeepMind's AI plays Minecraft
#news #games @hiaimediaen
16.04.202507:00
👐 Why Will Our Children Pity Us? A New Interview with Sam Altman
How does OpenAI plan to handle copyright? Is money corrupting the company? And what are the internal models really capable of?
TED curator Chris Anderson asked OpenAI CEO Sam Altman a series of unusually tough questions about power, safety, and the future of intelligence.
Here's what Altman revealed:
💬 On Copyright: ChatGPT does not generate in the style of living artists. Instead, it draws on broader "vibes," schools, and artistic movements. In the future, creators might opt-in to allow their styles to be used—and receive compensation. But, as Altman admitted, "how you divvy up the money" is still an open question.
💬 On the DeepSeek Challenge: OpenAI is working on a powerful open-source model, near the frontier. When asked why the company continues spending billions on proprietary systems while open models rapidly catch up, Altman explained: OpenAI is not just building a model—it's building a product. And in that product safety isn't an add-on, it's foundational.
💬 On AGI: AGI isn't a single breakthrough—it's a curve. Models will keep getting smarter than humans. What matters is learning how to manage that power, step by step. Or, as he joked:
💬 On the Future: Altman became a father in February 2025. He believes his son will grow up in a world where children are never smarter than AI—and where services by default understand people instantly. He hopes future generations will look back on us with pity and nostalgia, as those who lived before real progress began.
⁉️ One question for Sam Altman was generated by OpenAI's most advanced model, o1 pro:
Altman dodged a direct answer. Instead, he returned the question to TED curator Chris Anderson: "You've been asking me versions of this for the last half hour. What do you think?"
📱 Watch the full interview here.
More interviews with Sam Altman:
💬 On the Conflict with Musk and the Future
💬 Why Now Is the Best Time to Launch an AI Startup
#news #ChatGPT @hiaimediaen
How does OpenAI plan to handle copyright? Is money corrupting the company? And what are the internal models really capable of?
TED curator Chris Anderson asked OpenAI CEO Sam Altman a series of unusually tough questions about power, safety, and the future of intelligence.
Here's what Altman revealed:
💬 On Copyright: ChatGPT does not generate in the style of living artists. Instead, it draws on broader "vibes," schools, and artistic movements. In the future, creators might opt-in to allow their styles to be used—and receive compensation. But, as Altman admitted, "how you divvy up the money" is still an open question.
💬 On the DeepSeek Challenge: OpenAI is working on a powerful open-source model, near the frontier. When asked why the company continues spending billions on proprietary systems while open models rapidly catch up, Altman explained: OpenAI is not just building a model—it's building a product. And in that product safety isn't an add-on, it's foundational.
💬 On AGI: AGI isn't a single breakthrough—it's a curve. Models will keep getting smarter than humans. What matters is learning how to manage that power, step by step. Or, as he joked:
"If you got 10 OpenAI researchers in a room and asked to define AGI, you'd get 14 definitions."
💬 On the Future: Altman became a father in February 2025. He believes his son will grow up in a world where children are never smarter than AI—and where services by default understand people instantly. He hopes future generations will look back on us with pity and nostalgia, as those who lived before real progress began.
⁉️ One question for Sam Altman was generated by OpenAI's most advanced model, o1 pro:
Sam, given that you're helping create technology that could reshape the destiny of our entire species—who granted you, or anyone, the moral authority to do that? And how are you personally accountable if you're wrong?
Altman dodged a direct answer. Instead, he returned the question to TED curator Chris Anderson: "You've been asking me versions of this for the last half hour. What do you think?"
📱 Watch the full interview here.
More interviews with Sam Altman:
💬 On the Conflict with Musk and the Future
💬 Why Now Is the Best Time to Launch an AI Startup
#news #ChatGPT @hiaimediaen
无法访问
媒体内容
媒体内容
14.04.202507:29
🤖 What Is the Neo Home Robot Already Able to Do?
The Norwegian startup 1X has been working on the AI-powered humanoid robot Neo for several years. The company has attracted more than $125 million in investment, including from OpenAI and the Tiger Global fund.
In February, the developers showed off a new home robot named Neo Gamma. The robot is about 170 centimeters tall and "wears" knitted nylon—the soft material reduces the risk of damage from human contact.
🦾 When will robots do our household chores?
So far, Neo can't do much autonomously—he walks completely on his own, opens doors, sits and stands up, and keeps his balance even if pushed. He learned basic movements by himself, repeating them in a virtual simulation many times.
Neo is gradually learning to perform more complex tasks. In the future, he will be able to do all the housework: cleaning, watering flowers, washing dishes, making tea, doing laundry, and so on.
However, such processes are difficult to simulate in a virtual environment, so developers are collecting data to train AI in real homes. An operator uses a VR headset to control the robot, showing him how to perform actions correctly. Then, during reinforcement learning, the AI brain is rewarded for correct repetitions and penalized for mistakes. This approach helps the robots gradually achieve highly realistic movements. By the end of the year, the company will start testing Neo in several hundred volunteer homes.
In the not-too-distant future, every home will have a robot helper, according to 1X CEO Bernt Børnich. Still, in order for humanoids to integrate into everyday life, they must learn alongside humans.
More on the topic:
🤖 Will Robots Replace Waiters, Bartenders, and Baristas?
🤖 Why Do We Need Humanoid Robots?
#news #robots @hiaimediaen
The Norwegian startup 1X has been working on the AI-powered humanoid robot Neo for several years. The company has attracted more than $125 million in investment, including from OpenAI and the Tiger Global fund.
In February, the developers showed off a new home robot named Neo Gamma. The robot is about 170 centimeters tall and "wears" knitted nylon—the soft material reduces the risk of damage from human contact.
🦾 When will robots do our household chores?
So far, Neo can't do much autonomously—he walks completely on his own, opens doors, sits and stands up, and keeps his balance even if pushed. He learned basic movements by himself, repeating them in a virtual simulation many times.
Neo is gradually learning to perform more complex tasks. In the future, he will be able to do all the housework: cleaning, watering flowers, washing dishes, making tea, doing laundry, and so on.
However, such processes are difficult to simulate in a virtual environment, so developers are collecting data to train AI in real homes. An operator uses a VR headset to control the robot, showing him how to perform actions correctly. Then, during reinforcement learning, the AI brain is rewarded for correct repetitions and penalized for mistakes. This approach helps the robots gradually achieve highly realistic movements. By the end of the year, the company will start testing Neo in several hundred volunteer homes.
In the not-too-distant future, every home will have a robot helper, according to 1X CEO Bernt Børnich. Still, in order for humanoids to integrate into everyday life, they must learn alongside humans.
"The home provides real-world context and the diversity of data needed for humanoids to grow in intelligence and autonomy. It also teaches them the nuances of human life—how to open the door for the elderly, move carefully around pets, or adapt to the unpredictability of the surrounding world,"he explains.
More on the topic:
🤖 Will Robots Replace Waiters, Bartenders, and Baristas?
🤖 Why Do We Need Humanoid Robots?
#news #robots @hiaimediaen
25.04.202507:04
📞 inTouch: An AI That Calls Your Parents Every Day
inTouch—An AI service from a Czech startup can call your relatives for you for $29.90 a month. The founder and CEO of inTouch, financier Vassili le Moigne, is originally from France but has lived in Prague for many years and couldn't visit his aging mother regularly. "I needed the reassurance my 88-year-old mother was safe even on days when I couldn't call her," he says.
When signing up for the app, you provide details about your relative's interests, background, career, current lifestyle, and health concerns and select how often you'd like the calls to happen. With an AI-generated voice, the bot will call the family member according to the set schedule. The algorithms will also send notifications about whether a relative picked up the phone, a summary of the call, and an overview of their mood. The AI remembers all conversations and learns in the process, adjusting to the person's age and emotional and intellectual needs.
You can also exchange voice messages with your relatives through the app. The bot will read the messages out loud and then record the responses.
Some may be confused or even offended by the idea of delegating communication with parents to AI. But the elderly get used to talking to the bot pretty quickly, the inTouch founder insists, unlike younger relatives who subscribe to the service, who still might need some time to digest the concept.
More on the topic:
🔄 Experiment: Can AI Apologize Better Than a Human?
🔄 AI Learns to Recreate Human Personality After a Two-Hour Interview
#startup #health @hiaimediaen
inTouch—An AI service from a Czech startup can call your relatives for you for $29.90 a month. The founder and CEO of inTouch, financier Vassili le Moigne, is originally from France but has lived in Prague for many years and couldn't visit his aging mother regularly. "I needed the reassurance my 88-year-old mother was safe even on days when I couldn't call her," he says.
When signing up for the app, you provide details about your relative's interests, background, career, current lifestyle, and health concerns and select how often you'd like the calls to happen. With an AI-generated voice, the bot will call the family member according to the set schedule. The algorithms will also send notifications about whether a relative picked up the phone, a summary of the call, and an overview of their mood. The AI remembers all conversations and learns in the process, adjusting to the person's age and emotional and intellectual needs.
You can also exchange voice messages with your relatives through the app. The bot will read the messages out loud and then record the responses.
Some may be confused or even offended by the idea of delegating communication with parents to AI. But the elderly get used to talking to the bot pretty quickly, the inTouch founder insists, unlike younger relatives who subscribe to the service, who still might need some time to digest the concept.
"There's a feeling of guilt that sets in, because they feel like they're outsourcing their parents to artificial intelligence. But that's not the case at all. It's actually about supplementing the calls that the children will continue to make. But with InTouch, they'll have more information about their parents' health and mood,"explains Vassili le Moigne.
More on the topic:
🔄 Experiment: Can AI Apologize Better Than a Human?
🔄 AI Learns to Recreate Human Personality After a Two-Hour Interview
#startup #health @hiaimediaen
23.04.202507:35
🤩 Ghibli, Action Figures, and AI-Photoshop: GPT Image Generator Now in Our Bot!
Hi everyone! We've added OpenAI's trending GPT Images generator, the AI tool everyone uses past month for creating viral pics, into our @GPT4Telegrambot.
Now You Can:
➡️ Turn any photo into a Ghibli Studio anime, a Pixar movie, or give it another iconic style.
➡️ Generate unique individual sticker packs, action figures, or Lego bundles.
➡️ See what your child will look like when they grow up, or imagine your pet as a person.
➡️ Add or remove objects, change backgrounds: use the bot instead of Photoshop.
➡️ Create infographics, Amazon listings, and images with text.
➡️ Bring hundreds of ideas to life with photorealistic and aestethic generation!
➡️ Go to @GPT4Telegrambot
1️⃣ Purchase the ChatGPT Plus bundle in the /premium section
2️⃣ Send /photo to the bot and select GPT-4o Images
3️⃣ Send the bot a picture to edit or type in the chat what you want to create.
🔖 Here is our top 5 creative ideas for the new image generator (with instructions)!
🔖 You can bring all your creations to life in our bot using the Kling 2.0 AI video generator.
#ChatGPT #images #Ghibli @GPT4Telegrambot
Hi everyone! We've added OpenAI's trending GPT Images generator, the AI tool everyone uses past month for creating viral pics, into our @GPT4Telegrambot.
Now You Can:
➡️ Turn any photo into a Ghibli Studio anime, a Pixar movie, or give it another iconic style.
➡️ Generate unique individual sticker packs, action figures, or Lego bundles.
➡️ See what your child will look like when they grow up, or imagine your pet as a person.
➡️ Add or remove objects, change backgrounds: use the bot instead of Photoshop.
➡️ Create infographics, Amazon listings, and images with text.
➡️ Bring hundreds of ideas to life with photorealistic and aestethic generation!
How to Use?
➡️ Go to @GPT4Telegrambot
1️⃣ Purchase the ChatGPT Plus bundle in the /premium section
2️⃣ Send /photo to the bot and select GPT-4o Images
3️⃣ Send the bot a picture to edit or type in the chat what you want to create.
🔖 Here is our top 5 creative ideas for the new image generator (with instructions)!
🔖 You can bring all your creations to life in our bot using the Kling 2.0 AI video generator.
#ChatGPT #images #Ghibli @GPT4Telegrambot
20.04.202508:30
📣 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.
▎SAVE THIS — IT'S HELPFUL
💬 Top tips from Google on how to chat with AI chatbots.
▎TO READ
🔭 A high school student received $250,000 for discovering 1.5 million space objects with AI.
⛔️ Trump may ban DeepSeek AI services in the U.S.
🤖 What can the Neo Gamma home robot already do?
🧠 A new algorithm converts thoughts into speech in near real-time.
🐬 Google has created an AI model to decode dolphin language.
🌐 Google co‑founder Sergey Brin donates billions to develop new medicines.
▎TO WATCH
📺 "My kids will never be smarter than AI": A new interview with OpenAI CEO Sam Altman.
#AIWeek @hiaimediaen
▎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
▎SAVE THIS — IT'S HELPFUL
💬 Top tips from Google on how to chat with AI chatbots.
▎TO READ
🔭 A high school student received $250,000 for discovering 1.5 million space objects with AI.
⛔️ Trump may ban DeepSeek AI services in the U.S.
🤖 What can the Neo Gamma home robot already do?
🧠 A new algorithm converts thoughts into speech in near real-time.
🐬 Google has created an AI model to decode dolphin language.
🌐 Google co‑founder Sergey Brin donates billions to develop new medicines.
▎TO WATCH
📺 "My kids will never be smarter than AI": A new interview with OpenAI CEO Sam Altman.
#AIWeek @hiaimediaen
17.04.202510:59
🔭 High School Student Received $250k for Discovering 1.5 Million Space Objects with the Help of AI
Matteo Paz, an 18-year-old high school student from California, has been interested in astronomy since childhood. From the age of 15, he took part in the space exploration programs for high school students at Caltech.
Under the guidance of his Caltech mentor, astronomer Dave Kirkpatrick, Matteo developed the VARnet model, which identifies variable cosmic objects such as supermassive black holes and supernovae. Matteo's project won a prestigious science competition for high school students with a $250,000 prize.
Paz applied the algorithm to process a huge amount of data (more than 200 terabytes) collected for 10.5 years by the now-retired infrared telescope NEOWISE. The telescope primarily observed asteroids near Earth but also detected many distant space objects that flashed, pulsated, or dimmed during eclipses.
After analyzing the archived dataset, Paz's algorithm identified 1.5 million previously unknown objects. The results are still awaiting verification, and the complete catalog will be released by the end of the year.
The AI model is in the public domain and potentially suitable for other research. According to Mateo, it can be useful for analyzing the stock market, atmospheric effects, and other phenomena where the information similarly comes in a time series.
Mateo dreams of continuing his research and, in the future, properly measuring the rate of expansion of the universe, challenging our understanding of its origin. Mateo recently got his first paid job—at his mentor's research center for various NASA projects at Caltech.
More on the topic:
➡️ Startup Mercor, Founded by Three 21-Year-Old Entrepreneurs, Valued at $2 Billion
➡️ 17-Year-Old High School Student Launches a Million-Dollar AI Business
#news #space @hiaimediaen
Matteo Paz, an 18-year-old high school student from California, has been interested in astronomy since childhood. From the age of 15, he took part in the space exploration programs for high school students at Caltech.
Under the guidance of his Caltech mentor, astronomer Dave Kirkpatrick, Matteo developed the VARnet model, which identifies variable cosmic objects such as supermassive black holes and supernovae. Matteo's project won a prestigious science competition for high school students with a $250,000 prize.
Paz applied the algorithm to process a huge amount of data (more than 200 terabytes) collected for 10.5 years by the now-retired infrared telescope NEOWISE. The telescope primarily observed asteroids near Earth but also detected many distant space objects that flashed, pulsated, or dimmed during eclipses.
After analyzing the archived dataset, Paz's algorithm identified 1.5 million previously unknown objects. The results are still awaiting verification, and the complete catalog will be released by the end of the year.
The AI model is in the public domain and potentially suitable for other research. According to Mateo, it can be useful for analyzing the stock market, atmospheric effects, and other phenomena where the information similarly comes in a time series.
Mateo dreams of continuing his research and, in the future, properly measuring the rate of expansion of the universe, challenging our understanding of its origin. Mateo recently got his first paid job—at his mentor's research center for various NASA projects at Caltech.
More on the topic:
➡️ Startup Mercor, Founded by Three 21-Year-Old Entrepreneurs, Valued at $2 Billion
➡️ 17-Year-Old High School Student Launches a Million-Dollar AI Business
#news #space @hiaimediaen
15.04.202514:30
⚡️ 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.
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.
#news #Kling @hiaimediaen
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:🔴 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.
#news #Kling @hiaimediaen
13.04.202508:30
📣 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.
#AIWeek @hiaimediaen
▎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.
#AIWeek @hiaimediaen
24.04.202510:59
🎵 A Composer's Brain Writes Music Even After His Death
Composer Alvin Lucier passed away in 2021, but he still makes music. Or rather, a tiny "brain" grown from his cells is, thanks to a team of Australian neuroscientists and musicians.
The Revivification project began in 2018. In 2020, the then 89-year-old Lucier, who had Parkinson's disease, agreed to donate his blood for the experiment.
Scientists converted his blood cells into stem cells before using them to produce cerebral organoids. These clusters of neurons stimulate brain activity. Electrodes read signals from this mini-brain, which AI processes and transfers to small hammers that pound 20 curving brass plates suspended in a vast hall, filling the area with the late composer's "music."
What's remarkable is that the organoid doesn't just emit signals—it responds to ambient sound and adapts to it. "We're very curious to see whether the organoid will change or learn over time," says project co-creator and artist Guy Ben-Ary.
Alvin Lucier, an American, was a pioneer of experimental music. In 1965, he became the first to make music using brainwaves ⤴️ He researched the impact of sound on the human psyche, lectured at universities, and was involved in creative collectives.
Ben-Ary dreams that Lucier's grown brain will go on creating "new memories and stories" forever—a vision that Lucier's daughter Amanda says aligns with his own. "This is so my dad. Just before he died he arranged for himself to play for ever. He just can't go. He needs to keep playing», — Guy cites her as saying.
📺 Ben-Ary talks more about the project on YouTube, where you can also listen to Lucier's "posthumous music."
More on the topic:
🎶 Save the Guide: How to Create a Musical Avatar in Suno
🎶 Ray Kurzweil on How Our Consciousness Will Live Forever in the Cloud
#news #music @hiaimediaen
Composer Alvin Lucier passed away in 2021, but he still makes music. Or rather, a tiny "brain" grown from his cells is, thanks to a team of Australian neuroscientists and musicians.
The Revivification project began in 2018. In 2020, the then 89-year-old Lucier, who had Parkinson's disease, agreed to donate his blood for the experiment.
Scientists converted his blood cells into stem cells before using them to produce cerebral organoids. These clusters of neurons stimulate brain activity. Electrodes read signals from this mini-brain, which AI processes and transfers to small hammers that pound 20 curving brass plates suspended in a vast hall, filling the area with the late composer's "music."
What's remarkable is that the organoid doesn't just emit signals—it responds to ambient sound and adapts to it. "We're very curious to see whether the organoid will change or learn over time," says project co-creator and artist Guy Ben-Ary.
Alvin Lucier, an American, was a pioneer of experimental music. In 1965, he became the first to make music using brainwaves ⤴️ He researched the impact of sound on the human psyche, lectured at universities, and was involved in creative collectives.
Ben-Ary dreams that Lucier's grown brain will go on creating "new memories and stories" forever—a vision that Lucier's daughter Amanda says aligns with his own. "This is so my dad. Just before he died he arranged for himself to play for ever. He just can't go. He needs to keep playing», — Guy cites her as saying.
📺 Ben-Ary talks more about the project on YouTube, where you can also listen to Lucier's "posthumous music."
More on the topic:
🎶 Save the Guide: How to Create a Musical Avatar in Suno
🎶 Ray Kurzweil on How Our Consciousness Will Live Forever in the Cloud
#news #music @hiaimediaen
22.04.202511:08
🚺 Essence: AI Assistant for Women
Hormones greatly influence a woman's health, mood, and productivity. The hormonal balance naturally fluctuates throughout the different phases of the menstrual cycle.
The AI-powered Essence app helps synchronize a woman's lifestyle based on individual data and her natural rhythms. Algorithms analyze a user's habits (diet, exercise, hobbies), physiology (allergies, symptoms, diseases), and correlate them with their hormonal cycles.
The integrated AI assistant provides daily updates on hormonal fluctuations, offering insights into possible mood changes, energy level spikes, and related symptoms.
Users can sync the service with their Google calendar to get daily personalized recommendations on how to boost productivity and psychological well-being, avoid burnout, optimize skin care and sports activities.
Essence, based in the U.S., was launched by Russian entrepreneurs Elina Valeeva and Rustam Galiev. The startup attracted €500,000 in investments from several venture funds and won the best fem-tech product title at the Med-Tech World Awards. Some large companies, such as UniCredit Bank, purchase the app for female employees as part of their corporate social package.
➡️ Essence is available on iOS and Android. Subscription costs €9,99 per month, there is a 7-day trial mode.
More on the topic:
➡️ AI Will Help Couples Conceive with IVF
➡️ AI-Dermatologist: Appointments Are Available!
#startup #health @hiaimediaen
Hormones greatly influence a woman's health, mood, and productivity. The hormonal balance naturally fluctuates throughout the different phases of the menstrual cycle.
The AI-powered Essence app helps synchronize a woman's lifestyle based on individual data and her natural rhythms. Algorithms analyze a user's habits (diet, exercise, hobbies), physiology (allergies, symptoms, diseases), and correlate them with their hormonal cycles.
The integrated AI assistant provides daily updates on hormonal fluctuations, offering insights into possible mood changes, energy level spikes, and related symptoms.
Users can sync the service with their Google calendar to get daily personalized recommendations on how to boost productivity and psychological well-being, avoid burnout, optimize skin care and sports activities.
Essence, based in the U.S., was launched by Russian entrepreneurs Elina Valeeva and Rustam Galiev. The startup attracted €500,000 in investments from several venture funds and won the best fem-tech product title at the Med-Tech World Awards. Some large companies, such as UniCredit Bank, purchase the app for female employees as part of their corporate social package.
➡️ Essence is available on iOS and Android. Subscription costs €9,99 per month, there is a 7-day trial mode.
More on the topic:
➡️ AI Will Help Couples Conceive with IVF
➡️ AI-Dermatologist: Appointments Are Available!
#startup #health @hiaimediaen
19.04.202508:34
💬 Top Tips for Communicating with LLMs from Google
Google has released a detailed 68-page guide that explains in simple terms how to effectively interact with large language models to get the most efficient and useful responses.
Here are the most valuable takeaways:
1️⃣ Use examples. A prompt without examples (zero-shot) works well only for simple questions. If the task is more complex—like writing an email or a post—add one or more examples (one-shot / few-shot) so the model can understand the task's format, style, and essence.
2️⃣ For complex tasks, use a Chain of Thought. Add this phrase to your prompt: "Let's think step by step."
3️⃣ Provide clear and concise instructions. If a prompt is hard for a human to understand, it will also confuse the LLM.
❌ "I'm currently in New York, and I'd like to learn more about the attractions. I'm traveling with two three-year-old kids. Where should we go during our vacation?"
✅ "Act as a tour guide for tourists. Describe great places to visit in Manhattan with a three-year-old child."
4️⃣ Control response length. Specify the desired length of the response to get an appropriate answer. For example: "Explain quantum physics in a way that fits into a single tweet."
5️⃣ Take a step back. Start by asking for general ideas—this activates the model's knowledge. Then, use those ideas as context so no details are missed.
Step 1: "Based on popular first-person shooter games, name 5 fictional key locations that would make a level's storyline engaging and rich."
Step 2: "Using these themes, write one paragraph of a storyline for the location 'Abandoned Military Base.'"
6️⃣ Experiment with styles. The same query can be phrased as a question, statement, or instruction—and yield different results. This is especially useful when creating creative texts.
• "Why was this console revolutionary?"
• "Write a paragraph about the Sega Dreamcast and its features."
• Or start with: "Sega released the console in 1999. It..."
More useful resources:
📌 Interesting lectures on LLMs and machine learning
📌 Top 5 free AI services for summarizing YouTube videos
#top #education #Google @hiaimediaen
Google has released a detailed 68-page guide that explains in simple terms how to effectively interact with large language models to get the most efficient and useful responses.
Here are the most valuable takeaways:
1️⃣ Use examples. A prompt without examples (zero-shot) works well only for simple questions. If the task is more complex—like writing an email or a post—add one or more examples (one-shot / few-shot) so the model can understand the task's format, style, and essence.
2️⃣ For complex tasks, use a Chain of Thought. Add this phrase to your prompt: "Let's think step by step."
3️⃣ Provide clear and concise instructions. If a prompt is hard for a human to understand, it will also confuse the LLM.
Example:
❌ "I'm currently in New York, and I'd like to learn more about the attractions. I'm traveling with two three-year-old kids. Where should we go during our vacation?"
✅ "Act as a tour guide for tourists. Describe great places to visit in Manhattan with a three-year-old child."
4️⃣ Control response length. Specify the desired length of the response to get an appropriate answer. For example: "Explain quantum physics in a way that fits into a single tweet."
5️⃣ Take a step back. Start by asking for general ideas—this activates the model's knowledge. Then, use those ideas as context so no details are missed.
Example:
Step 1: "Based on popular first-person shooter games, name 5 fictional key locations that would make a level's storyline engaging and rich."
Step 2: "Using these themes, write one paragraph of a storyline for the location 'Abandoned Military Base.'"
6️⃣ Experiment with styles. The same query can be phrased as a question, statement, or instruction—and yield different results. This is especially useful when creating creative texts.
For example,to gather information for a text about the Sega Dreamcast, you can ask:
• "Why was this console revolutionary?"
• "Write a paragraph about the Sega Dreamcast and its features."
• Or start with: "Sega released the console in 1999. It..."
More useful resources:
📌 Interesting lectures on LLMs and machine learning
📌 Top 5 free AI services for summarizing YouTube videos
#top #education #Google @hiaimediaen
17.04.202504:50
👐 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.
What's New:
➡️ Record-breaking performance. The models show unprecedented results in math, coding, and scientific knowledge benchmarks. While Gemini 2.5 Pro might compete with o4-mini in some tests, only OpenAI's Deep Research rivals the tool-using o3.
➡️ On the Codeforces platform, both models score over 2700 ELO points, which is equal to the ranking of the top 200 coders in the world.
➡️ Models think with images, not just chains of sentences. For example, they can rotate or zoom in on an image during response generation to study it better, use Python for calculations and data analysis, and apply other tools.
➡️ The models reason deeper, make fewer errors, and avoid hallucinations better.
🔴 You can already try the world's most powerful AI systems o3 and o4-mini in our bot!
➡️ 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
#news #ChatGPT @GPT4Telegrambot
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."
What's New:
➡️ Record-breaking performance. The models show unprecedented results in math, coding, and scientific knowledge benchmarks. While Gemini 2.5 Pro might compete with o4-mini in some tests, only OpenAI's Deep Research rivals the tool-using o3.
➡️ On the Codeforces platform, both models score over 2700 ELO points, which is equal to the ranking of the top 200 coders in the world.
➡️ Models think with images, not just chains of sentences. For example, they can rotate or zoom in on an image during response generation to study it better, use Python for calculations and data analysis, and apply other tools.
➡️ The models reason deeper, make fewer errors, and avoid hallucinations better.
🔴 You can already try the world's most powerful AI systems o3 and o4-mini in our bot!
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
#news #ChatGPT @GPT4Telegrambot
15.04.202510:59
🌟 What Are Dolphins Saying? Google's Model Tries to Find Out
Researchers from Google DeepMind and Georgia Tech created DolphinGemma, an AI model designed to help decode dolphins' "language."
Scientists have long observed that dolphins send signals to each other while playing, hunting, or traveling. Some believe that dolphins may even form full "sentences," which could suggest they have a simple form of language.
🐬 What Did Google Do?
To decode dolphin speech, Google DeepMind trained a model of about 400 million parameters based on Gemma architecture, using annotated dolphin recordings collected over 40 years by the Wild Dolphin Project.
The key innovation is SoundStream, a tool that turns dolphin sounds into tokens the model can understand. It listens to whistles and clicks and tries to predict what sound might come next. It works a lot like how LLMs autocomplete human sentences.
The model can generate dolphin-like responses, but we still don't know what those sounds mean. The main goal is to link sound patterns with dolphin behavior—and, eventually, decode their communication.
More on the topic:
🔜 MeowTalk: AI Cat-to-Human Translator
🔜 Lost in Translation: Will AI soon help us talk to animals?
#news #Google @hiaimediaen
Researchers from Google DeepMind and Georgia Tech created DolphinGemma, an AI model designed to help decode dolphins' "language."
Scientists have long observed that dolphins send signals to each other while playing, hunting, or traveling. Some believe that dolphins may even form full "sentences," which could suggest they have a simple form of language.
"If dolphins have language, then they probably also have culture,"notes Thad Starner, a research scientist at Google DeepMind.
🐬 What Did Google Do?
To decode dolphin speech, Google DeepMind trained a model of about 400 million parameters based on Gemma architecture, using annotated dolphin recordings collected over 40 years by the Wild Dolphin Project.
The key innovation is SoundStream, a tool that turns dolphin sounds into tokens the model can understand. It listens to whistles and clicks and tries to predict what sound might come next. It works a lot like how LLMs autocomplete human sentences.
The model can generate dolphin-like responses, but we still don't know what those sounds mean. The main goal is to link sound patterns with dolphin behavior—and, eventually, decode their communication.
More on the topic:
🔜 MeowTalk: AI Cat-to-Human Translator
🔜 Lost in Translation: Will AI soon help us talk to animals?
#news #Google @hiaimediaen
12.04.202508:01
💿 "Tron: Ares": What Awaits the Audience in the Continuation of the Legendary Saga
Disney unveiled the first trailer for the third installment of the Tron series at Comicon. While details are still scarce, it's already clear that the new chapter will mirror the story of the previous two films: this time, a program enters the real world.
Ares is the first AI in history that humanity will encounter face-to-face. Alongside him, other programs break into reality, seeking to unify everything in their path and create a digital "Matrix" in the physical world. In the trailer, they are represented by a massive flying ship called the Recognizer.
Fans can also expect the iconic light cycle chases, which have become a hallmark of the series 🟡
Jared Leto stars as Ares, while Jeff Bridges, who played the lead role in the original film, will return to the saga.
The film is set to premiere in October 2025.
Are you excited about Tron: Ares?
👾 — yes, I'm a true fan of the saga!
❤️ — maybe I'll check it out...
🤔 — haven't seen the previous films...
#mustsee @hiaimediaen
Disney unveiled the first trailer for the third installment of the Tron series at Comicon. While details are still scarce, it's already clear that the new chapter will mirror the story of the previous two films: this time, a program enters the real world.
Ares is the first AI in history that humanity will encounter face-to-face. Alongside him, other programs break into reality, seeking to unify everything in their path and create a digital "Matrix" in the physical world. In the trailer, they are represented by a massive flying ship called the Recognizer.
Fans can also expect the iconic light cycle chases, which have become a hallmark of the series 🟡
Jared Leto stars as Ares, while Jeff Bridges, who played the lead role in the original film, will return to the saga.
The firstTron film from 1982 tells the story of an obsessed programmer who gets trapped inside a program he created. In the 2010 sequel, the protagonist's son accidentally enters the Grid while searching for his father and discovers that an antivirus program has established a tyranny in this world, while his father has become a "digital dissident."
The film is set to premiere in October 2025.
Are you excited about Tron: Ares?
👾 — yes, I'm a true fan of the saga!
❤️ — maybe I'll check it out...
🤔 — haven't seen the previous films...
#mustsee @hiaimediaen
24.04.202506:59
🦿 Robots Ran Among Humans in Humanoid Half Marathon
On April 19, 21 robots competed alongside 12,000 human marathoners in a 21.1 km endurance race in Beijing. The bots had a separate track but faced the same hills, heat, and crowds. Some ran fully on their own, others were remotely controlled. All had engineers nearby for support.
Only six robots finished the race. The winner, Tiangong Ultra, created by the X-Humanoid researchers, crossed the finish line in 2 hours and 40 minutes, including battery swaps. That's just 2.5 times slower than the winning human, who ran it in 1 hour and 2 minutes.
Meanwhile, the Unitree G1—famous for its kung-fu moves in promo videos—fell right at the start. The company later said the robot wasn't using its official software.
🏅 Rapid Progress
The Beijing Half Marathon was the largest robot race ever, and it was the first time they ran alongside humans, but it was not the first time robots raced. In 2011, Japan hosted a marathon for five small humanoid robots. They had to run 422 laps of 100 meters indoors. The winner took almost 55 hours to finish.
In 2021, a U.S. robot named Cassie (essentially just a pair of legs) ran 5 km autonomously without human assistance. A year later, it set a Guinness World Record by running 100 meters in 24.73 seconds.
😅 Not Just For Fun
Robot races are real-life tests for how robots walk, balance, navigate, and manage battery life. The same tech could eventually be used in delivery, construction, healthcare, or space missions.
More on the topic:
🤖 What Is the Neo Home Robot Already Able to Do?
🤖 Why Do We Need Humanoid Robots?
#news #robots @hiaimediaen
On April 19, 21 robots competed alongside 12,000 human marathoners in a 21.1 km endurance race in Beijing. The bots had a separate track but faced the same hills, heat, and crowds. Some ran fully on their own, others were remotely controlled. All had engineers nearby for support.
Only six robots finished the race. The winner, Tiangong Ultra, created by the X-Humanoid researchers, crossed the finish line in 2 hours and 40 minutes, including battery swaps. That's just 2.5 times slower than the winning human, who ran it in 1 hour and 2 minutes.
Meanwhile, the Unitree G1—famous for its kung-fu moves in promo videos—fell right at the start. The company later said the robot wasn't using its official software.
🏅 Rapid Progress
The Beijing Half Marathon was the largest robot race ever, and it was the first time they ran alongside humans, but it was not the first time robots raced. In 2011, Japan hosted a marathon for five small humanoid robots. They had to run 422 laps of 100 meters indoors. The winner took almost 55 hours to finish.
In 2021, a U.S. robot named Cassie (essentially just a pair of legs) ran 5 km autonomously without human assistance. A year later, it set a Guinness World Record by running 100 meters in 24.73 seconds.
😅 Not Just For Fun
Robot races are real-life tests for how robots walk, balance, navigate, and manage battery life. The same tech could eventually be used in delivery, construction, healthcare, or space missions.
More on the topic:
🤖 What Is the Neo Home Robot Already Able to Do?
🤖 Why Do We Need Humanoid Robots?
#news #robots @hiaimediaen
22.04.202506:59
🙏 Being Polite To ChatGPT Costs OpenAI Tens of Millions
OpenAI CEO Sam Altman revealed on X that polite user prompts to ChatGPT cost the company tens of millions of dollars. "Well spent—you never know [what might happen]," he added, hinting at a future where rudeness to AI might backfire.
While Altman's tone was ironic and not meant to provide exact figures, his comment sparked a debate: should we waste resources being polite to a machine?
💰 Where Do the Costs Come From?
Each prompt to a chatbot triggers a chain of computations: chips spinning up, cooling systems running, and other infrastructure working in data centers. Even a few "extra" tokens in a prompt increase server load and power consumption.
Researchers at Epoch AI estimated that a single GPT-4o query requires about 0.3 watt-hours of electricity—enough to boil a teaspoon of water. Words like "please" and "thank you" are a small part of a conversation. Still, across billions of interactions, they add up to megawatts of power—and serious operating costs.
🙏🏻 Who Says "Thanks" to AI?
In the U.S., 67% of people who use AI are polite to it. Of those, 55% say it's simply "the right thing to do," and 12% say it's just in case AI or robots ever become conscious and remember who treated them with respect. Some Reddit and X users say it plainly feels wrong to be rude, even to an algorithm.
And maybe they're onto something: research shows that politeness can improve AI responses. Curtis Beavers, a director on the design team for Microsoft Copilot, explains that LLMs mirror the user's tone, responding not only politely but also more professionally. We covered this in more detail here.
So, are you polite to AI?
❤️ — yes, always!
🤔 — sometimes, if it answers good
🎃 — nope, it's just a code!
#news #OpenAI #ChatGPT @hiaimediaen
OpenAI CEO Sam Altman revealed on X that polite user prompts to ChatGPT cost the company tens of millions of dollars. "Well spent—you never know [what might happen]," he added, hinting at a future where rudeness to AI might backfire.
While Altman's tone was ironic and not meant to provide exact figures, his comment sparked a debate: should we waste resources being polite to a machine?
💰 Where Do the Costs Come From?
Each prompt to a chatbot triggers a chain of computations: chips spinning up, cooling systems running, and other infrastructure working in data centers. Even a few "extra" tokens in a prompt increase server load and power consumption.
Researchers at Epoch AI estimated that a single GPT-4o query requires about 0.3 watt-hours of electricity—enough to boil a teaspoon of water. Words like "please" and "thank you" are a small part of a conversation. Still, across billions of interactions, they add up to megawatts of power—and serious operating costs.
🙏🏻 Who Says "Thanks" to AI?
In the U.S., 67% of people who use AI are polite to it. Of those, 55% say it's simply "the right thing to do," and 12% say it's just in case AI or robots ever become conscious and remember who treated them with respect. Some Reddit and X users say it plainly feels wrong to be rude, even to an algorithm.
And maybe they're onto something: research shows that politeness can improve AI responses. Curtis Beavers, a director on the design team for Microsoft Copilot, explains that LLMs mirror the user's tone, responding not only politely but also more professionally. We covered this in more detail here.
So, are you polite to AI?
❤️ — yes, always!
🤔 — sometimes, if it answers good
🎃 — nope, it's just a code!
#news #OpenAI #ChatGPT @hiaimediaen
18.04.202510:59
🌐 Google Co-founder Sergey Brin: From Google Search to Drug Discovery
At the end of 2019, Sergey Brin, currently the eighth richest person in the world with an estimated fortune of about $120 billion, stepped down as president of Alphabet (Google's parent company) to focus on addressing fundamental global challenges: finding cures for incurable diseases and combating climate change.
💰 Brin has already invested approximately $1.8 billion in Parkinson's disease research, making it one of the largest sums ever donated to fight a specific illness. Brin himself is a carrier of a rare genetic mutation that increases his risk of developing Parkinson's. He inherited this gene from his mother, who suffered from the disease for over 20 years.
💰 Brin's other philanthropic efforts are also closely tied to his personal life. He has donated nearly $75 million to bipolar disorder research (a condition affecting a family member) and almost $50 million to autism research, as his youngest daughter has been diagnosed with the condition.
💰 Through Catalyst4, the nonprofit organization he founded in 2019, Brin funds various startups. These startups mainly work on projects ranging from treatments for neurological diseases to developing more effective vaccines and gene and cell therapies.
💰 Brin has also invested $2 billion in green energy projects, particularly those focused on hydrogen extraction and carbon capture technologies to combat climate change.
Despite stepping back from Google's day-to-day leadership, Sergey Brin retains his position as a board member and has significant influence over the company's decisions. It is rumored that the AI race has motivated the billionaire to take a more active role in corporate affairs once again.
Recently, Brin reportedly emailed employees urging them to return to the office daily and commit to working up to 60 hours per week.
More on the topic:
🌟 Gemini 2.5 Pro: Google's New Champion LLM
🌟 Elon Musk VS Larry Page
#startup #google @hiaimediaen
At the end of 2019, Sergey Brin, currently the eighth richest person in the world with an estimated fortune of about $120 billion, stepped down as president of Alphabet (Google's parent company) to focus on addressing fundamental global challenges: finding cures for incurable diseases and combating climate change.
💰 Brin has already invested approximately $1.8 billion in Parkinson's disease research, making it one of the largest sums ever donated to fight a specific illness. Brin himself is a carrier of a rare genetic mutation that increases his risk of developing Parkinson's. He inherited this gene from his mother, who suffered from the disease for over 20 years.
💰 Brin's other philanthropic efforts are also closely tied to his personal life. He has donated nearly $75 million to bipolar disorder research (a condition affecting a family member) and almost $50 million to autism research, as his youngest daughter has been diagnosed with the condition.
💰 Through Catalyst4, the nonprofit organization he founded in 2019, Brin funds various startups. These startups mainly work on projects ranging from treatments for neurological diseases to developing more effective vaccines and gene and cell therapies.
💰 Brin has also invested $2 billion in green energy projects, particularly those focused on hydrogen extraction and carbon capture technologies to combat climate change.
Despite stepping back from Google's day-to-day leadership, Sergey Brin retains his position as a board member and has significant influence over the company's decisions. It is rumored that the AI race has motivated the billionaire to take a more active role in corporate affairs once again.
Recently, Brin reportedly emailed employees urging them to return to the office daily and commit to working up to 60 hours per week.
"Competition has accelerated immensely and the final race to A.G.I. is afoot. I think we have all the ingredients to win this race, but we are going to have to turbocharge our efforts," he wrote.
More on the topic:
🌟 Gemini 2.5 Pro: Google's New Champion LLM
🌟 Elon Musk VS Larry Page
#startup #google @hiaimediaen
16.04.202511:01
💬 A New Algorithm Decodes Thoughts into Speech in Near Real-Time
Scientists from the University of California, San Francisco, tested their technology on Ann, the patient, who has been paralyzed for almost 20 years after a stroke. In 2023, the same team of researchers implanted a device on the surface of the brain's speech region to record the neurons' activity.
An algorithm then converted these signals into real speech. Initial experiments two years ago allowed Ann to speak at a rate of up to 80 words per minute. This is about twice as slow as a natural conversation (about 160 words per minute) but times faster than her current assistant communication device. However, there was almost an eight-second delay between her thoughts and the sounds, which made the speech look more like messaging than a conversation.
The new method reduces this delay between an intent to speak and the words to less than one second. It allows one to communicate in a much more natural format. This was achieved by a speech decoding algorithm similar to those used in voice assistants such as Alexa or Siri, says senior study author Gopala Anumanchipalli.
The AI model that streams Ann's voice was trained on recordings from her wedding video made before the stroke. The model even synthesized words that were not part of the training dataset vocabulary.
🧠 One of the key research team members, neurosurgeon Edward Chang, recently became the head of Echo Neurotechnologies Corp., a company developing implants to connect the brain to a computer. Here, we told more about this startup, which had recently received $50 million in funding.
#news #health @hiaimediaen
Scientists from the University of California, San Francisco, tested their technology on Ann, the patient, who has been paralyzed for almost 20 years after a stroke. In 2023, the same team of researchers implanted a device on the surface of the brain's speech region to record the neurons' activity.
An algorithm then converted these signals into real speech. Initial experiments two years ago allowed Ann to speak at a rate of up to 80 words per minute. This is about twice as slow as a natural conversation (about 160 words per minute) but times faster than her current assistant communication device. However, there was almost an eight-second delay between her thoughts and the sounds, which made the speech look more like messaging than a conversation.
The new method reduces this delay between an intent to speak and the words to less than one second. It allows one to communicate in a much more natural format. This was achieved by a speech decoding algorithm similar to those used in voice assistants such as Alexa or Siri, says senior study author Gopala Anumanchipalli.
The AI model that streams Ann's voice was trained on recordings from her wedding video made before the stroke. The model even synthesized words that were not part of the training dataset vocabulary.
🧠 One of the key research team members, neurosurgeon Edward Chang, recently became the head of Echo Neurotechnologies Corp., a company developing implants to connect the brain to a computer. Here, we told more about this startup, which had recently received $50 million in funding.
#news #health @hiaimediaen
15.04.202505:59
👐 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.
Why choose GPT-4.1 mini:
➡️ Outperforms the previous small model, GPT-4o mini, in all tasks.
➡️ Free for all users of our bot.
2️⃣ Select the GPT-4.1 mini in the /model section
🧑💻 Developers have already announced that the GPT-4.5-preview model will be removed from the API in July, and they recommend switching to GPT-4.1.
#OpenAI #ChatGPT @hiaimediaen
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
Why choose GPT-4.1 mini:
➡️ Outperforms the previous small model, GPT-4o mini, in all tasks.
➡️ Free for all users of our bot.
How to Use?1️⃣ Go to @GPT4Telegrambot
2️⃣ Select the GPT-4.1 mini in the /model section
🧑💻 Developers have already announced that the GPT-4.5-preview model will be removed from the API in July, and they recommend switching to GPT-4.1.
#OpenAI #ChatGPT @hiaimediaen
11.04.202510:59
👁 Humanity's Golden Age or its End? Where AI Will Take Us
In 2021, a year before the release of ChatGPT, former OpenAI Governance team employee Daniel Kokotajlo predicted that within five years, AI algorithms would learn to create realistic photos and videos, be able to diagnose more accurately than doctors, hit a scaling "wall," and AI agents will start its expansion.
In April 2024, he left OpenAI due to disagreements with the company's AI safety policies. Teaming up with a group of researchers, he has recently unveiled a new report with forecasts about AI development for the next five years. Its main characters are the fictional US company OpenBrain and its Chinese competitor DeepCent.
⌚️ Main predictions:
2025-2026: AI agents become more autonomous and master all programming languages. OpenBrain builds the largest data center in history, and DeepCent opens a computing cluster at one of its most powerful nuclear power plants.
2027: OpenBrain releases an AI agent on a new architecture, which allows it to continuously improve by learning "online." China secretly steals the model's data but lacks the capacity to run the system fully. Meanwhile, OpenBrain claims to have achieved AGI and releases a mini-version of the next agent for everyone.
2028-2029: The arrival of superintelligence. AI researchers will achieve efficiency levels 40 times greater than humans, paving the way for groundbreaking advancements in fusion technology, quantum computing, and the development of highly effective pharmaceuticals.
The AI economy is growing rapidly: businesses and laboratories run entirely by artificial intelligence are emerging worldwide, and poverty is eliminated with universal basic income.
😎 Utopia or cyberpunk?
According to Kokotajlo and his team, if governments can overcome conflicts, agree, and ensure transparent use of superintelligence, then a technological breakthrough in the 2030s could boost the colonization of the Solar System.
The apocalyptic scenario includes the arrival of a reality similar to The Matrix movie, where artificial intelligence enslaves and potentially destroys humans and all living organisms.
Kokotajlo himself refers to himself as a techno-pessimist. He believes that AI will catastrophically harm humanity with a 70% probability.
What do you think: will AI lead us to prosperity or a catastrophe?
👍 — I'm a techno-optimist; AI is a breakthrough
😈 — I'm a techno-pessimist; AI is a threat
🤔 — it's hard to say definitively
#news #forecasts #OpenAI @hiaimediaen
In 2021, a year before the release of ChatGPT, former OpenAI Governance team employee Daniel Kokotajlo predicted that within five years, AI algorithms would learn to create realistic photos and videos, be able to diagnose more accurately than doctors, hit a scaling "wall," and AI agents will start its expansion.
In April 2024, he left OpenAI due to disagreements with the company's AI safety policies. Teaming up with a group of researchers, he has recently unveiled a new report with forecasts about AI development for the next five years. Its main characters are the fictional US company OpenBrain and its Chinese competitor DeepCent.
⌚️ Main predictions:
2025-2026: AI agents become more autonomous and master all programming languages. OpenBrain builds the largest data center in history, and DeepCent opens a computing cluster at one of its most powerful nuclear power plants.
2027: OpenBrain releases an AI agent on a new architecture, which allows it to continuously improve by learning "online." China secretly steals the model's data but lacks the capacity to run the system fully. Meanwhile, OpenBrain claims to have achieved AGI and releases a mini-version of the next agent for everyone.
Kokotajlo's team believes that by the end of 2027, AI systems will surpass humans in everything.
2028-2029: The arrival of superintelligence. AI researchers will achieve efficiency levels 40 times greater than humans, paving the way for groundbreaking advancements in fusion technology, quantum computing, and the development of highly effective pharmaceuticals.
The AI economy is growing rapidly: businesses and laboratories run entirely by artificial intelligence are emerging worldwide, and poverty is eliminated with universal basic income.
😎 Utopia or cyberpunk?
According to Kokotajlo and his team, if governments can overcome conflicts, agree, and ensure transparent use of superintelligence, then a technological breakthrough in the 2030s could boost the colonization of the Solar System.
The apocalyptic scenario includes the arrival of a reality similar to The Matrix movie, where artificial intelligence enslaves and potentially destroys humans and all living organisms.
Kokotajlo himself refers to himself as a techno-pessimist. He believes that AI will catastrophically harm humanity with a 70% probability.
What do you think: will AI lead us to prosperity or a catastrophe?
👍 — I'm a techno-optimist; AI is a breakthrough
😈 — I'm a techno-pessimist; AI is a threat
🤔 — it's hard to say definitively
#news #forecasts #OpenAI @hiaimediaen
显示 1 - 24 共 182
登录以解锁更多功能。