In 2020, the Future of Life Institute published a report entitled "The Windfall Clause: Distributing the Benefits of AI for the Common Good." (They also wrote an abridged version to be published in the Proceedings of AIES. Also, fun aside: Metaculus' own @Anthony got a shout-out in the acknowledgements.) Essentially, they expanded on an idea from Bostrom's Superintelligence that a firm working on transformational AI technology could limit global inequality as a result of their breakthrough by pre-committing to keep only a fraction of the fabulous profits, and distribute the rest.
Though there are numerous obstacles barring the path to making this plan a reality, the FLI research lowers many of these bars, including investigating the legal apparatus for enacting such an agreement and envisioning some of the logistical means by which a windfall might be distributed. What really remains is buy-in. If this plan will work, it will work because the companies positioned to develop windfall AI technologies signed on to a Windfall agreement in advance.
Before 2025, will any of the following companies/organizations ratify a Windfall Clause?
- Alphabet (or any of its subsidiaries, such as Deepmind or X)
For the purposes of this question, a Windfall Clause is any legally-binding agreement which caps profits resulting from a new technological innovation at any value less than 50% of Gross World Product. Ratification means encoding the Windfall Clause within a institutional structure with external enforcement mechanisms (e.g. as a publicly-promulgated legal contract, as a Smart Contract on a Blockchain. Internal commitments without third-party arbiters will not suffice.)