Digital Ghosts

What are you trying to do?We are unblocking the ability to talk to someone who’s time-poor, deceased, or the composite of a group.How is it done today? What are the limitations of the current system?You can read someone’s intentional writing & notes and generate group composites using consensus mechanisms.What is new in your approach and why do you think it will be successful?This will succeed because disk space is cheap, ML is getting cheaper, and we have language models that can generate a model of an individual or a composite of people.If successful, what difference will it make?You’d be able to have virtualizable people; enables high parallelization and delegation to a highly trusted individual or unblock groups who are operating in an “ask forgiveness, not permission” mode.How much will it cost?If we want to fine-tune a model of ~100 people, that’d cost on the order of $200; $40k for an org of 10k people + 2 engineers for 4-6 months (@$150k/yr), ~ $500k for an initial prototype + testing with 4 organizations.How long will it take?It would be 4-6 months to a first prototype.What are the mid-term and final exams to check for completeness?Prototype in 4 months, test with companies in 6 months to validate success.What we want/needFunders would be great; Novel monetization strategies for open source products, and LLMs with open weights.Contact: [email protected]