Now, many investors, organizations, and entrepreneurs are deeply committed to building an artificial intelligence (AI) ecosystem that prioritizes people and planet agencies, equity and sustainability. However, current investment options remain limited. Governments and funders seeking to support public interest AI are faced with the choice of frustration to invest in costly “sovereign AI” infrastructures (high-end computing, basic models, energy) without strategic autonomy or matching frontier capabilities and clear paths to ensure US or Chinese hyperscaler applications. Neither option helps generate the contextually rich data needed for communities to tackle shared challenges. Neither can adequately address the concentration of fundamental forces that persists in a small number of private companies.
Alternative approaches to building AI systems based on collective information science (CI) can address these shortcomings at once.
As we recently explored together in discussions with entrepreneurs and investors at Human+Tech Week, now extracts technically feasible (there are advances in modeling techniques that provide privacy) and inexpensive (due to the constant cost of computing and software) and large-scale language models (from building LLMS) – individuals, teams, and community intelligence. The ultimate vision of this approach is to complement our efforts to achieve monolithic artificial general information to grow thousands or millions of intelligent communities around data and culture, meaningful equity in AI infrastructure and applications, and institutions to use these systems to self-statisticize sharing priorities.
Elements of this vision already exist in technical prototypes, policy proposals, and a dedicated community of technology entrepreneurs, scientists and investors. What is needed is a coordinated effort across these actors to sew a more consistent field that can innovate impactful use cases that block noise. To this end, we have distilled three guiding principles for framing, designing and tuning of AI ecosystems that are useful for people and planets.
1. Frame AI as a social technology
It's important to know how you talk about AI. AI that works for people must recognize human contributions to AI. LLM is pre-trained at a huge high school of human-created, human-infused internet content. It was refined through reinforcement learning using human feedback and further adjusted through human use patterns. AI consistently delivers the best performance as a hybrid system that combines machine speed and scale with collective human expertise and know-how.
Still, these human contributions to AI are not found in mainstream conversations. A new narrative is needed from a paradigm protecting AI policy debate from AI, to a paradigm that ensures that human contributions to human AI systems (deliberation, deliberation, intuition, social context) are protected and quite valuable. As some of us (TK, AP, MR) have recently proposed, there is an opportunity to reconstruct the generator AI as generative collective intelligence or “genci.” The ability of algorithms combined with human expertise can be used to address complex, real-world challenges that humans and machines cannot deal with on their own.
2. Design AI that is faithful to human agents
Humans are agents and human agents should be a central concern in AI development, but even so, in today's enthusiasm for autonomous AI systems or “agent AI,” these fundamental truths require explicit defense. Investors, policymakers, and end users need to advocate for AI algorithms, architectures, and approaches that amplify human institutions and social intelligence.
It is possible to build algorithms that capture and enhance the shared beliefs, objectives, and activities of groups and organizations developed by platforms like Common Good AI. Similarly, approaches like pol.is and deliberation or deliberation are used to scale inclusive and grounded dialogues using summary models and adaptive polling, while maintaining a voice of nuance and diversity. Human-team approaches like vibe teams can position AI tools to support the creativity and quality of human-to-human problem solving.
Emerging AI agents, on the other hand, are “by design” and can be treated as a trustee of a human, team or community, not just a company, but not just a company). Machine learning methods that provide innovative data governance (models such as the Human Genome Project) and privacy can help consolidate the local into larger community-governed ensembles, and companies such as trust and cooperatives.
3. Coordinate the big bet application centered on
Innovative applications can demonstrate why expanding alternative AI ecosystems is important. AI systems based on CI science and design principles have a natural competitive advantage in addressing challenges that a single actor cannot solve on its own. Like a regional or global green energy trading platform that uses local to transparently validate and exchange carbon intensity data from scratch. A trusted, AI-driven public service that uses local to consolidate sensitive individual and government datasets. Or scale up pioneering prototypes like interspecific money. This uses the principles of CI design to build AI that represents and evaluates the agency of non-human life. To help mobilize the necessary scale of the infrastructure (high-end calculation, data, and talent) needed to develop world-leading use cases, the “Airbus for AI” model (the public consortium of the National AI Labs of the Middle-class powers) will work together to collaborate with these ideas to help the market and help develop utilities and Europe's first proposal).
