The European Commission states that “humans are increasingly influencing the climate and the earth’s temperature by burning fossil fuels, cutting down rain forests and farming livestock.” We call this effect on our climate, which was unintended and not recognized early during the Industrial revolution, Global Warming. I would suggest that for AI, there will be a similar awakening in the future where we might state that humans are increasingly losing control over the growing influence of AI technologies on their social, economic and political climate. We can call this Global Artificial Intelligence Swarming or Global Swarming for short, where swarming describes a situation of being overrun by AI technologies.
So what is happening with Global Warming? Well, there is now, after a century of “burning fossil fuels, cutting down rain forests and farming livestock”, a realization that the threat is real and only a collaborative global effort might help mitigate or slow down the adverse effects. This realization is what led to the Paris Agreement on Climate Change.
The Paris Agreement “brings all nations into a common cause to undertake ambitious efforts to combat climate change and adapt to its effects, with enhanced support to assist developing countries to do so. As such it charts a new course in global climate effort.”
Now this brings up another question. What if these same Parties to the Paris Agreement had reached such an accord 10 years ago instead of in 2017? Or 30 years ago? Or 50 years ago? First it is well known that safety prevention is generally cheaper than reactive safety measures. Second, we know that over time, as a safety concern arises, risk management options may be reduced.
We think that the global AI community can learn from what is going on with climate change and not wait until it is too late to agree to collaborate in an effective global framework. For this reason, the Consortium for Safer AI is unique in asking it members to take a pledge to share information, without jeopardizing their business well-being, which would help elevate the overall understanding of all the potential risks associated with the fast growing penetration of AI technology into our world. Max Tegmark has noted that as a technology gets more powerful, it is more risky to learn from mistakes. Learning from mistakes has been one of the ways of managing the risks of our products and processes. With the exponential growth in the power of AI technology, we cannot afford to wait and learn from failures of commercialized products much longer. This pledge by Consortium members is modeled on a feature of the Paris Agreement called the nationally determined contributions (NDC). The NDC asks that each Party to the agreement take actions and report those actions so that other Parties to the Agreement can benefit. This is not a binding agreement and so its effectiveness is based on the “common cause” that “charts a new course in global climate effort”.
The Consortium for Safer AI was formed to help create a Paris Agreement like environment to fund research and share learnings as we are still early in the AI revolution. We do not want to be in the same position with AI revolution as we are now trying to deal with the consequences of our Industrial revolution. As the Paris agreement notes, since we have let too much time pass, we must make “ambitious efforts” and must learn to “adapt” to the effect of climate change. For the AI revolution, we still have time to use our human intelligence collaboratively to be proactive about the safety risks of the coming AI revolution.