“AI as CW: Are Quakers Called to Transform Artificial Intelligence into Collaborative Wisdom?”
I am excited by the roster of sessions coming up for the Summer Research Seminar of QIF this July 17th-21st. (The program will be posted in the next few days.) In the meantime, I would like to share a leading I have been following for a long time and whose weight has been pressing ever more heavily on me since the release of ChatGPT last November and the subsequent flood of events, news, and debates around Artificial Intelligence (AI).
The misuse of AI and perhaps even its emergence as an increasingly autonomous power seems increasingly to threaten our world. How can we respond to this not out of fear but hope? How can we enter into that power of love which is divinely available at all times to teach, guide, lead, heal, and transform each individually — and all of us collectively?
It is becoming clear that the exponential growth in the power of AI promises not only extraordinary wonders but serious dangers as well. These include complex “wicked” problems and perhaps even existential threats as just as significant as climate change, the sixth great extinction, and nuclear war.
For people of faith, the central challenge is to respond to threats of this sort by not simply asking “How can we save our world from destruction by these forces?” We need to ask: “How we may be called to redeem the world through creation guided by love?”
We are called to go deeper in our understanding of the forms of reasoning and decision making that have brought us to the brink and seek out the practical methods of guiding thought and action by that Inward Light that transforms ego driven calculation into collaborative wisdom.
Quakers, Gandhians, Indigenous communities, and others have centuries of experience in experimenting with methods of decision making that are, at their best, spirit-led forms of communal reasoning that provide ways to discern, demonstrate, and non-violently defend emergent, objective, moral truths. There is a desperate need to seek out ways to encode these methods into the hearts of our civilization’s institutions and technological systems.
The machines, institutions, users, and natural organisms which are all being assimilated into the emerging AI systems are members of a great field of activity. We are called to enter in, not as missionaries who would colonize the heathen, but as members of a community in which we are genuinely seeking, as humbly and faithfully as we can, to see what love might do.
We need to not only work to ensure that AI systems are ethical, legal, and safe. We also need to promote research and development of ways to transform the rationality operating in the machines that play key roles in driving our life systems and social institutions. They should not be run by the narrow economic rationality of most current Artificial Intelligence but, instead, by ecologically balanced Collaborative Wisdom. We need AI as CW.
Two decades ago, the Quaker Institute for the Future was founded as a think tank that might serve to help discern and develop a testimony on ecology that might be as fundamental and as powerful as the Friends’ Testimony on Peace. A host of individuals and organizations have worked with much success to make a vision of “an earth restored” become central to Friends’ public witness, lobbying, and actions – including down to earth activities of building solar panels and creating organic farms.
We are at a moment now in which developments in AI call for a similar effort, one which will likewise require Friends to respond with prayerful discernment, community organizing, and practical engagement in the transformation of technology systems of AI.
I am hoping that Friends will enter into dialogues to engage seriously with this concern in the weeks and months to come. I have appended below some further thoughts that might spark discussion and prompt further inquiry and experiment. I would also note that QIF will soon publish a book I authored which addresses these issues in depth: Smarter Planet or Wiser Earth? Dialogue and Collaboration in the Era of Artificial Intelligence. I would also note that several people doing research/action projects on these issues are interested in organizing QIF workshop in the coming months to consider these issues from a Quaker point of view. Please be in touch if you would be interested in participating.
An Appendix: Quaker Testimonies and the Emerging Forms of Artificial Intelligence
Since the release of ChatGPT in November of 2022 and the flood of new Generative Artificial Intelligence programs, there are increasingly clear and pressing questions about these new systems. The questions concern their ethics, safety, and legality as well as their compatibility with a life framed by faith and led by spirit.
The misuse of these systems to manipulate people’s economic, political, and other life choices threatens to undermine the most fundamental power of citizenship in a free democracy, that is, the power for individuals and communities to effectively deliberate decisions in rational ways that pursue their individual and collective interests. In that sense, these systems provide a fundamental threat to the sovereignty of states and the freedom of citizens.
For Quaker and other religious communities, these systems further threaten what we understand to be the most definitive feature of humans, the ability to be open to the will of God and embody Jesus’ teachings through guiding our lives by the power of love and the Inward Light. It is this ability which both empowers us and calls compellingly to us to act in faith as Friends in order to, as FCNL has put it: “create a world free from war, a society with equity and justice for all, a community where every person’s potential may be fulfilled, and an earth restored.”
Like most citizens, many Friends may find the current wave of Generative AI systems difficult to label and describe, let alone understand and evaluate. But key features and concerns are becoming painfully clear. In key respects we find ourselves in a situation comparable to 1945 and the advent of nuclear power. We are faced with a technology whose inner workings and unprecedented power for both good and bad may mystify us. We can, however, recognize key ethical and existential threats it poses. In the case of the Generative AI systems like ChatGPT, we can see:
a.) These systems being developed and deployed throughout our society rely increasingly on black box, opaque systems that their creators cannot well understand, predict, or control.
Previous systems were transparent because they were designed and assembled by programmers using principles of formal logic. In contrast, these systems are provided with a malleable and evolving set of random initial behaviors and then their creators give them data equivalent to hundreds of thousands of years’ worth of lives found on the internet. They then develop and train the AI models to imitate that internet behavior. They do so with methods that are analogous to those used to breed and train dogs in at least one, crucial respect: Despite whatever knowledge they may have of the basic blocks that serve like the genes and neurons of the system, they have no understanding of the emergent design that structures the billions of parameters operating in it.
b.) Such systems are becoming more powerful at exponential rates that outpace most organizations’ abilities to assess their impacts and deliberate effectively concerning the ethics, safety, and legality of their behavior and effects.
For instance, the 3.5 Generative Pre-Trained model of AI that was used to create ChatGPT surprised its creators by proving capable not only of learning natural languages like English but of learning to write code in Python and other computer languages from internet data on sites like GitHub. The extraordinary powers of the new system were a key reason it became the fastest growing app in history, being adopted by over a million users in five days and 100 million in less than two months
c.) There are many fundamental ethical, safety, and legal threats these systems pose for which the AI research community has not been able to provide solutions or even, in key cases, demonstrations that solutions are possible.
The enormous reams of internet data used to train the AI behavior are poorly filtered and includes a host of unethical, unsafe, and illegal role models. The probabilistic reasoning methods used to reinforce ever better imitations of those models are designed to innovate with new text that, by its nature, regularly “hallucinates” with innovations that are absolute fictions in fact or fallacies in formal reasoning. Further, those probabilistic methods are, in themselves, only designed, at best, to imitate what people actually say and do on the internet, not what they ought to say or do. Human feedback and control are essential to review and revise them in order to move them beyond probabilistic inference aimed at imitation to genuinely dialogical forms of reasoning guided by emergent, objective, normative values. There are currently no viable designs for providing human review and maintaining human control over the five hundred thousand years’ worth of reading time they use to access data, the billions of parameters they use to process them, and the literally lightning-fast speeds at which their electrical circuits operate. No plausible proofs have been discovered, as of yet, for thinking that there will be, in the future, any way of inventing effective systems for comprehensive human review and control of these black box systems.
d.) There are significant threats to our economy, ecology, and democracy which similar but less powerful forms of artificial intelligence have already demonstrated are serious.
Less powerful forms of machine learning have been used, for instance: in mortgage lending software that discriminated racially: in facial recognition software that consistently misidentified women and people of color; and in social media software that dramatically diminished teenagers’ self-esteem to suicidal levels, polarized communities in ways that made civil dialogue extremely difficult, and served political operatives in effectively subverting democratic elections.
e.) The malfunction or misuse of the more powerful versions of these systems pose threats of exponentially greater impact and of even of existential danger to the human species.
The extraordinary exponential growth in the use of these systems is feeding intensive investment in further research and development which is leading to new apps and versions at a pace that has led leading researchers and entrepreneurs in the industry to call for a pause. Even while they feel pushed by the flood of competition to keep advancing in their own work, most acknowledge the industry as a whole needs to be slowed down, virtually all acknowledge that the new systems are continually surprising them with unpredicted and often disturbing emergent properties and capacities, and a majority of them judge that there is a significant risk that we are entering a period in which there is a 10% or greater chance that AI may in fact threaten the existence of our species
In discerning a Quaker response to these concerns, it may help to recall the most central of all testimonies of early Quakers was affirmed in the label they chose for themselves: The Religious Society for Friends of the Truth. Artificial systems that subvert, coopt, or destroy our collective abilities to reason in shared dialogue and discern truth are fundamentally inimical to our faith and practice.
It is imperative that Friends learn more about these systems and consider how we can best act as individuals, as Monthly Meetings, and as a Religious Society, to slow the development of these technologies and direct them in ways that are guided by loving reason and spirit-led discernment. At the dawn of the age of industrial nation-states, Friends recognized the dangers of unfettered capitalism and unbridled militaries and they sought out experiments that might let us “see what love can do.” We have entered now into a period of dramatic social and technological change that calls perhaps even more pressingly for spirit-led discernment and action rooted in a commitment to turning ever more fully towards Truth and that ocean of Light which is the birthright resource of every human.