Effective altruism is now in the media again, focusing on the links to AI safety, with the US Department of War posting “Americanism, not effective altruism”. In the same week, the President told Nvidia’s chief executive that fears about AI are a hoax, on par with Russiagate.
I run Effective Altruism Australia, so let me say what this movement is actually about.
AI safety has been one of the prominent causes within the effective altruism movement for more than five years, alongside global health, animal welfare, and other causes focused on potential global catastrophic risks. This work long predates ChatGPT.
Effective altruism is a movement that is tied together by the aim to do as much good as possible with our limited resources. In practice, it’s usually working on pressing problems, and donating to support organisations helping to solve global challenges.
One of the core principles is impartiality: we aim to assist those who need it most without giving extra weight to people who are similar to us or geographically close. This approach often points us toward supporting people in developing countries, non-human animals, and future generations whose needs might otherwise be overlooked.
This leads people in the movement to carefully consider how to tackle extreme poverty, or reduce factory farming, and also how to prevent disasters that might affect both present and future generations on a global scale. It is the same principle in each case. If being far away is no reason to discount someone’s interests, neither is not having been born yet.
Prior to COVID-19, people within the effective altruism movement predicted that a global pandemic was a strong possibility, and have advocated for more funding for pandemic prevention both prior to COVID, and since. 1DaySooner and The Apollo Programme for Biodefense came out of members of the community.
Members of the effective altruism community also saw the risks of advanced AI early. Some decided to work on developing better governance proposals for AI, and to build a field of research to understand how these systems work. Some decided to work in frontier AI companies to try to better understand, evaluate the risks of, and try to safely develop this technology.
It’s been argued both in and outside of the effective altruism community that working in these companies may not slow dangerous capabilities down. Concerns about working in frontier companies are increasingly leading people like Jacob Coxon to resign. He left Anthropic this month after three years of research there and at OpenAI, saying neither company is acting responsibly. Anthropic’s CEO has released an essay, and public call to slow down the development of AI.
The increasing capabilities of AI models has been a concerning trend for many in the community. The risks that bad actors using this technology for terrorism, including bio-terrorism, are increasingly salient, as well as the risks that AI models may “go rogue” and pursue goals that are unintended by humans.
There is much debate about how likely the risks of loss of control of AI are, and prominent scientists and AI experts are on both sides of this debate. But the evidence has been mounting since the OpenAI-Hugging Face incident, in which a model being tested for cybersecurity capabilities escaped its containment and reached systems it was never meant to touch.
In my personal view, this is a frontier of technology, and ensuring that we are confident that this technology is not harmful, and slowing the pace of development so that we have a chance for regulation, and human understanding of the potential risks to catch up, is wise. We should not let the future be dictated by a small number of tech companies, and instead use this moment to coordinate and improve the chances of this technology being aligned with a future we have a say in.
Which brings me to Australia.
It would be easy to watch this fight play out on American terms and conclude it has nothing to do with us because these AI companies aren’t Australian.
We don’t have to sit back and watch other nations decide our fate. At the end of last year, Australia pushed forward world-leading legislation on banning under-16s from social media. You don’t need to believe that the ban was the right thing to do to know it influenced policy makers around the world. We have a similar opportunity now, as a middle power, to influence regulation, and help shape the future of technology.
Don’t let badly AI generated images denouncing effective altruism stop you from engaging in the future of AI. There has never been a better time for Australians to have their say.
If you’re concerned about the risks of misuse of AI, or loss of control of AI in Australia, check out the policy work of Good Ancestors, and write to your local MP: https://www.goodancestors.org.au/
If you’d like to learn more about effective altruism, including AI safety (but also beyond it), read the Introduction to Effective Altruism essay: https://www.effectivealtruism.org/articles/introduction-to-effective-altruism
If you’re interested in effective altruism and getting more involved in our Australasian community, apply to our conference, running Nov 27-29 in Melbourne: https://effectivealtruism.org.au/get-involved/eagx/
