Speaking ahead of a global gathering he has convened next week
to examine the risks of the technology, Sunak said he hoped the
participants could agree on the nature of the risks and
establish a global panel to assess them.
Britain will set up an artificial intelligence (AI) safety
institute to "examine, evaluate and test new types of AI so that
we understand what each new model is capable of, exploring all
the risks from social harms like bias and misinformation through
to the most extreme risks," Sunak said.
Representatives of AI companies, political leaders and experts
will meet at Bletchley Park - where Britain's World War Two
code-breakers worked - in southern England on Nov. 1-2.
"Get this wrong, and AI could make it easier to build chemical
or biological weapons," Sunak said. "And in the most unlikely
but extreme cases, there is even the risk that humanity could
lose control of AI completely."
Sunak wants Britain to be a global leader in AI safety, carving
out a role after Brexit between the competing economic blocs of
the United States, China and the European Union.
Around 100 participants at next week's meeting will discuss
subjects including the unpredictable advances of AI and the
potential for humans to lose control of it, according to the
agenda.
Sunak said China had been invited but he could not guarantee
that a representative of the country would attend.
U.S. Vice President Kamala Harris and Google DeepMind CEO Demis
Hassabis are on the guest list.
Leaders of the Group of Seven (G7) economies, comprising Canada,
France, Germany, Italy, Japan, Britain, the United States and
the European Union, in May called for adoption of standards to
create trustworthy AI and to set up a ministerial forum dubbed
the Hiroshima AI process.
(Reporting by Paul Sandle; Editing by Mike Harrison, William
Schomberg and Christina Fincher)
[© 2023 Thomson Reuters. All rights
reserved.] Copyright 2022 Reuters. All rights reserved. This material may not be published,
broadcast, rewritten or redistributed.
Thompson Reuters is solely responsible for this content.
|
|