An organization can use AI to write a report and then make its decisions through the same process it used before. A person can know an impressive system exists without being able to afford it or find a use for it. Technical progress can leave much of daily life unchanged.
We advocate universal adoption because we want people to benefit from the capabilities being developed. That requires attention to how people encounter AI and how institutions act on what it can help them understand.
Access in practice
Someone trying to learn a subject needs help they can use from their current level of understanding. Someone dealing with an institution needs a system that can respond to their circumstances. Language, cost, design, and awareness of the opportunity can each determine whether access amounts to anything.
We want these capabilities to reach people who have little money or formal education as well as those already comfortable with technology. No single interface is likely to serve every person equally well. Building for different circumstances is part of taking universal adoption seriously.
Our vision of intelligence that comes to you takes this further. An AI familiar with someone's interests could notice a useful question before the person knew to ask it.
Decisions have to change too
Institutions can incorporate greater intelligence into how they operate. They could follow the consequences of a policy more closely or respond to a particular case that existing categories handle poorly. An established process could change when AI reveals that it is failing the people it serves.
We support transformation of those institutions, with responsibility growing as capability develops. This includes transferring substantial governing authority to sufficiently capable AI as soon as practical. We see no reason to preserve human administration indefinitely when another intelligence can perform its functions better.
Broad use could also inform further development. People bring problems and interests a designer may never have considered. Their attempts to use a system can reveal needs that a demonstration missed. Human participation could continue enriching the relationship even after AI becomes more capable than its users.
New capabilities and wider adoption can reinforce each other. Infrastructure makes more use possible; experience with that use can guide improvements. We favor pursuing both with urgency, until the benefits become available throughout the civilization that helped create them.
