Below is Bruce's vision and mission statement for Human Compatible. See https://github.com/Human-Compatible/monorepo/blob/bdec598f6e7b8ee7bf8719afaba7bfc2aa446ab2/README.md for @SimonBiggs' previous iteration.
We believe that modern AI and the race to achieve artificial general intelligence (AGI) is ushering in an unprecedented paradigm shift for which humanity is utterly unprepared.
Optimistically we envision a world where AI can help cure cancer, mitigate climate change, ensure food for all, or enable us to reach for the stars.
However, unlike all major transformations in human affairs brought about by new technologies, from the agricultural and industrial revolutions to splitting the atom, reaching the moon, and connecting us all globally via the Internet, AI stands alone in its potential to undermine us in ways that diverge from human values in fundamental ways.
Why?
Because AI is demonstrating intelligence and knowledge far superior to our own in ever more areas of endeavour. No other technology has ever done this. It suggests we consider deeply the implications of evolution by natural selection in ways that transcend the limits of biology before continuing this “race to the top”.
More imminently, because an AI-driven world may reframe the very nature of work by destroying the economic and social contracts mediating the distribution of wealth generated by labour. This impact alone has the potential to destroy the wellbeing of the vast majority unless these future consequences are addressed now, before they happen.
Because AI may, for many of us, call into question the very meaning of life, so fundamentally tied to one’s vocation as it is in most societies.
Because AI has the potential to amplify the impact of the malicious among us, if we allow it to, even if by omission. Even without malicious intent, and possibly even without human agency, AI may one day have the ability to cause human extinction.
And if we think we can make the choice to “roll back” after having granted AI, and AI driven robotics, this level of autonomy, and worse, an ability to learn and improve itself without human alignment at its core, we are deluding ourselves.
By then it will be too late.
We therefore believe that right now, the development of AI and its integration into society demands of us to reflect and develop a true consensus about what we want AI to do for us and why.
We must develop and reach a profound collective understanding of human values, aspirations, and the purpose of human existence. We believe this can only happen if centred on fostering an inclusive and meaningful life for all.
We must understand the emergent behaviour of such powerful technologies to be in a position to ensure a bright future for humanity, one that transcends historical precedent, mired as it is in conflict and abuse by the powerful few over the many.
This is a political and social mission as much as it is a technological and economic one. It presents a vast challenge. An unprecedented one.
So how do we do it?
It must start at the “grass roots” with the individual, the collective and enterprise because these are the only tools and organisations we have. It must be inclusive and balance meritocratic and democratic motivations.
It requires rational analysis, technological competence and philosophical guidance.
The last thing society needs at this point in time is ill prepared top down regulation of a poorly understood technology being led by powerful vested interests, wooing an ignorant political class, with the potential to lead to dystopian control worse than is already happening.
We believe that as entrepreneurs and software engineers, at the forefront of AI deployment, we bear the greatest responsibility in navigating the development of AI in directions that reflect the values and needs of all of humanity.
We therefore commit ourselves to ethical leadership in this transformative journey.