The emergence of artificial general intelligence (AGI) would be the single most significant technological advance in the history of humanity, writes Professor Richard Susskind in the 30th anniversary edition of his book The Future of Law.* ‘We would be sharing the planet with entities that are as capable as us. That is an eventuality for which we are comprehensively ill-prepared.’
Despite being regarded as a futurologist, Susskind is not necessarily predicting that AGI will become available. Still less is he willing to put a timescale on it. But he says the technology companies believe they will deliver a version of AGI between 2030 and 2035. Despite working with law firms, he has found hardly any lawyers who understand its likely impact, least of all on the young trainees and associates they are now recruiting.
AGI is a step up from the generative AI that we use at present. AGI systems, in Susskind’s definition, will be able to match human performance across the full range of our cognitive tasks.
But superintelligence goes even further. Artificial superintelligence involves machines that are designed to improve themselves. This ‘recursive self-improvement’ could produce computers that are unfathomably more capable than us.
It has often been said that the law lags 10 years behind new technology. And until a few days ago, the government had shown little awareness of what may lie ahead. But two developments have brought AI to the top of the political agenda.
One was a social media post which had been viewed more than 170 million times by the middle of this week. Jacob Coxon, 27, said he had resigned from his research job at Anthropic, one of the leading AI developers, because he believed that Anthropic and its chief rival, OpenAI – where he had also worked – were ‘racing straight to self-improving superintelligence and gambling with our lives’. The people building AI ‘earnestly believe that it could kill us all by the end of the decade’, he added.
Coxon’s concerns were acknowledged by industry leaders. Dario Amodei, head of Anthropic, said developers needed to slow the pace of progress. He feared that, within six to 12 months, a swarm of AI agents could be capable of taking over the entire internet, potentially causing hundreds of billions of dollars in damage. Amodei’s call was supported by Sam Altman of OpenAI and Elon Musk, who produced the AI chatbot Grok.
On the very day that Coxon’s warning appeared, parliament’s joint committee on human rights was signing off a report calling for new legislation. MPs and peers say the law should prohibit ‘the development and provision of very powerful AI systems (such as artificial general intelligence and artificial superintelligence) which risk causing widespread and very serious harm, including the capacity to evade effective human control’.
Two years have passed since Labour’s election manifesto promised to ‘ensure the safe development and use of AI models by introducing binding regulation on the handful of companies developing the most powerful AI models’. In the King’s speech shortly afterwards, ministers said they would ‘seek to establish the appropriate legislation’. But none has appeared.
At the House of Lords on Tuesday, a Liberal Democrat peer asked the government about the risk ‘of AI causing human extinction within the next decade’. The minister did not sound too worried about the end of the world – though she did say the government was treating AI risks with ‘extreme seriousness’.
Louise Haigh, standing in for the prime minister at the TUC, went a little further. ‘There are clearly huge risks to our national security and society if the right guardrails are not put in place,’ the first secretary of state said. ‘This is an issue that goes beyond our borders and we are ready to work with our international partners to make sure we get this right.’
Haigh did not explain what we could do while our biggest international partner – under its current president, at least – was not willing to work on this with us.
The government now has a minister for AI – the high-flyer Kanishka Narayan (pictured) – and MPs hoped he would be called in to answer an urgent question before the Commons adjourned for the party conference season. As a fallback, Narayan wrote a piece for The Times calling for international efforts to protect the public.
The minister was pleased to see AI pioneers recognising the need to slow development. But these companies needed to be more transparent about emerging capabilities and risks, he wrote. They should put robust safeguards in place before increasingly powerful models were deployed – not afterwards.
Or, as Susskind is fond of putting it, better to build a guardrail at the top of the cliff than keep an ambulance at the bottom.
*The Future of Law – Reflections and Predictions will be published by Oxford University Press on 24 September
joshua@rozenberg.net
No comments yet