I was thinking about steps towards personal digital assistants. In a separate thread, I was also thinking about the psychology of aging. In a third thread, I was thinking about projects I might want to build to help me learn more. It makes sense to bring all these threads together: thinking of systems I can build to improve the quality of life I’ll enjoy in the future.

I think this might be a better fit for my experimental learning than either a hypothetical market or specific people. After all, I’ll always have a future self who could benefit. (And if I don’t, I’ll be past caring!) If the things I build along the way turn out to be useful for others, all the better.

Anyway, I was thinking about the kind of simple, deterministic, idiosyncratic assistant I could build to make life a teensy bit better in the medium term and the long term.

I could start with a text box interface on a webpage, then move to alternative inputs like dictation or neuro-integration(!) when that becomes reliable. It would be great to have some kind of offline buffering, too.

In terms of logic, I could start with stateless well-defined responses, add synonyms, support conversational interfaces, use weighted factors, add feedback mechanisms, and then eventually reach proactive notification and action. Inferences would be awesome, but I don’t have to wait for them to be sorted out. Ditto for program generation and adaptation.

In terms of sensing and acting, I can start with existing APIs and tools, write specific adapters for other sites, push into the physical world with sensors and actuators, use context and probability to simplify, and then take advantage of improvements in fields like computer vision or biometric analysis as other people build and commoditize cool tech.

But first, it starts with building a simple tool. Hmm, maybe a little thing that suggests what to do next (and coincidentally makes it easy to track)…