OpenAI is offering $10 million (about £7 million) in grants to facilitate research that focuses on controlling artificial intelligence (AI) systems that are more intelligent than an average human.
The American AI company hopes that the results from its Superalignment Fast Grants will help better understand how AI systems can be used to assess the output of newer AI systems and play a key role in building an AI lie detector.
Evaluating millions of lines of code
According to OpenAI, fully understanding superhuman AI systems will be an arduous task. Notably, humans will not be able to reliably evaluate whether a million lines of complicated codes generated by an AI model are safe to run.
Taking to X (formerly Twitter), OpenAI shared a post stating: "Figuring out how to ensure future superhuman AI systems are aligned and safe is one of the most important unsolved technical problems in the world."
We're announcing, together with @ericschmidt: Superalignment Fast Grants.
— OpenAI (@OpenAI) December 14, 2023
$10M in grants for technical research on aligning superhuman AI systems, including weak-to-strong generalization, interpretability, scalable oversight, and more.
Apply by Feb 18! https://t.co/eCKwZWLSZE
The AI lab went on to suggest that the problem is solvable. "There is lots of low-hanging fruit, and new researchers can make enormous contributions," the company wrote.
While we are able to supervise current AI systems, OpenAI says the future superhuman AI systems will be much smarter than humans. So, the company is sparing no effort in a bid to figure out how humans can still be in charge.
Claiming the OpenAI grants
OpenAI is offering grants for individual researchers, non-profits and academic labs. Moreover, graduate students can avail of an OpenAI-sponsored $150K Superalignment Fellowship.
The company wants to support researchers who are willing to work on alignment for the first time. It is worth noting that no prior experience working on alignment is required and you can apply by February 18.
OpenAI's research implies there are 7 practices that help keep AI systems safe and accountable. Now, the company wants to fund further research that can find answers to some open questions that surfaced from the previous study.
"We are launching a program to award grants of between $10,000 - $100,000 (about £8,200 and £82,000) to fund research into the impacts of agentic AI systems and practices for making them safe," the company said.
Interestingly, OpenAI investor Vinod Khosla recently said that China is more likely to kill us than sentient AI.
Agentic AI system
Agentic AI systems allude to a superintelligent AI that can perform a wide range of actions. Moreover, it is sometimes capable of autonomously acting on complex goals on behalf of the user.
In the future, humans will need to supervise AI systems much smarter than them.
— OpenAI (@OpenAI) December 14, 2023
We study an analogy: small models supervising large models.
Read the Superalignment team's first paper showing progress on a new approach, weak-to-strong generalization: https://t.co/BiBpGdqtKj pic.twitter.com/jmbbwcrmm8
For instance, you can ask your agentic personal assistant to help with shopping for Christmas and it will print out a list of things you might want to buy and recommend websites and offline stores where you can buy these items.
Before taking advantage of agentic AI systems, OpenAI researchers suggest it is imperative for us to make them safe by minimising their failures, vulnerabilities and abuses.
This doesn't come as a surprise given that new research shows that cybercriminals can take advantage of OpenAI's AI-powered chatbot ChatGPT to commit malicious acts.
