In the near future, everyone should have an assistant that is aligned to them, and physically with them.
Following the current status quo, we will not get there. The leaders of the race are beholden to shareholders and investors.
Not a hypothetical: see the collapse of OpenAI's governing structure.
For-profit companies beholden to shareholders and investors will never be able to be truly aligned with the individual.
Alignment is NOT a technology problem.
Alignment is a provider problem.
We propose a fully capable aligned assistant, provided by an organisation that is not beholden to profit maximisation, but rather to the individuals and communities it serves.
Peace & AI's legal structure is designed to optimize for alignment.
Our job as the provider is choosing the models and technologies that are as aligned as possible while still being fully capable.
This only works if the provider stays independent.
No exclusivity contracts
Nothing that makes swapping a provider expensive. The moment a swap costs us something, the choice stops being the individual's.
Rigorous testing of every model
Each one tested for alignment with the individual, not for benchmark scores, and the results published either way.
The aligned assistant is not a static product. It changes as the most aligned option changes.
Which is the part a company with a model to sell cannot promise. If you are building this, funding it, or would rely on it, the next step is a short call.