The AI@Home Consortium (…is the future of work)
Why did Artificial intelligence suddenly take off in the last decade?
- It wasn’t because we found a powerful master algorithm for building artificial intelligence. Just the opposite. AIs became possible because the brute force processing power needed to train artificial neural networks suddenly became affordable.
- The cost of processing power is still a gating factor for the improvement of AI’s today. AI training has an endless appetite for processing power. The more you have, the better the AI you can build (and the more variation, experimentation, and tinkering that can be accomplished).
- The affordability of processing power ALREADY poses a problem for academic researchers. They simply can’t afford it — a recent example required 800 GPUs running flat out for over a month. The growing cost of training AIs is forcing experimentation out of the Universities and into the big tech companies.
The AI@home Consortium

This situation is an opportunity to introduce something new: A decentralized training system for academic AI researchers that crowdsources the processing required to do it. A system that could make hundreds of millions of GPUs available for AI training. Let’s dive into this a bit:
- This is very similar to the approach used by folding@home (protein folding for medical cures), SETI@home (the search for extraterrestrial life)and many others. In the past, these crowdsourcing efforts have been accomplished merely on a charitable basis. However, that shouldn’t apply in this case. The commercial opportunity of AIs is simply too great.
- The best solution is one that doesn’t only support academic research, it is one that pays substantial royalties on the commercial use of the AIs produced or derived by this training. How would it work? Universities, commercial start-ups and independent researchers would sign over potential commercial patents on the research conducted on the consortium’s platform, in exchange for a generous share of the benefits and access to an unparalleled training system.
- Win-win-win-win... As people contributed time on their smartphone processors, the consortium would automatically allocate their contributions to a blockchain like ethereum and regularly award royalties on the patents created with their contribution. A growing annuity paid out over long time scales, would serve to attract millions of participants with minimal attrition over time unlike any other method (contrast this to Amazon’s Mechanical Turk).
What does this mean?
A platform of this type would:
- Generate competitors. A blockchain competitor that worked based on bitcoin mining rules could emerge. However, the benefits of such a system would be dominated by <100 large mining pools. This would radically limit the potential for the emergent benefits derived from a large number of individual participants.
- An AI@home consortium could branch off in new directions or exhibit emergent properties. For example, the participants could be incentivized to create massive training databases (recent findings show that AIs have an unlimited appetite for curated data as well) for which they are paid an ongoing royalty. Further, a huge number of participants could vote with their wallet by favoring the AIs produced by the consortium over those produced elsewhere.
- A consortium like this would send a shockwave throughout the commercial world. It wouldn’t only spur the creation of dozens of competitors, it’s size would vastly exceed by revenue and “employee” count any established corporate entity. Alternatively, it could break apart into dozens of smaller parts, with each part focusing on different opportunities (from finance to farming).
I hope this gets you thinking….
Sincerely,
John Robb
PS: The future of work in an AI fueled global economy is an existential problem. Get it right and everyone prospers. Get it wrong and the paroxysms of extreme political turbulence that follow, may kill us all (we still live under a nuclear sword of Damocles). This means there is really no choice. We must figure this out. Get to work!
