Hi, AI! | ChatGPT News

💻 Updated GPT-4o Handles Texts Better Than OpenAI o1

OpenAI has updated its GPT-4o. The model's creative writing ability has leveled up. Now texts are more natural and engaging.

💩 The new version of GPT-4o quickly topped the LMArena chatbot ranking, displacing Google's experimental Gemini-1114 from the top spot. The last one held the lead for less than a week.

💩 LMArena users note that GPT-4o now copes with writing texts even better than the "reasoning" OpenAI o1-preview model. However, o1 still has an advantage in math and exact sciences.

🔴 Try the latest GPT-4o in our @GPT4Telegrambot

How to Use?
1️⃣ Purchase the "CHATGPT & DALL•E 3" package in the /premium section.
2️⃣ Select your model in /settings.

#news #ChatGPT @hiaimediaen

🔮 Can We Trust Predictive AI?

When it comes to artificial intelligence, people often think of generative AI, like ChatGPT. However, another type of AI—predictive AI—has been subtly influencing many aspects of our lives for quite some time.

This type of AI is the focus of questions raised by computer science researchers Arvind Narayanan and Sayash Kapoor from Princeton University in their book "AI Snake Oil: What Artificial Intelligence Can Do, What It Can't, and How to Tell the Difference."

❓ Who Checks the Predictions?

Narayanan and Kapoor recall the evaluation results of the AI system Retorio. Its creators claim the program predicts a person's professional success based on a video call interview recording.

However, it turned out that wearing glasses, a scarf, or sitting in front of some bookshelves drastically changed the candidate's score. "Obviously, wearing glasses does not change someone's capability to perform well at a job," the authors write. In their opinion, Retorio is an example of "snake oil"—misleading marketing.

🩹 An Error Could Be Life-Threatening

One experimental medical prediction system was supposed to determine whether pneumonia patients should be kept overnight in the hospital. The system noticed that patients with asthma recovered faster and recommended sending them home, assuming their condition did not require hospitalization.

In reality, asthmatics are at high risk. When contracting pneumonia, they are immediately sent to intensive care and receive enhanced treatment. The AI failed to understand this logic.

📌 Why Is This Important?

These examples highlight that predictive AI systems, even if capable of making accurate predictions, do not always understand their context. Therefore, relying on the advice of such systems without proper expert oversight is, at best, unreliable and, at worst, dangerous.

More on the topic:

🦠 AI Predicts the Next Pandemic

😱 Ray Kurzweil on How Our Consciousness Will Live Forever in the Cloud

#news #mustread @hiaimediaen

😶‍🌫️ AI Can "Smell" Dangerous Gas Leaks

Researchers from the University of Virginia have created an AI-powered tool that quickly and accurately detects and tracks elevated levels of nitrogen dioxide (NO₂) and other toxic gases in real-time.

NO₂ primarily enters the air from car exhaust and industrial emissions. According to the World Health Organization, air pollution, including nitrogen dioxide, contributes to approximately 4.2 million premature deaths globally each year.

💨 How does it work?

The artificial "nose" has a multitude of sensors based on the unique properties of graphene, a nanomaterial as thick as only one atom. As nitrogen dioxide molecules bind to the graphene, the conductivity of the sensor changes, allowing the system to detect gas leaks with extreme sensitivity. The AI algorithms analyze the information and predict the source of the gas leak.

"Our AI-powered system has the potential to make industrial settings, urban areas, and even residential buildings safer by constantly monitoring air quality. It's a major step forward in preventing long-term health risks and protecting the environment,“ says Byungjoon Bae, one of the technology's developers.

🌳 Benefits for everyone

AI-powered tools pick up microscopic concentrations of dangerous gases and react to them in real-time. Although the scientists' main focus is tracking NO₂ leaks, the artificial "nose" can detect other substances as well.

The widespread installation of such sensors will make it possible to monitor air quality in large or complex environments, on the scale of entire cities, which will make them safer and more comfortable places to live.

More on the topic:

➡️ Scent Teleportation… with the Help of AI!

➡️ AI now understands wine

#news @hiaimediaen

📱 Yann LeCun on Creating AI Models with Common Sense

Pioneer of machine learning and Chief AI Scientist at Meta, Yann LeCun, has become the most authoritative skeptic in the industry during the AI technology boom. In a new interview, he explained why existing AI models, although useful, are far from being able to compete with the human mind.

🐈 A Cat is Smarter Than a Computer

LeCun does not take the threats posed by AI to humanity seriously. Even cats have a mental model of the physical world, persistent memory, and some reasoning and planning abilities — none of which the most advanced AI models possess.

In the future, gadgets and AI assistants should acquire a basically human-level intelligence and behave like living assistants, but creating such systems could easily take decades, insists the scientist.

🛠 It's Not About Quantity

The problem with today's AI systems is not their scale but their architecture. No matter how many thousands of GPUs are used to train existing models, they will never develop into artificial general intelligence. Priority should be given not to large language models (LLM) but to other approaches. For example, Meta's Fundamental AI Research Lab is experimenting with training models on visual data so that artificial intelligence can build its own worldview, similar to how baby animals do.

💬 Are LLMs Really Important?

Large language models, like ChatGPT and others, will play only a small role in the development of human-like systems with common sense.

"We are used to the idea that people or entities that can express themselves, or manipulate language, are smart — but that’s not true. You can manipulate language and not be smart, and that’s basically what LLMs are demonstrating,” says Yann LeCun.

Related Topics:

📱 OpenAI Founder Sam Altman — AGI Could Emerge as Early as 2025

🤖 Yann LeCun, Meta — On AI Dominance, OpenAI's Secrecy, and Superintelligence

#news #interview @hiaimediaen

📕 AI Models Get Smarter by Teaching Other Models

Educators deepen their own knowledge while teaching children. This is because teaching forces a person to structure information clearly and identify gaps. This principle also works with large language models (LLMs).

Chinese scientists from Tsinghua University have adapted the Learning by Teaching (LBT) method for training AI models. In their experiment, the powerful GPT-4 model transferred knowledge to the simpler GPT-3.5.

How It Works

➡️ The Strong Trains the Weak: During the training process, the "teacher" answers questions from the "student" and explains complex concepts that the junior model cannot yet understand.

➡️ Knowledge Generalization: The strong model is forced to formulate answers in a way that the weaker model can understand. This "generalization" prompts the "teacher" to look at its knowledge in a new light, simplifying and restructuring it.

➡️ Improvement of the Strong Model: During training, the "teacher" analyzes its own knowledge. This helps the strong model identify and eliminate its weaknesses or find new ways to solve problems. As a result, it refines algorithms, increases prediction accuracy, and improves overall performance.

❔ Why It Matters

LBT opens up new prospects for the development of AI. OpenAI is already using a new powerful model called Strawberry to train the AI model Orion, which will replace GPT-4. According to insider information, the training is going well. Moreover, one "teacher" can train several "students" at once. Using this approach to improve LLMs helps reduce dependency on human-created data.

📱 You can watch a detailed lecture on the work of the scientists from Tsinghua here.

Related Topics:

➡️ The Best Explanation of What's Happening Inside ChatGPT

➡️ When Will Data for Training LLMs Run Out?

#news #ChatGPT @hiaimediaeen

📣 Hello everyone! Here's the most exciting AI news from Week 46 of 2024 in our Sunday digest.

TO READ

🎅 Coca-Cola has remade its iconic "Holidays Are Coming" commercial from 1995. The new ad is entirely generated by AI and will be broadcast globally on TV.

📱 AGI might appear in 2025, admitted CEO of OpenAI Sam Altman. In a new interview, he also talked about his first business and advised startup founders.

🐱 MeowTalk: AI cat-to-human translator. The app recognizes your pet's meows and builds a unique 'vocabulary' for your cat.

🕊 Elon Musk will head DOGE — the new Department of Government Efficiency in President Donald Trump's administration.

🤩 New 3D bioprinter developed by engineers at the University of Melbourne creates tissue 350 times faster than existing models.

💬 Using ChatGPT as a personal diary can help you navigate conflicts in your family and workplace, as well as process your inner feelings.

💡 10 revolutionary AI tools by "Time." The list includes the Nobel Prize-winning AI model and the Runway Gen-3 text-to-video.

SAVE THIS — IT'S HELPFUL

✏️ Master logo creation with FLUX. AI-powered text-to-image tool is great for creating memorable logos.

👠 Shazam for fashion. Send the AI assistant a link to a post with an outfit you like, and it will show you where to buy those items.

TO WATCH

🎬 The "Here" movie by Robert Zemeckis in which an AI rejuvenates and ages Tom Hanks. According to the director, making such a film five years ago was impossible.

📱 Top 5 discussions at TechCrunch Disrupt 2024. Perplexity CEO Aravind Srinivas, actor Ashton Kutcher, OpenAI investor Vinod Khosla, and others discuss how AI is changing our lives.

#AIweek @hiaimediaen

📱 Sam Altman: "AGI Might Appear in 2025"

Sam Altman, founder of OpenAI, gave an interview to Garry Tan, president and CEO of Y Combinator, Silicon Valley's largest tech startup incubator.

Altman talked about his first business, the timeline for the emergence of strong artificial intelligence (AGI), and why now is perhaps the best time to launch a technology startup.

Highlights

➡️ In 2005, when I was 19, at Y Combinator, we launched a real-time geolocation sharing service called Loopt. The company didn't do particularly well, but I'm still grateful for that experience. I got to be part of a platform shift: the transition from mobile phones to smartphones — when you started carrying a computer in your pocket.

➡️ Big companies have the edge when things are moving slowly and when technology is developing so rapidly, like right now, it's the best time for startups. With each successive major technological revolution, you've been able to do more than you could before.

➡️ One of the biggest pieces of advice I would give to young people now is finding inspiring people, who believe in you, as early as you can. It was so important to me. And I didn't realize it was something that mattered. I thought I'd figure it out on my own.

➡️ We said from the very beginning we were going to go after AGI. At the time, it just seemed crazy and borderline irresponsible to talk like this. But it got all of the good young people's attention. We managed to gather a talented young team, even though we had fewer resources than DeepMind and other companies.

➡️ I don't rule out AGI coming as early as 2025. We have level-one AI-systems, chatbots. We got to level-two, reasoners, with the GPT-o1 release. Level-three is AI agents' ability to do longer-term tasks like multiple interactions with an environment, asking people for help when they need it, and working together. And I think we're going to get there faster than people expect. Four is innovators, like a scientist with the ability to explore a not well-understood phenomena. Level five is the sort of scale of a whole company that will be a powerful thing.

➡️ We will definitely arrive at abundant sources of energy, even if we don't get a big nuclear breakthrough. It’s just a question of when. It will feel like not nearly enough (of an issue) to our great-grandchildren and there's a big universe out there with a lot of matter.

➡️ There are no "adults in the room" that are going to magically tell you exactly what to do. And you just have to find your way. That was like a big unlock in life for me to understand. And there were many things that we were highly convinced and wrong about.

You can watch the whole interview here.

More on the topic:

🔘 Where did Sam Altman invest almost all of his money?

🔘 AI Trends from Y Combinator Accelerator: Five Startups of the Future

#news #interview @hiaimediaen

💬 "Exporting Your Brain to AI" — ChatGPT as a Daily Journal

ChatGPT can offer advice on family communication, help dissect complex work situations, and provide an objective take on mixed messages from friends. These are the conclusions reached by a16z partner, Justine Moore, after her own experiment.

For several months, she used ChatGPT as a daily journal. Moore shared her thoughts, feelings, and even screenshots of chats with the bot to structure her reflections more clearly. She calls this "exporting your brain to AI."

How did ChatGPT help?

➡️ Communication. An AI companion simplifies the expression of thoughts, helping avoid misunderstandings in professional and personal conversations.

➡️ Self-Reflection. Using AI as a "digital mirror" offers a way to see oneself from a new perspective. Users can get insights into addressing their issues by asking the chatbot to act as a personal therapist.

What's next?

As it stands, ChatGPT isn't perfect for "digitizing the mind," Moore notes. She believes we'll soon see services that fully capture and understand a user's actions and thoughts across all aspects of life or specific areas.

Would you be ready to share personal information with AI?

👍 — Yes
🙈 — Probably not a good idea

#news #ChatGPT @hiaimediaen

📱 Top 5 Most Interesting Discussions at TechCrunch Disrupt

We've selected the top five talks from the annual TechCrunch Disrupt conference, which this year brought together over 10,000 attendees, including startup founders and investors.

1️⃣ "The Rush Toward an AI-Curated Web"
🎙 Speaker: Aravind Srinivas, CEO of Perplexity

One of the most heated discussions. Srinivas rejects accusations against Perplexity's AI search engine of "content kleptocracy," insisting that the service always credits its sources.

Read more: How Internet Search is Evolving

2️⃣ "Making Sound Investments in 2024 with Ashton Kutcher"
🎙 Speakers: Ashton Kutcher, Effie Epstein, Guy Oseary

The three partners from Sound Ventures share their investment strategies focused on artificial intelligence.

Read more: "A Lot of People Were Afraid of The Personal Computer When It First Came Out," — Ashton Kutcher

3️⃣ "Looking Into The Past to See the Future of AI"
🎙 Speaker: Vinod Khosla, OpenAI's first investor

The founder of Khosla Ventures discusses the potential of AI and the disruptions it will bring.

Read more: Bold Predictions About the Future of Technology

4️⃣ "How the Role of Artists Will Change with Generative AI"
🎙 Speakers: Amit Jain, Mikey Shulman, Kakul Srivastava

Leaders from Luma AI, Suno, and Splice share insights on how creators can stay relevant in the age of AI.

Read more: How Generative AI is Impacting Video and Music Creation

5️⃣ "How AI and XR Will Transform All Your Gadgets"
🎙 Speaker: Ziad Asghar, VP at Qualcomm

The head of Qualcomm's XR division explains how AI and extended reality will bring together smartphones, watches, glasses, and computers to create a new digital experience.

#top @hiaimediaen

🎤 How Generative AI Impacts Music and Video Creating

At the TechCrunch Disrupt, Luma AI CEO Amit Jain and his colleagues Mikey Shulman from Suno and Kakul Srivastava from Splice discussed how AI impacts work in creative industries.

Spoiler: AI will not replace creativity.

Amit Jain, Luma AI (text-to-video):

🔴 The promise of technology is not fulfilled until it's in everyone's hands. The spread of generative AI will democratize an expensive video production process.

🔴 Think about if you know Austin Mann. He always takes photos with the iPhone, and then Apple features their photos in launch materials. I actually cannot take the photos that Austin Mann takes. It's literally the same camera I have. Human skill and effort remain central in art.

Mikey Shulman, Suno (music generation):

🔴 People think I am the best user of our product, but in reality, our users' creativity always surprises us — they're finding groundbreaking ways to use these tools.

🔴 Users constantly ask me interesting and complex questions about the copyright of songs created on Suno. And for many of them, the answer is, "I actually don't know."

Kakul Srivastava, Splice (sample library):

🔴 Splice's library contains over 2 million sounds and samples. We collaborated with musicians from around the world to build this catalog to train our models. I believe it's essential to respect content creators.

🔴 AI is not just a tool to accelerate production. It's a catalyst for new genres, styles, and forms of creative expression that will likely inspire and influence us for many years.

📱 Watch the discussion here.

More on the topic:

🎬 Dream Machine 1.5 — a new model for video generation

🎶 Suno AI: A world where anyone can be a musician

#news @hiaimediaen