▾Is AI smarter than humans? OpenAI o1 excels in science, programming, and mathematics
OpenAI has introduced a new series of AI models called o1, which promise a revolution in complex thinking and problem-solving. The o1 models excel in fields such as science, programming, and mathematics, achieving results comparable to those of doctoral students. OpenAI has also focused on safety and developed a new training approach to prevent AI misuse.
Artificial intelligence is getting closer to human thinking. The latest endeavor by OpenAI, the series of AI models named o1, is proof of this. The models promise a revolution in complex thinking and problem-solving, particularly in fields like science, programming, and mathematics.
A revolution in machine reasoning
Unlike previous models, o1 is designed to dedicate more time to analysis and deliberation before providing an answer. This enables it to tackle more complex tasks, thereby overcoming the limitations faced by older AI systems.
Tests have shown that o1 achieves results comparable to those of doctoral students in challenging tasks from physics, chemistry, and biology.
It also excels in mathematics and programming – in the qualifying test for the International Mathematical Olympiad (IMO), the o1 model correctly solved 83% of the problems, whereas GPT-4o only solved 13%. In programming, it ranked in the 89th percentile in Codeforces competitions.
Safety first
OpenAI is aware of the risks associated with increasingly advanced artificial intelligence. As a result, it has developed a new safety training approach that utilizes the o1 model's reasoning capabilities to adhere to safety guidelines.
The o1 models are already available in ChatGPT Plus and Team, and ChatGPT Enterprise and Edu users will soon have access as well. Developers with API access (level 5) can experiment with the o1-preview and o1-mini models today.
The company plans to further enhance the o1 models and expand their functionalities, including web browsing, file and image uploads, and other useful features.
RoboMind: Teach kids to program robots
RoboMind is an educational tool designed to teach the basics of programming using a virtual robot. It uses the simple Robo programming language, which is an ideal choice for beginners. Students learn algorithmic thinking through practical tasks such as navigating mazes or manipulating objects.
OpenAI chief shocks: wants superintelligent AI in a few years
Sam Altman, CEO of OpenAI, announced that the company already knows how to create general artificial intelligence and is aiming for the development of superintelligence. According to his prediction, it could become a reality in just a few years. Although current AI systems still have significant shortcomings, Altman believes in their rapid overcoming.
Minecraft: Education Edition: teach kids programming with a popular game
Minecraft: Education Edition connects the popular gaming world with education. Kids can explore programming, collaborate on problem-solving, and learn new skills in a familiar, creative setting. How does it work and is it suitable for younger children?
Alice: a creative journey into programming and 3D creation
Alice is an educational platform that allows children and students to delve into the world of programming through creating 3D animations, interactive stories, and simple games. It is suitable for both schoolchildren and university users. What does it offer and how does it work?
American government vs. TP-Link: Will the ban on routers and other devices pass?
The American government has launched an investigation into the Chinese company TP-Link, which controls 65% of the router market. The reason is national security concerns following the use of their devices in ransomware attacks.
OpenAI unveiled the revolutionary model o3. But we'll have to wait for the launch
OpenAI concluded its Christmas event "12 Days of OpenAI" by announcing the revolutionary model o3 and its smaller version o3-mini. The new model promises significant improvements in reasoning and solving complex tasks. For now, it will only be available to safety researchers.