AI Fringe Day 5: Looking ahead: What next for AI?

As the curtains close on the AI Fringe in London, the air is dense with anticipation. After a week of fervent discourse at the edge of technological foresight, the final day unfolds, offering a collective reflection on the gathered wisdom and a gaze into the crystal ball of AI’s future.

The day commences with a panel discussion that feels less like an epilogue and more like the blueprint for a new chapter. Moderated by Evan Davis, a journalist and presenter from the BBC, the panel features an eclectic mix of voices: Matt Clifford, Dame Angela McLean, and Peter Kyle. Each of them holds a lantern to the past week's events, shedding light on the transformative dialogue that transpired and the steps to be undertaken hereafter.

Dame Angela McLean, the UK Government's Chief Scientific Adviser, is the first to share her thoughts. Her words weave through the summit’s past sessions, highlighting the progress made. The sentiments of pride are palpable, reflecting on a summit that has successfully circumnavigated the maelstrom of AI hype to focus on substantial, serious conversations. A particularly noteworthy session chaired by McLean dealt with the future scientific direction, where even Twitter adversaries shared a civilised debate.

McLean distills the panel discussion into four pivotal themes: the necessity to engineer AI with safety from the onset rather than retrofitting it, the management of risks with the current AI models, the scientific approach to AI safety research, and the quintessential need for inclusivity. The call for gender inclusivity and linguistic diversity resonates, emphasising the significance of incorporating global perspectives and minorities voices into the AI narrative.

The conversation takes a thoughtful turn when Davis probes the panel on their ability to balance conversations on existential AI risks with the more immediate concerns of AI’s nuanced discriminations. McLean acknowledges the summit’s adept navigation through these varied waters, from the misuse by malefactors to societal impacts on employment, illustrating the summit's comprehensive approach.

Next, Matt Clifford, representing the Prime Minister at the AI Safety Summit, reflects on his interim departure from venture capitalism to public service. His takeaway? The non-partisan nature of AI discourse. Clifford’s tenure might be fleeting, but his impact is indelible, noting the substantial ground of consensus reached at the summit, a significant step towards depolarising the debate on AI's future.

Clifford touches on the UK’s unique position to convene international dialogues on AI, a role underscored by the bilateral participation of global leaders. This, he argues, marks a potential inflection point in global AI governance. The summit's success lies not just in its discourse but in its demonstration that a multitude of voices can find common ground in an area as dynamically charged as AI.

Turning his thoughts to the future, Clifford emphasises the urgency of the work ahead. The UK’s new AI Safety Institute stands as a testament to this commitment, and Clifford's legacy will be a blueprint for future engagements, even as he steps away from his public service role.

Following on, Peter Kyle, Shadow Secretary of State for Science, Innovation, and Technology, offers a unique perspective. Not having been present at the summit, Kyle’s insights are formed through the lens of the public and through individual meetings with key figures. He celebrates his ‘social ostracisation’ for providing an invaluable outsider’s viewpoint and opportunities for in-depth conversations with global AI influencers.

Kyle appreciates the summit’s significance and the effort poured into it. However, he expresses concern about the messaging to the public and legislators, highlighting a dissonance between the perceived existential threat and the call for tempered regulatory actions. The mixed messages suggest a disparity between recognising a significant risk and proposing robust action to mitigate it.

Next, Clifford emphasises Britain’s ambition to spearhead AI development, not just for economic gains, but to ensure foundational democratic values are intertwined from inception. Kyle suggests that the regulatory landscape in the UK needs testing for efficacy to ensure it doesn't just promote public service and economic growth but also addresses the stability and protection of democratic processes, especially with an important year for democracy on the horizon. The consensus was clear: safeguarding democratic values is crucial, particularly as the role of AI and technology in elections and democracy becomes more pronounced.

McLean broaches the subject of misinformation, emphasising the need for public education on the potential deceptions of AI, including doctored photos and videos. The panel agreed that beyond regulating “the liars,” there is a profound need to educate the populace on recognising and defending against such tactics.

Clifford made an interesting comparison with biosecurity, suggesting that “hardening the world” is as vital as regulating AI models themselves. In terms of industrial policy, Kyle reflected on the potential friction between each country’s desire to dominate the AI sphere and the need for globally aligned safety standards. He held an optimistic view that historical precedence suggests nations can unite over shared threats to find common solutions.

Kyle shared a compelling personal story about the transformative impact of AI in healthcare. He narrated how AI integration into radiotherapy departments is not only speeding up cancer diagnostics but also enabling medical professionals to upskill, focusing on more specialised and impactful work. His narrative was a powerful testament to the potential for AI to enhance and save lives, directly addressing the concern of whether AI technology aligns with societal benefits.

The discussion also ventured into the regulatory realm. Clifford pointed out that a race to the top could be more beneficial than a race to the bottom in terms of regulation. There was a shared belief that transparent, swift, and appropriate regulation could attract businesses, with the public adoption of AI being contingent on its safety.

The panel discussion, weaving through reflections, aspirations, and constructive critique, sets a tone of cautious optimism. It underscores the enormity of what lies ahead: a complex interplay of regulation, innovation, and the ceaseless pursuit of an inclusive future shaped by artificial intelligence.

Next, the second panel of the day, “AI safety: a global question,” sought to contextualise and integrate a week of vibrant discussions on the future of artificial intelligence within the broader international stage. The panel, moderated by Kevin Allison of Minerva Technology Policy Advisors, featured an intriguing mix of speakers representing both governmental and civil society interests, with Charlotte Watts (Chief Scientific Adviser, Foreign, Commonwealth and Development Office), Daniel Remler (AI Policy Coordinator, U.S. Department of State), and Linda Bonyo (Founder, Lawyers Hub) providing their unique perspectives.

The timing of the AI Fringe Hub was nothing short of impeccable, positioned amidst a bustling 'Autumn Festival of AI diplomacy'. It became a pivotal point, an inflection in the ongoing and multifaceted dialogue about AI’s place on the world stage. Just prior to the meeting in London, there were talks at Bletchley Park, and on the horizon lay the OECD meetings in Paris and the inaugural meeting of the Global Partnership on AI in Delhi. The pace of these discussions, the different venues and voices, highlighted the growing acknowledgment of AI as a matter of pressing international concern.

Kevin Allison began the panel by emphasising the nuanced reality of global AI diplomacy, a field marked by gradual progress often perceived as frustratingly slow to the external observer. Yet, this incremental nature is intrinsic to the delicate dance of international relations, especially when it comes to the contentious and complex topic of AI governance.

Charlotte Watts shared insights on the recent summit at Bletchley Park, revealing the diplomatic triumph of bridging 28 countries, including major powers like the US, EU, and China, in conversation about AI safety. The summit concluded with an understanding that while AI harbours immense potential, it also presents significant risks, particularly for future generations and iterations. A major takeaway was the recognised need for coordinated international action on frontier AI issues. Watts highlighted the Bletchley statement's significance, a testament to intensive diplomatic groundwork and a clear milestone in AI safety discourse.

Furthermore, the establishment of AI safety institutes in the UK and the US underscored a commitment to moving from dialogue to tangible measures. The commitment to shared benefits and the inclusion of diverse stakeholders from government, private sector, and civil society marked the summit as a successful step in a longer journey, with follow-up meetings in South Korea and France already scheduled.

Daniel Remler contextualised the summit within the broader tapestry of global AI policy development. He drew attention to the longstanding efforts by bodies like the OECD, which have been forging AI principles since 2018, and the G7’s ongoing work since 2016. Remler underscored that while skepticism is healthy in the face of burgeoning initiatives, the Bletchley Park summit successfully filled a gap by focusing on long-term risks and bringing together a diverse group of stakeholders at high levels. This success, he argued, signified a promising direction for future concerted international efforts in AI policy.

Linda Bonyo then turned the discussion towards Africa, voicing concerns about the continent's underrepresentation in global AI discourse. Bonyo’s remarks were potent, pointing to the dichotomy between the global North’s advanced AI debates and Africa's struggle with foundational digital infrastructure. She underscored the importance of recognising the continent's diversity and agency, emphasising that Africa should not be lumped together with other regions under the monolithic term "global South."

The panel thus concluded by foregrounding the complexity of AI diplomacy, which must accommodate a mosaic of global perspectives while striving for cohesive international strategies. As the audience applauded, it was clear that the event had laid bare both the accomplishments and the challenges that lie ahead in the pursuit of safe, equitable, and globally considerate AI development.

This panel served as a mirror to the world, reminding the audience of the diverse and sometimes divergent interests that must be navigated as we forge a future with AI as a global partner. The path ahead is one of cautious optimism, lined with the knowledge that AI’s greatest test will be in its capacity to enhance, not eclipse, the human experience across all corners of the planet.

Following on in the final day schedule was a fireside chat, aptly titled "View from the Summit and Looking Ahead," was moderated by Melissa Heikkilä, Senior Reporter for AI at MIT Technology Review, and featured two esteemed voices in AI: Francine Bennett, Interim Director at the Ada Lovelace Institute, and Lila Ibrahim, Chief Operating Officer at Google DeepMind.

This pivotal discussion wasn't simply a recap but a strategic compass pointing towards the future trajectory of artificial intelligence. The forum's dialogue reflected on the insights from the UK AI Summit, analysing their impact on the AI landscape and the expectations of what's to come.

Bennett, fresh from the Summit, quipped about the "lot of tech guys looking uncomfortable in suits," marking a departure from the industry's usual casual norms. However, it was the unexpected presence of Chinese representation that sparked a scramble for translation devices and ignited a flurry of interest from the attendees. The inclusion of such a broad range of international stakeholders underscored the global reach and implication of AI developments.

Ibrahim reflected on the Summit and the week's events as a beacon for inclusive dialogue among diverse stakeholders. With the White House's earlier commitments and the flurry of activity post-Summit, there was a tangible sense of momentum. From setting up the Frontier Model Forum to discussing AI safety funds, the week was marked by concrete steps toward harnessing AI for the future.

Heikkilä’s personal commentary on the evolution of AI policy—from a niche subject to a hot topic—highlighted the dramatic shift in public and political attention towards AI. Both panelists acknowledged this surge in interest as a crucial moment to seize and drive meaningful action in the sector.

The panelists unanimously agreed that the time for broad declarations and general conversations has passed; the next vital step is operationalisation. Bennett emphasised the need for "hard legislation" and "hard rules" to ensure that AI technology works beneficially for society. Ibrahim highlighted DeepMind’s efforts in incorporating diverse voices into AI development and policy formation, championing responsible innovation.

Bennett called for a stronger lead from governments, suggesting that they should define what responsible AI looks like, setting a stage for companies to act within those parameters. Such government-led initiatives could create a framework for responsible AI development that aligns with societal values and norms.

Ibrahim pointed out the less "headline-grabbing" yet crucial work that lies ahead in operationalising the principles discussed at the Summit. She cited examples of DeepMind's initiatives, such as engaging with experts in various fields to understand AI's implications across different sectors.

Bennett added the necessity of broadening the concept of evaluation to include the societal impact of AI, beyond the technical aspects, suggesting a more holistic approach to understanding and regulating AI’s influence.

The discussion also touched upon the polarising debate around AI safety and ethics, with Heikkilä noting the often-conflicting viewpoints that portray AI either as an existential threat or as a tool whose immediate risks are manageable. The Summit, according to the speakers, played a role in bridging these viewpoints, creating a more nuanced understanding among policymakers about the real risks and potential of AI.

As the fireside chat concluded, it was clear that the AI Fringe was not just an event but a catalyst. The conversation called for a continued and collective effort to build on international coordination, evaluate the risks responsibly, and establish structures that could hold the burgeoning technology accountable.

In the calm after the week’s storm of ideas, two things were evident: AI’s future is not a path to be walked alone, and the roadmap for responsible AI development has many architects—from tech companies to governments, from the thinkers to the doers. The UK AI Summit was a step, but the journey ahead is a marathon that requires stamina, collaboration, and a shared vision for an AI-integrated society that values ethics as highly as innovation.

The industry, now more than ever, is poised at a crossroads where every decision and non-decision can sculpt the landscape of our digital future. What is clear from this session is that while the summit has provided an elevation from which to view the horizon, the real work lies in the climb ahead. The AI community, bolstered by events like the AI Fringe and invigorated by the influx of public interest, stands ready to take on the challenge.

These sessions were followed by a panel and fireside focussed on ‘AI & Creativity: Protecting Creators in the Age of AI’ in which I participated. I will be writing those up separately and sharing the video links.