India pushes in, raising concerns about AI safety
In the last two years, many new nonprofit organisations have
been created to focus on the risks of artificial intelligence (AI). These are
not AI companies building products, but groups working to make AI safer.
Some of these organisations include Fathom, Current AI, the
International Association for Safe and Ethical AI (IASEAI), and the AI Futures
Project. They work with governments and companies to create rules and safety
guidelines for AI, which is rapidly changing industries and economies.
As AI becomes more widely used, safety is becoming very
important. AI researchers, founders, and governments are now discussing how to
control risks.
Call for Global AI Regulation
At a recent AI Summit, Sam Altman, founder of OpenAI, said
powerful technologies need strong safeguards. He suggested the world may need a
global organisation for AI, similar to the International Atomic Energy Agency
(IAEA), to coordinate safety efforts.
Andrew Freedman, co-founder and CEO of Fathom, said AI is
developing so fast that governments are struggling to keep up. He believes AI
governance must be different from traditional regulation.
Concerns from AI Experts
Well-known AI researchers like Stuart Russell and Yoshua
Bengio (often called the “Godfather of AI”) have warned about serious risks
from advanced AI.
Stuart Russell, a professor at UC Berkeley and president of
IASEAI, said there are two main safety problems:
- How
safe AI systems actually are
- What
level of risk is acceptable when using them
He believes unsafe AI systems should not be released and
should require proper testing and licensing.
Yoshua Bengio has started a nonprofit called LawZero, which
works on technical solutions to ensure AI systems are developed and used
safely.
Different Approaches to AI Safety
Fathom is creating legal and technical frameworks to check
if AI systems are safe. It works with verification platforms like Avery, METR,
and Apollo, especially in regulated sectors like insurance, finance,
healthcare, and construction.
Current AI is focusing on inclusive AI. In partnership with
Bhashini, it has launched a hardware device using open-source AI models. The
device helps rural people and people with disabilities communicate in their
native languages. The project will later be opened to startups to build more
tools.
Funding and Impact
Most of these organisations receive funding from AI
companies, donations, or philanthropic groups. Current AI received $400 million
in funding at the AI Impact Summit in Paris. Fathom is mainly donation-funded.
Some of their work is already influencing policy. For
example, model AI legislation created by Fathom is being discussed in six US
states, including California, Virginia, and Ohio. One law called SB-53
(Transparency in Frontier Artificial Intelligence Act) has already been adopted
in California and came into effect on January 1.
However, experts say it is still early. Discussions are
happening, but creating strong AI safety systems will take time.
