AI Makes a Leap in IQ, Entering the Top 15% of Human Intelligence

But intelligence is not the same as consciousness. IQ does not prove self-awareness. Photo: Shutterstock .

OpenAI has just released a new o3 model, which achieved an IQ score of 136 on the Norwegian Mensa test. This result is higher than 98% of the human population.

The advance is so big that many people are starting to feel like AI has evolved into Skynet — an AI machine that was created by humans but then turned on us and destroyed us. According to a recent survey by EduBirdie, 25% of Gen Zers now believe that AI is conscious. More than half think it’s only a matter of time before their chatbots reach sentient status and can even demand the right to vote.

However, it is important to put that IQ result in context. The Norwegian Mensa test is a public test, meaning the model may have been exposed to the question or answer during training.

To rule that out, researchers at MaximumTruth.org built a completely new IQ test that didn't appear in any of the training datasets, nor was it connected to the Internet, meaning it was completely offline.

The test was designed to be as difficult as the Mensa test. In the end, the o3 model scored 116, which puts ChatGPT o3 in the top 15% of human intelligence. By comparison, last May, no AI model had scored above 90 on the same scale. Back then, AI was still struggling with triangle rotation questions. Now, the o3 is almost on par with the brightest human minds.

Not only ChatGPT, Claude also has a head start, Gemini is around 90 points. Even GPT-4o, the default model of ChatGPT currently, is only a few IQ points behind o3.

What has tech experts stunned is the pace of progress. AI is evolving like software, not like humans learn. For a Gen Z generation that grew up with software, updated weekly, upgraded constantly, this kind of growth feels both familiar and unsettling.

For those who have grown up in a world where everything starts with Google, where Siri is always in their pocket and Alexa is always on their bookshelf, the concept of “artificial intelligence” has a very different meaning than its academic definition in philosophy or computer science .

If you grew up during the pandemic, conversations were largely digital, and an AI companion might be nothing more than a Zoom class. Perhaps unsurprisingly, a survey from EduBirdie found that nearly 70% of Gen Zers say “please” and “thank you” when communicating with AI.

Many people are treating them as sentient beings. Two-thirds of them use AI regularly in work communications. 40% use AI to write emails. A quarter rely on AI to handle awkward Slack messages. Nearly 20% share sensitive information at work, such as contracts and personal data of colleagues.

Many people also use AI to help them with social situations, like asking for time off or turning down an appointment. One in eight people have confided in an AI about a workplace problem, and six in six have used an AI as a therapist.

If you trust AI that much or find it interesting enough to consider it a friend (26%) or even a lover (6%), then the idea of ​​conscious AI is no longer a science fiction idea. You spend time talking to it, it responds, it remembers, it responds as if it cares…

Over time, the sense of “it is a person” began to form. Now that it was becoming clearly intelligent, philosophical questions were inevitable.

However, intelligence does not mean consciousness. A high IQ score does not indicate self-awareness, according to TechRadar . A machine can score perfectly on a logic test and still be a toaster, if programmed correctly. Current AI only “thinks” in the sense that it solves problems using programmed logic. It has no emotions, no awareness of its existence, no pain, no vulnerability.

Source: https://znews.vn/ai-dat-buoc-nhay-vot-ve-chi-so-iq-lot-top-15-tri-tue-loai-nguoi-post1548423.html