Artificial Intelligence (AI) continues to evolve, with systems becoming increasingly intelligent and capable. OpenAI has recently announced a new initiative called ‘Superalignment’.
This initiative aims to steer and control AI systems that are much smarter than us, addressing a critical challenge in the AI landscape.
Superintelligence, the concept of machines that surpass human intelligence, is predicted to be the most impactful technology humanity has ever invented. It holds the potential to solve many of the world’s most pressing problems.
However, the vast power of superintelligence also poses significant risks, including the potential disempowerment of humanity or even human extinction.
OpenAI believes that superintelligence could arrive this decade, and managing its risks will require new institutions for governance and solutions for superintelligence alignment. The key question is: How do we ensure AI systems much smarter than humans follow human intent?
Currently, there is no solution for steering or controlling a potentially superintelligent AI, and preventing it from going rogue. Existing techniques for aligning AI, such as reinforcement learning from human feedback, rely on humans’ ability to supervise AI.
However, these techniques will not scale to superintelligence as humans won’t be able to reliably supervise AI systems much smarter than us.
Superalignment – OpenAI’s new project
To address this, OpenAI’s approach is to build a roughly human-level automated alignment researcher. This researcher can then use vast amounts of compute to scale efforts and iteratively align superintelligence.
The process involves developing a scalable training method, validating the resulting model, and stress testing the entire alignment pipeline.
OpenAI is assembling a team of top machine learning researchers and engineers to work on this problem, dedicating 20% of the compute they’ve secured to date over the next four years to solving the problem of superintelligence alignment.
The goal is to solve the core technical challenges of superintelligence alignment in four years.
Could you be that person? If so, OpenAI are looking for you to join their team. Check out this post on their website if you think you’ve got what it takes.
Keep up to date with all the latest AI news from TrufflerAI