Contribute to the research
The question: How do AI systems fail when the people responsible for human life use them, and how do we use them without losing control?
Why it matters: AI is moving into the systems where care decisions are carried out: the medical record, the chart, the insurance company. A physician doesn't own those systems and can't look inside them. If a machine overrides a physician's judgment there, the physician won't see the decision, only the outcome. You don't want the only people who understand these systems to be the most unscrupulous ones.
What exists now: 63 documented failure cases in 21 types, and a curriculum being released one piece at a time on Parallel Genius as each is reviewed.
Funding pays for:
- Reviewing the cases and building the curriculum, so it gets out faster and I can teach people
- With more funding, formal testing
This work goes ahead either way. How much we can take on, and how fast, depends on funding. We're a small team.
AI research contributions aren't refundable, because the research is already under way.