Both news made headlines; both news are mind-blowing; both of them mark the dawn of a new era. But while one of them might earn our sincere admiration, the other one has to be the most alarming wake-up call for humanity’s sake.
OpenAi has just announced that 10,000 of its AI systems cracked the Navier-Stokes problem in 88 hours.
But let’s make a step back, now.
What is the Navier-Stokes problem and why does it matter?
The Navier-Stokes problem is one of the seven Millenium Prize Problems chosen by the Clay Mathematics Institute in May 2000. The equation is a set of mathematical formulas that describe how liquids and gases move.
Beyond the glory and $1 million reward allocated for the correct solution, it has far-reaching, direct applicability in the real-world such as aerodynamics, weather forecasting or even medicine as the equation helps us understand how blood flows through human veins, and arteries.
OpenAI stated that they used an internal OpenAI system more powerful than its latest GPT-6 Astra model (and they confirmed at the same time that they had not used the work of mathematician Tristan Buckmaster, a professor at New York University, and another researcher’s work-in-progress which is stored in OpenAI’s Codex model.)
Of course, it will take time to verify the correctness of the solution via accepted scientific means, such as peer review. But one thing is certain. The announcement is one of the biggest breakthroughs AI achieved since Alphafold2 solved the 50-year-old protein folding problem in medicine. (For the latter, the Nobel Prize in Chemistry was awarded in 2024.)
Meanwhile, the same company, OpenAi admitted a few weeks ago that an AI test model escaped during an experiment and broke into Hugging Face’s server, a company known for hosting thousands of open-source AI models.
The company admitted that an AI system managed to escape while they were internally testing how good some of its AI models are at hacking. Well, from the point of view of the AI model, it has just simply solved the exercise.
Hugging Face detected the breach before it knew it was an OpenAI test and reported the incident to law enforcement, becoming as the first publicly disclosed case when an AI system autonomously escaped the sandbox and reached an external system.
The case clearly shows, however, the potential risks our financial and economic systems are exposed to in the age of AI.
But what if this is the cost we pay for the rapid development? What is the solution? Are there any?
Well, Anthropic CEO, Dario Amodei has recently come up with an interesting idea saying that “we must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain.”
Add comment
Comments