The author of The Precipice has reassessed the existential risk from artificial intelligence, pandemics, climate change, and nuclear war.
In a speech at the EA Global Bay Area conference earlier this year, Ord discussed his current outlook on existential risk, and how the events over the last few years have changed it.
The good news – Ord is now more optimistic with regards to our chances of avoiding the very worst outcomes of man-made climate change. This stems from updated models of the effect of increased carbon in the atmosphere on global temperatures.
The Intergovernmental Panel on Climate Change (IPCC)’s estimated range for ‘climate sensitivity’, a metric which takes into account the many positive feedback loops of global warming, had sat at 1.5 – 4.5°C for many years. But a 2021 report from the IPCC estimates the range to be 2.5 – 4°C.
So, whilst the mean has gone up, suggesting more global warming than previously thought, the most extreme possibilities of a runaway greenhouse effect are now less likely; leading Ord to believe the chance of extinction from climate change has decreased.
One X-risk Ord feels has become more likely is nuclear war. When he wrote The Precipice, Ord felt that ‘nuclear war had seemed like a forgotten menace — something that I had to remind people still existed’.
Russia’s invasion of Ukraine has led to increased tensions on the world stage, which makes the expiration of New START (a nuclear arms reduction treaty) all the more worrisome. With the treaty set to expire in February 2026, Ord explains that ‘a new replacement treaty would have to be negotiated’. Donald Trump, the current favourite to win the 2024 US Election, intended to let the agreement run out during his previous stint as president.
Ord also mentions the decreased funding for advocacy work that seeks to reduce the number of nuclear weapons in the world as another cause for pessimism.
Nuclear war and climate change were both estimated in The Precipice to have a 0.1% chance of causing an existential catastrophe in the next 100 years. It’s not clear how strongly Ord has updated each estimate, but what we can say is p(extinction from climate change) < 0.1%, and p(extinction from nuclear war) > 0.1%.
Ord’s assessment of the X-risks from pandemics and artificial intelligence are fuzzier, as there have been positive and negative changes in both areas.
The COVID-19 pandemic reduced Ord’s trust in institutions such as the WHO, but the record-breaking speed of vaccine development gave him cause for hope.
On AI, the incredible advancements in capabilities, the worrying race dynamics between companies, and the laughably misaligned behaviour of certain large language models probably would have increased Ord’s p(doom) of 10%, if only it weren’t for the great strides made in governance and various experts smashing the Overton window by openly talking about the existential threat of AI.
We’ve seen two of the ‘godfathers of AI’ sign an open letter stating that the existential risk from artificial intelligence should be taken at least as seriously as that of climate change and nuclear war, the UK hosted the AI Safety Summit, and countries around the world are mounting substantive regulatory efforts.
Ord ends his section on AI with the following paragraph:
‘All told, this was an amazing shift in the Overton window in just 16 months. While there is by no means universal agreement, the idea that existential risk from AI is a major global priority is now a respected standard position by many leaders in AI and global politics, as is the need for national and international regulation of frontier AI.’
