AI Alignment Network (ALIGN) - Intellectual Property Incentives to Help Address AI Alignment Problems
7 Pages Posted: 26 Jul 2023 Last revised: 27 Jul 2023
Date Written: July 24, 2023
Recent public releases of large language models have captured the public’s attention and re-surfaced the so-called AI alignment problem. A large number of well recognized computer scientists such as Geoffrey Hinton and Stuart Russell have noted that AI is rapidly improving, may be approaching general superhuman intelligence and, if misaligned, could endanger human civilization. Indeed in 2023, world-renowned AI researchers, scholars, AI tech CEOs and others signed the following statement: "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war". Despite these warnings, there is a dramatic mismatch between the resources expended on the development of AI and work on the AI alignment problem.
This article proposes intellectual property incentives to help address AI alignment problems and potentially AI safety more broadly. These intellectual property incentives are achieved through private ordering solutions. Private ordering solutions have a number of advantages over legislation or treaties (including speed of implementation). Such IP-based private ordering solutions have succeeded recently in other contexts. The LOT Network, Open Invention Network and the Open COVID pledge are examples of collaboration among IP holders to solve a ubiquitous problem and they provided inspiration for the current proposals.
This article includes two proposals which could be implemented together or separately/individually. The first proposal is a collaborative, networked, royalty-free license grant (aka a pledge) similar to the Open COVID pledge but directed to the AI alignment problem as opposed to the COVID pandemic. The second proposal is a collaborative, networked licensing agreement that provides preferred IP license terms (e.g., royalty-free) to those AI systems and ecosystems that incorporate a set of best practices for AI alignment, e.g., as determined by a 3rd party non-profit comprised of subject matter experts.
Keywords: patent, license, AI alignment
Suggested Citation: Suggested Citation