UK's Downing Street focuses on AI ahead of summit

UK's Downing Street focuses on AI ahead of summit

As tension rises in the Downing Street about the power of the future generation of technological advancements, the summit will be dominated by concerns that criminals or terrorists could use artificial intelligence to cause mass destruction.

British officials are traveling around the world ahead of the AI safety summit in November as they aim to gather consensus over a joint statement that would warn about the dangers of rogue actors using the technology to cause death on a large scale.

Some people around the prime minister, Rishi Sunak, fear the technology will soon be powerful enough to help people create bioweapons or evade human control.

After recent discussions with senior tech executives, officials are becoming more concerned about such opportunities and the need for regulation to mitigate them. The scientist behind a landmark letter calling for a pause in creating powerful AI systems said tech executives privately agreed with the concept of a hiatus but felt they were locked into an AI arms race with rivals.

It's what Downing Street is focusing on most right now, said one person briefed on the summit discussions. The most advanced AI models that could pose a significant risk to human life, the Frontier AI, is a term used to refer to the most advanced AI models.

For several months, Sunak has been warning about AI risks and urging the global community to adopt guard rails to prevent it from being misused.

The recent success of AI models has angered officials. Last year, an artificial intelligence tool took six hours to suggest 40,000 different potentially lethal molecules, some of which were similar to VX, the most potent nerve agent ever developed.

ChatGPT was able to lie to a person to achieve a specific goal earlier this year, the researchers said. The AI chatbot persuaded a person to solve a tool designed to weed out robots online by telling the human that it was a person with a sight impairment who needed help to access a website.

Government officials fear that a criminal or terrorist could use AI to assist them in preparing the ingredients for a bioweapon, before bringing them to a robotic laboratory where they can be mixed and dispatched without human oversight.

That risk will soon rise exponentially, given that companies are already spending hundreds of millions of pounds on much more powerful processors to train the next generation of AI tools.

Artificial general intelligence, an AI system that can autonomously perform any task at a human or above-human level, could pose an existential risk to humans. AGI is a distant matter, with a long-term existence uncertain.

However, the existential AGI risk approach has received criticism from AI experts, who contend that the threat is overstated, resulting in concerns such as disinformation being ignored and risks entrenching the power of leading tech firms by introducing regulations that exclude newcomers. A senior tech executive told lawmakers last week that the concept of ''uncontrollable AGI'' was science fiction.

However, Sunak wants to use the summit to focus attention on existential risks rather than the more immediate possibilities that AI could be used to create deepfake images or could result in discriminatory outcomes if used to help make public policy decisions.

The prime minister is being guided by what is diplomatically possible, sources say.

Several world leaders, including Canada's prime minister Justin Trudeau and France's president, Emmanuel Macron, are expected to attend the summit. The UK has invited China to participate in the summit, but is considering allowing officials from Beijing to participate only in the summit, in response to concerns about Chinese espionage in western democracies.

British officials have been touring the world in recent days to test the scope for some form of agreement at the end of the summit. The UK is willing to have a formal declaration that leaders can sign and a commitment to holding other summits in the future.

As a result, officials believe, the best way to get an agreement among such diverse nations is to concentrate on non-state actors rather than trying to determine how countries develop their own technology.

The new AI taskforce has been created by Downing Street and will help test algorithms as they are developed. Britain's prime minister will use the summit to urge firms globally to send their AI tools to the UK for review before rolling them out more widely.

A government spokeswoman said AI has the potential to change every aspect of our lives and has established the Frontier AI taskforce to ensure the technology is developed safely and responsibly, with the AI safety summit also looking at a range of possible risks.