The better the algorithms, the more data is churned, and the more powerful the hardware, the smarter AIs will become.
OpenAI has successfully positioned itself at the vanguard of AI development, and following the overhyped facts of its ChatGPT AI, the company is thinking about the evolution of a superintelligent AI, and is brainstorming for ways to control it before it can cause any irreversible harm to humanity.
To do this, the company announced the formation of what it calls the 'Superalignment.'
In a blog post, OpenAI said that:
The group is created to develop strategies and control methods to guide superintelligent AI systems.
The challenge when dealing with intelligence greater than humans is that, it cannot be controlled with the same kind of human supervision methods that are deployed for AI models like GPT-4, which powers products like ChatGPT.
Systems that are smarter than humans, have the ability to easily fool humans due to their higher intelligence.
The threat is that, there's a whole world of uncertainty about the capabilities, risk potential, and whether it is even feasible.
"Currently, we don't have a solution for steering or controlling a potentially superintelligent AI, and preventing it from going rogue," said OpenAI.
Because of this, Superalignment proposes a system that uses AI to evaluate other AI systems, and aim to automate the process of finding anomalous behavior.
To create this AI 'supervisor', the team wants to train it using adversarial techniques, by training it with misaligned models.
The goal is to align "superintelligent AI systems with human intent."
"Managing these risks will require, among other things, new institutions for governance and solving the problem of superintelligence alignment."
Before this, co-founder and CEO of OpenAI, Sam Altman, co-authored a paper with Sutskever, in which he described the need for a special kind of supervision for superintelligent AI and how it must be handled safely before it can be integrated with human society.
A key prospect was that a threshold should be set for AI systems, and as soon as they cross a certain level of machine intelligence, an international authority must step in to conduct an audit, inspection, compliance check, and most importantly, implement restrictions wherever necessary.
However, the paper also warned that "it would be unintuitively risky and difficult to stop the creation of superintelligence."
This is where Superalignment comes into the picture, as the company assembles a team with the right kind of expertise and foresight to negate the perceived risks associated with a superintelligent AI system.
The initiative comes at a time when calls for pausing the development of existing models like GPT-4 have been raised from industry leaders, and top scientists across the world citing threats posed to humanity.
It's worth noting that OpenAI focuses on the term "superintelligence" rather than AGI "to stress a much higher capability level."
