Effective altruism and AI

In short

Effective altruism and AI is where the effective-altruism project meets artificial intelligence. The project tries to find better ways of helping others and put them into practice. For some members, that has meant careers on AI risks, among other problems.

Also called: EA and AI
Topic: Safety, ethics and risk
Level: Some background helps
First described: William MacAskill writes that, on 5 December 2011, directors of Giving What We Can and 80,000 Hours voted on a name for their umbrella organisation, and Centre for Effective Altruism won by a clear margin.³ He had been pushing for the words effective altruism. A colleague preferred rational altruism.³ 80,000 Hours had soft-launched in February 2011. The post does not discuss AI.³

The effective altruism site defines the project as trying to find the best ways of helping others, and putting them into practice.² It describes a philosophy that aims to identify those ways, and a practical community that tries to use the research to make the world better.² The Centre for Effective Altruism lists its core principles as scope sensitivity, impartiality, a scout mindset and recognition of tradeoffs.¹

The same FAQ points to global-health charities recommended by GiveWell, animal charities recommended by Animal Charity Evaluators, grants from Open Philanthropy, and funds for global health and development, animal welfare, the long-term future, and infrastructure.² AI is not the whole list. 80,000 Hours says many members work on pressing problems including, but not limited to, AI risks.⁴

80,000 Hours says it helped found the community in 2011, and that the principles of effective altruism have led many members to use their careers to address AI risks.⁴ It is 80,000 Hours' own view that risks from advanced AI are the biggest issues facing humanity right now.⁴ Conferences it recommends also cover biosecurity and ending factory farming.⁴

The Centre for Effective Altruism says there are substantial disagreements among people involved, driven by different values and different methodological assumptions.¹ 80,000 Hours says some people disagree strongly with the ideas, or with the way they have been put into practice, and that the AI-safety and effective-altruism communities overlap rather than being the same group.⁴

Example

The Centre for Effective Altruism looks for problems that affect many beings, have relatively few people already working on them, and offer a tangible path to further progress.¹ A person using that test could look at global-health charities, which the effective altruism FAQ lists as community work, or at AI-risk careers, which 80,000 Hours says the principles have led many members to choose.²⁴

Why it matters

80,000 Hours says effective-altruist principles have led many members into careers on AI risks, and that those risks are, in its view, the biggest issues facing humanity right now.⁴

Questions people ask

Is effective altruism mainly about AI?

No. The effective altruism FAQ lists global health, animal welfare, the long-term future and infrastructure among the work.² 80,000 Hours says members work on pressing problems including, but not limited to, AI risks, and that its conferences also cover biosecurity and factory farming.⁴

Why do some people in this project work on AI risk?

80,000 Hours says the principles of effective altruism have led many members to use their careers to address AI risks.⁴ It also says that it, 80,000 Hours, thinks risks from advanced AI are the biggest issues facing humanity right now.⁴ That is one organisation's view.

Do these sources share one view of AI?

The Centre for Effective Altruism says disagreements among people involved can be substantial, and can come from different values and different methodological assumptions.¹ 80,000 Hours says some people disagree strongly with the ideas or with how they have been practised.⁴ The importance, neglectedness and tractability test on the principles page is a general way to compare causes, not an AI policy.¹

Related terms

Sources

Back to the glossary