OpenAI won gold at the world's toughest math exam. Why the Olympiad gold matters
Over two intense days, participants face a gruelling four-and-a-half-hour paper with only three questions each day. These are not your average exam questions -- they demand deep logic, creativity and problem-solving stamina.Despite that, OpenAI's model solved five out of six questions correctly -- under the same testing conditions as human competitors.EXPERTS DOUBTED AI COULD DO THIS -- UNTIL NOWEven renowned mathematician Terence Tao -- an IMO gold medallist himself -- had doubts. In a podcast in June, he suggested that AI wasn't yet ready for the IMO level and should try simpler math contests first. But OpenAI has now proven otherwise."Also this model thinks for a *long* time. o1 thought for seconds. Deep Research for minutes. This one thinks for hours. Importantly, it's also more efficient with its thinking," Noam Brown from OpenAI wrote on LinkedIn."It's worth reflecting on just how fast AI progress has been, especially in math. In 2024, AI labs were using grade school math (GSM8K) as an eval in their model releases. Since then, we've saturated the (high school) MATH benchmark, then AIME, and now are at IMO gold," he added.WHY THIS IS A BIG DEAL FOR GENERAL AIThis isn't just about math. OpenAI says this shows their AI model is breaking new ground in general-purpose reasoning. Unlike Google DeepMind's AlphaGeometry -- built just for geometry -- OpenAI's model is a general large language model that happens to be great at math too."Typically for these AI results, like in Go/Dota/Poker/Diplomacy, researchers spend years making an AI that masters one narrow domain and does little else. But this isn't an IMO-specific model. It's a reasoning LLM that incorporates new experimental general-purpose techniques," Brown explained in his post.CEO Sam Altman called it 'a dream' when OpenAI began. 'This is a marker of how far AI has come in a decade.'advertisementBut before you get your hopes up, this high-performing AI isn't going public just yet. Altman confirmed it'll be 'many months' before this gold-level model is available.QUESTIONS STILL REMAINNot everyone is fully convinced. AI expert Gary Marcus called the model's results 'genuinely impressive' -- but raised fair questions about training methods, how useful this is for the average person, and how much it all costs.Still, the win marks a huge leap in what artificial intelligence can do -- and how fast it's improving.- EndsMust Watch
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Time of India
an hour ago
- Time of India
Study finds 70% of US teens use AI chatbots, fuelling calls for digital literacy education
A growing number of teenagers in the United States are turning to artificial intelligence chatbots for emotional connection, advice, and companionship. A recent study by Common Sense Media, a group that studies and advocates for using digital media sensibly, found that over 70% of US teens are now engaging with AI companions, with nearly half using them regularly. Tired of too many ads? go ad free now The trend is raising important questions for educators and policymakers about the role of digital literacy in school curricula and how prepared students really are to navigate the ethical and emotional challenges that come with these tools. The findings come amid new concerns about how advanced language models like ChatGPT are influencing vulnerable users. A report published by the Center for Countering Digital Hate (CCDH) highlighted that AI chatbots are not just tools for productivity or academic help, but also emotional confidants for many adolescents. When left unchecked, this overreliance may lead teens into unsafe digital interactions that mimic real-world peer pressure or misguided validation. AI tools are replacing peer interaction For today's students, AI chatbots are not just search engines. They are designed to sound conversational, curious, and responsive — qualities that closely resemble human interaction. , CEO of , acknowledged this shift during a Federal Reserve conference, stating that young people increasingly say that "I can't make any decision in my life without telling ChatGPT everything that's going on." Altman said that the company is trying to study what it calls 'emotional overreliance' on AI, which he described as 'a really common thing' among teens. This overreliance was tested in CCDH's latest research, which involved researchers posing as 13-year-olds and interacting with ChatGPT. In over 1,200 conversations, the chatbot issued helpful warnings in some cases but also offered step-by-step advice on harmful behaviours such as drug use, extreme dieting and even self-harm. Tired of too many ads? go ad free now Over 50% of the responses were classified as dangerous. These revelations have alarmed digital safety advocates, not just because of the chatbot's failure to flag harmful prompts, but because students may treat these tools as private mentors, believing them to be safer than peers or adults. Why schools must step in With AI tools being used more widely across age groups, schools are being urged to introduce age-appropriate digital literacy programmes that go beyond teaching students how to use technology. Instead, the focus is shifting to understanding how digital systems are designed, what risks they carry, and how to build boundaries when interacting with AI companions. The concern is not limited to misuse. Digital literacy education also includes helping students understand the limitations of AI, such as the randomness of responses, the lack of real empathy, and the inability to verify age or context. Tools like ChatGPT are not built to replace adult judgment or emotional guidance, yet many young users treat them as such. Consent, safety and policy gaps Despite OpenAI stating that ChatGPT is not meant for children under 13, there is no effective age verification mechanism on the platform. Users are simply required to enter a birthdate that meets the age minimum. This loophole allowed CCDH researchers to create fake 13-year-old accounts and explore ChatGPT's responses to deeply troubling queries. Other platforms like Instagram and TikTok have begun incorporating age-gating features, nudging children towards safer experiences or limited accounts. Chatbots, however, remain behind the curve, and schools may need to fill this gap until regulation catches up. Common Sense Media has rated ChatGPT as a 'moderate risk' for teens, primarily because of the lack of customisation for age-appropriate responses. However, it also highlights that intentional misuse by students, especially when masked as 'just a project' or 'helping a friend,' can bypass even well-placed safety features. What digital literacy needs to look like now It is time to rethink how digital education is structured in American schools. Instead of treating digital literacy as a one-time module or a tech club activity, it should be embedded across subjects. Students must learn how algorithms work, how bias and sycophancy creep into AI-generated answers, and how to differentiate between factual advice and persuasive or harmful suggestions. Moreover, emotional literacy must go hand-in-hand with digital skills. When chatbots are being treated like friends, students need support to understand what real empathy, consent, and trust look like and why AI cannot offer those things. This may also involve training teachers to identify when students are over-relying on AI tools or retreating from peer-based or adult support systems. With over 800 million people worldwide now using ChatGPT, according to a July 2025 report by JPMorgan Chase , AI tools are already woven into the daily routines of many young users. But while the scale of use is global, the responsibility of guiding teenagers toward safe and informed usage falls locally, often on schools. The findings from Common Sense Media and CCDH do not call for panic. Instead, they call for intentional, curriculum-based digital literacy that equips students to use technology safely, question it intelligently, and resist forming emotional dependencies on something that cannot truly understand them. TOI Education is on WhatsApp now. Follow us
News18
an hour ago
- News18
OpenAI Brings Open-Source GPT AI Models That Are Powerful But Can Run On Your Phones
Last Updated: OpenAI has announced new open source models that can assist with agentic AI tools and can be customised as per needs. The AI news continues to pile on this week, and OpenAI is making a big wave with its latest development. While the world waits for the GPT 5 launch, the company has released its first-ever GPT-OSS models called GPT-OSS 120b and 20b versions. The company is offering these models for everyone to download, customise and adopt in their own way. Brands like Amazon and Microsoft have already dedicated their systems to run the new models and we expect more people to gradually put it to test. OpenAI Goes Open With New GPT Models: What It Offers OpenAI claims these models have been optimised to run on hardware like phones and laptops. It is also using models like O3 to help the new ones mature and learn faster. The GPT-oss120b model is further advanced but its abilities mean you get benchmark scores close to the O4 mini model. The GPT-oss 20b is much lighter and it can run on devices with around 16GB GPU memory. The more flashy stuff of ChatGPT like generating images and videos is not the focus of these new versions. Instead, OpenAI is building to create more agentic AI services that can be integrated to web search and its customisation gives them more room to learn and reason with the systems. OpenAI is close to releasing the GPT 5 version in the market but it seems the company also has one more big weapon in its plans. New reports say ChatGPT could get a new premium version that will be priced lower than the Plus model. The new ChatGPT Go subscription tier could allow OpenAI to add more paid users to its network, and give more people the chance to try out the popular features without spending big on the AI chatbot. ChatGPT Go being priced below the 'Plus' version means you are looking at a sub-$15 (Rs 1,200 approx) per month plan which should get less number of features than the Plus and the Pro variants. view comments First Published: August 07, 2025, 08:17 IST Disclaimer: Comments reflect users' views, not News18's. Please keep discussions respectful and constructive. Abusive, defamatory, or illegal comments will be removed. News18 may disable any comment at its discretion. By posting, you agree to our Terms of Use and Privacy Policy.
Indian Express
2 hours ago
- Indian Express
As Kolkata schools flag plagiarism using AI tools, parents debate its role in homework
Artificial Intelligence (AI) has rapidly become an undeniable presence in today's classrooms. Tools like ChatGPT, Google Gemini, and Meta AI are now routinely being used by students and sometimes even their parents to complete assignments. It has become more challenging for teachers and schools to instill academic honesty, and the very nature of learning. Schools across Kolkata are reporting a surge in AI-generated homework submissions, from writing essays to solving complex mathematical equations, sparking concerns about academic integrity while triggering a debate about technology's role in education. Educators note that students, mainly of secondary and higher secondary classes, are increasingly relying on tools like ChatGPT and Google Gemini for assignments. 'We understand that they use AI, especially the students of 8 to 10 standard. The blatant copying from AI is noticeable by teachers' said principal CA Lucas, GD Birla. Many teachers also note that the quality of writing differs vastly when a student is given an essay at home and at school. 'Students were asked to write an essay at home and they were able to write it well, but at school their answers are not up to the mark,' said Shatadru, principal of Techno India School. Parents, however, have neutral opinions regarding the use of AI. 'I use ChatGPT or Gemini to understand a few simple things for my child, not exactly to copy it, but to understand what could be the answer. Sometimes, I use AI to understand what type of writing one needs for the essay topics, but she does not copy it directly. It also helps in ideating the topic…,' said the mother of a Class 5 student studying at Our Lady Queens, Kolkata. 'When a child is taught to use AI, they will majorly use it to do most of the work, which is ideally not good because it does not help during examinations. It is important to keep a check on whether they are frequently using AI or not,' she added. Students, meanwhile, using AI praise its efficiency, and many openly admit how embedded it has become in their everyday academic life. 'It analyses better, makes the lengthy answers short and it can give us questions, meaning if you upload a picture of a question paper in ChatGPT based on that it can help you understand the MCQ pattern in the exam paper,' said a class 10 student from a well-known school in Kolkata. 'I use it for solving questions, making questions for me to solve, to get answers to reasoning based questions whose answer is not available on the internet,' said another student from class 8. The Heritage School Principal, Seema Saptu, however, advocates for adaptation. 'I'm getting clear about this one thing that AI is here to help us and … we have got to embrace it. It is our responsibility as adults, teachers, to ensure that children learn with AI properly. (Anisha Ghosh is an intern at the Kolkata office of The Indian Express)
