Advanced Artificial Intelligence

Viscount Chandos Excerpts

Monday 24th July 2023 Lords Chamber

[Viscount Chandos (Lab)]

My Lords, I found it hard to know where to start in trying to address the subject of this debate. Professor Geoffrey Hinton, whom other noble Lords have already invoked, has said that

“it's quite conceivable that humanity is just a passing phase in the evolution of intelligence”.

From the existential risk to humanity—not, in my view, to be lightly dismissed as scaremongering—through to the danger of bad actors using AI for bad things, to the opportunity to re-energise productivity growth by responsibly harnessing generative AI, a mind-boggling range of issues are raised by today’s Motion.

The introduction from the noble Lord, Lord Ravensdale, could not have been bettered as a summary and agenda—an encouragement, perhaps, that humanity has not yet been displaced in the hierarchy of intelligence. The AI entrepreneur Jakob Uszkoreit is quoted in today’s Financial Times:

“In deep learning, nothing is ever just about the equations … it’s a giant bag of black magic tricks that only very few people have truly mastered”.


This echoes the question from the noble Lord, Lord Ravensdale. Who among us, in or out of government, really understands this?

Not only is it a black box but, from what we lay people know, it is one that is changing breathtakingly fast, as my noble friend Lady Primarolo has said. Moore’s law, which observed that the number of transistors on a semiconductor doubles every two years, is nothing compared with the pace of AI development. As a foundation model or platform, GPT saw the number of parameters between GPT 1 and GPT 2 increase by 1.3 billion in less than two years. The increase in parameters in a single year between GPT 2 and GPT 3 was over 100 times greater, at 173 billion—as my noble friend Lord Watson said, it is exponential, not linear. As a simplified representation of the speed of AI’s development, this simultaneously indicates the opportunities to harness it for good on the one hand and demonstrates the formidable challenge for regulation on the other.

I am temperamentally an optimist and therefore excited by the positive contribution that advanced AI can make to our world. However, in the time available, I want to focus on two of the challenges: the macro issue of regulation and the more specific question of fake news. In that context, I highlight two of my interests declared in the register. I am a trustee of the Esmée Fairbairn Foundation, which has an endowment with investments in US and other VC funds with holdings in advanced AI companies. I am chair of the Thomson Foundation, which trains journalists, principally in low-income and/or low press freedom countries.

Regulation of advanced AI is inevitably complex. It may be right to focus in the short term on ensuring existing regulators adequately consider the impact of AI. However, in the longer term, there has to be specific overarching regulation, as I think is implied by the Government’s declared aspiration for the UK to be the geographical centre of global AI regulation. Could the Minister say what the Government’s objectives are for the global AI summit being convened for later this year? What arguments have they used for the UK to be the global centre of regulation? When, six years ago, the founder of a Silicon Valley software company described GDPR to me as the de facto global standard for data protection, that reflected the sheer size of the EU market to which it applied. That is not a factor which applies now to the UK’s negotiating position.

Finally, the media, both mainstream and social, are a vital source of information for us all; the integrity of that information lies at the heart of democracy. Your Lordships may have seen the amusing faked photograph of Pope Francis in a fashionable puffer jacket. Other future faked images could be deeply dangerous to stability and security, both globally and, in particular, in low-income countries. Regulation cannot be the only answer to this; education and training are also essential. Can the Minister urge his colleagues in the FCDO to protect and increase funding for media development and journalist training in low-income countries, with a central focus on countering AI-generated fake news, from whatever source?