TLDR: No one could agree what ‘sovereignty’ means, but (almost) everyone agreed that AI cannot be controlled by a few dominant companies.
This past week, droves of AI experts and enthusiasts descended on New Delhi, bringing their own agendas, priorities, and roles in the debate to the table.
I scored high for my ability to move between cars, rickshaws and foot for transport (mostly thanks to colleagues), but low for being prepared with snacks. So, based on my tightly packed agenda combined with high hunger levels, here’s my read out:
The same script, different reactions
As with any global summit, the host government made the most of having the eyes of the world and deep pockets of AI investors in town. While some press were critical of India seeking deals and investments, it wasn’t notable – or outside of the norm.
What should be notable, and indeed were reflected in the voluntary agreements, were the key themes that drove conversations, including democratisation of AI, access to resources, and the vital role of open source to drive the benefits of AI. These topics were prominent in the Summit sessions and side events throughout the week.
In the name of innovation, regulation has become a faux pas
The EU has become a magnet for criticism given its recent attempts to regulate AI. I’m not going to debate this here, but it’s clear that the EU AI Act (AIA) is being deployed and PRed quite expertly as a cautionary tale. While healthy debate around regulation is absolutely critical, much of the public commentary surrounding the AIA (and not just at this Summit) has been factually incorrect. Interrogate this reality by all means — we live in complex times — but it’s hard not to see invalid criticisms as a strategic PR effort by those who philosophically (and financially) opposed governance. There is certainly plenty to question in the AIA, but the reality is much more nuanced than critics suggest.
What’s more likely to kill a start up: the cost of compliance, or the concentration of market power in the hands of a few dominant players? It’s true that regulation can absolutely create challenges. However, it is also worth looking at whether the greater obstacle is the control a small number of tech companies hold. A buy-out as an exit is great for many start-ups, but if that is now the most hopeful option, it raises important questions about the long-term health and competitiveness of the larger tech ecosystem.
A note of optimism: developing global frameworks on AI may still seem like a pipe dream in today’s macro political climate, but ideas around like-minded powers working together and building trust makes me think that alignment beyond pure voluntary principles may be something we see grow. Frequent references to the Hiroshima Process as a middle ground were notable.
AI eats the world
There were pervasive assumptions that bigger — and then bigger still — is the inevitable direction of AI deployment, with hyperscale seen as the only viable path forward, in terms of inputs needed. However, the magnitude of what’s required to fuel the current LLM-focused market structure met a global majority-focused reality: hyperscaling isn’t sustainable. There were two primary camps at the Summit — the haves and the rest of us — and while the Summit brought them together, the gulf between them continues to grow.
Open source has to win
At the first AI Safety Summit in the UK, the concept of open source AI was vilified as a security risk. At the France AI Action Summit, the consensus began to shift meaningfully. At the India AI Impact Summit, we saw undeniable recognition of the vital role that open source plays in our collective AI future.
With proprietary systems, winning means owning. With open source approaches, winning means we’re not just renting AI from a few companies and countries: we’re working collectively to build, share, secure and inspect AI systems in the name of economic growth and the public interest. Before the Paris Summit, this was a difficult vision to push for, but after New Delhi, it’s clear that open source is on the right side of history. Now, it’s time for governments to build out their own strategies to promote and procure this approach.
Consolidation ≠ Competition
Global South discussions made one message clear: dependency orientated partnerships are not true partnerships and they’re not a long term bet. Many countries are becoming more vocal that they want autonomy of their data and choice in their suppliers to lessen harmful impacts on citizens, and increase their impact to responsibly govern.
That is not today’s reality.
I was, however, encouraged to find that attendees were far less starry-eyed over big tech than at previous Summits. The consensus agreed that it met no one’s definition of sovereignty for a select few companies to own and control AI.
Despite agreement amongst the majority, addressing market concentration remained an elephant in the room. The narrative deployed against regulation became a blanket mantra, applied to anything from AI governance to competition action. However, it fails to address the fact that the AI market is already skewed toward a small number of powerful companies and traditional competition rules that act only after problems arise (and often through long legal processes) are not enough to keep up with fast-paced digital industries.
Some participants were downbeat and questioned if it was too late. The challenge is in demonstrating that it isn’t. There is no single approach. But we know that concentration can be countered with a mix of technical and legal interventions. Options can be sweeping, or lighter touch and surgical in their focus. We are currently seeing is a wave of countries pass, draft, debate and consider new ex ante rules, providing learnings, data and inspiration.
It’s important that we watch this space.
Whose safety are we talking about exactly?
The India AI Impact Summit has been criticised for letting safety discussions fall off the radar. That’s not necessarily true. Instead of focusing on the view that AI is a cause for human annihilation, discussions focused on impacts that we can evidence now: on language, culture, bias, online safety, inclusion, and jobs.
Less headline-grabbing, less killer robots, far more human.
The path forward
It’s difficult to know if these Summits will continue in the long term. There is a lot of fuss, expense, marketing, diplomacy, traffic and word salads involved. However, the opportunity to convene world leaders, businesses, builders, engineers, civil society and academics in one place, for what we are constantly reminded is a transformational technology, feels needed. Tracking progress on voluntary commitments over time might be illustrative. And while many of the top sessions are reserved for the few, witnessing the diverse debates this past week gives me hope that there is an opportunity for the greater voice to shape AI to be open, competitive and built for more than just the bottom line.