Strategic Implications of Openness in AI Development

This paper attempts a preliminary analysis of the global desirability of different forms of openness in AI development (including openness about source code, science, data, safety techniques, capabilities, and goals). Short-term impacts of increased openness appear mostly socially beneficial in expectation. The strategic implications of medium and long-term impacts are complex. The evaluation of long-term impacts, in particular, may depend on whether the objective is to benefit the present generation or to promote a time-neutral aggregate of well-being of future generations. Some forms of openness are plausibly positive on both counts (openness about safety measures, openness about goals). Others (openness about source code, science, and possibly capability) could lead to a tightening of the competitive situation around the time of the introduction of advanced AI, increasing the probability that winning the AI race is incompatible with using any safety method that incurs a delay or limits performance. We identify several key factors that must be taken into account by any well-founded opinion on the matter. Policy Implications • The global desirability of openness in AI development – sharing e.g. source code, algorithms, or scientific insights – depends – on complex tradeoffs. • A central concern is that openness could exacerbate a racing dynamic: competitors trying to be the first to develop advanced (superintelligent) AI may accept higher levels of existential risk in order to accelerate progress. • Openness may reduce the probability of AI benefits being monopolized by a small group, but other potential political consequences are more problematic. • Partial openness that enables outsiders to contribute to an AI project’s safety work and to supervise organizational plans and goals appears desirable. The goal of this paper is to conduct a preliminary analysis of the long-term strategic implications of openness in AI development. What effects would increased openness in AI development have, on the margin, on the long-term impacts of AI? Is the expected value for society of these effects positive or negative? Since it is typically impossible to provide definitive answers to this type of question, our ambition here is more modest: to introduce some relevant considerations and develop some thoughts on their weight and plausibility. Given recent interest in the topic of openness in AI and the absence (to our knowledge) of any academic work directly addressing this issue, even this modest ambition would offer scope for a worthwhile contribution. Openness in AI development can refer to various things. For example, we could use this phrase to refer to open source code, open science, open data, or to openness about safety techniques, capabilities, and organizational goals, or to a non-proprietary development regime generally. We will have something to say about each of those different aspects of openness – they do not all have the same strategic implications. But unless we specify otherwise, we will use the shorthand ‘openness’ to refer to the practice of releasing into the public domain (continuously and as promptly as is practicable) all relevant source code and platforms and publishing freely about algorithms and scientific insights and ideas gained in the course of the research. Currently, most leading AI developers operate with a high but not maximal degree of openness. AI researchers at Google, Facebook, Microsoft and Baidu regularly present their latest work at technical conferences and post it on preprint servers. So do researchers in academia. Sometimes, but not always, these publications are accompanied by a release of source code, which makes it easier for outside researchers to replicate the work and build on it. Each of the aforementioned companies have developed and released under open source licences source code for platforms that help researchers (and students and other interested folk) implement machine learning architectures. The movement of staff and interns is another important vector for the spread of ideas. The recently announced OpenAI initiative even has openness explicitly built into its brand identity. Global Policy (2017) doi: 10.1111/1758-5899.12403 © 2017 The Authors Global Policy published by Durham University and John Wiley & Sons, Ltd. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Global Policy