True Autonomous Transactions

Trust yourself

In the last two years, machines generate more information than humans. We believe that, in the near future, the same will apply to transactions.

This can't work unless we trust our devices (our autonomous cars, our smart fridges, our smart electricity meters) to transact in our best interests.

To do that, we have to replace the existing transaction model that gives all the power to the seller, to one based on agency:

The seller's side of the transaction is represented by a seller algorithm, the buyer's side of the transaction is represented by a buyer algorithm.

Each side represents its own best interest, and we achieve a trustworthy, negotiated endpoint: a platform on which to build the future of autonomous transactions.

The team

• Moises Baly | @mses-bly | Cornell Tech CS
• Gregory Pekar | @gpekar | Cornell Tech ORIE
• Johanan Ottensooser | @oatsandsugar | Cornell Tech LLM
• Pablo Watfi | @pablowatfi | Cornell Tech ORIE

Our original pitch

Our team believes that the existing machine to machine transaction system (the interesting part of IoT for us) has an agency flaw: we won't trust automated transactions where those transactions are optimizing for the seller's profit. Existing automated transaction providers dodge this problem by making it easier for you to consent to these transactions (e.g. Amazon Dash) or by getting you to consent to a repeated string of transactions (e.g. subscription based payments): neither of which are truly autonomous.

What we propose is an agency based solution:

• your interests are represented by an algorithm that optimizes for your utility/happiness (considering price and value) which negotiates with seller algorithms that optimize for their own profit
• which interact on a marketplace that we would propose to build and manage.

This solves the agency problem by having your interests represented (you don't merely trust the other party not to exploit you), which would build trust in true automated transactions.

An interesting use-case would be managing transactions relating to autonomous vehicles:

• when/where should they refuel (considering different petrol prices, location of petrol stations, possible paying rides to that location)
• whether the car A should "bribe" the autonomous car in front of it to get out of its way if car A is in a rush (and whether Car B should agree to that bribe)

Our true target would be to allow any autonomous transactions to be negotiated within our system/using our protocol.

Our profit would be derived from:

• per transaction fees for transaction negotiated on our network
• per transaction fees for using our utility maximizing agents (though we would allow other parties to build these as well)