Mitigating Existential Risks: Aging, Cosmic Threats, Nuclear War, and AI

Key Quotes

"I think that the best thing that can be done about it is nuclear arms control and diplomacy and I can't really think of a scenario where the technology goes away it's also very difficult to think of a technology that can with 100 accuracy remove the risk posed by an oncoming warhead they go incredibly fast they're hypersonic you have to be incredibly accurate and fast to have a hope of catching one of them let alone all of them and a hundred could be launching at once."

The author, Tom Ough, argues that nuclear arms control and diplomacy are the most effective ways to mitigate the risk of nuclear war. He highlights the difficulty of developing defensive technologies against hypersonic missiles, suggesting that negotiation is the primary defense against this existential threat.

"The thick skinned weirdo who is the the archetype who's most likely to call attention to some big risk before it's commonly accepted that it is a risk and so there's there's you know people who did that about asteroids and there are people who did that about ai and we might call them cassandras but certainly there's like a type of psychology where you're quite disagreeable and quite open minded and quite resilient and those typically are the first people within some some field maybe longevity as well at the moment you had to be quite thick skinned to to be part of that field I think."

Tom Ough describes a specific psychological archetype, the "thick-skinned weirdo," who is crucial in identifying and raising awareness about significant risks before they are widely recognized. He notes that individuals with traits like disagreeableness, open-mindedness, and resilience often pioneer these fields, enduring skepticism until their concerns are validated.

"I think that the best counter argument to what I've just said is that humanity is continuing despite aging existing right so it doesn't really matter much to the continuation of the species that three of us are going to die in fact it might be quite good in some ways because it creates room for it creates room for the next generation."

Tom Ough presents a counterargument to the idea that aging is an existential risk, acknowledging that humanity has persisted despite it. He notes the perspective that natural death and aging can be beneficial by making space for new generations, a viewpoint he contrasts with his own sympathy for longevity science.

"I think that the most pragmatic thing to do in respect to those is to have lots of redundancy in in our systems and so one good example of these are kind of submarine internet cables and I write in the book about how there's an awful lot of them going through the straits of Malacca which is quite close to a few active volcanoes and if a volcano erupts then there's a chance that places like that might have kind of turbidity currents under under the surface of the sea and that could mess with the cables so I would like to see us having you know more satellite internet and and more submarine cables that kind of thing so you know we can't stop these volcanoes going off but we we can sure that there's some mitigation ready."

Tom Ough suggests that redundancy in systems is the most practical approach to mitigating risks from natural disasters like volcanic eruptions. He uses submarine internet cables as an example, advocating for more satellite internet and diversified cable routes to ensure connectivity even if certain areas are affected by volcanic activity.

"I think that the ai safety movement has you know raised the alarm early and there's a lot of concern around the time that gpt 4 was unveiled and that the fear is that kind of you know within a short number of years life would be unrecognizable and that hasn't happened yet I I think for for many reasons principally because it's actually quite difficult to get far above human performance in in certain tasks because all the data is based on human performance and we we can't train ais in a very energy efficient way yet so so things haven't accelerated so far in the way that the kind of most worried safety people were forecasting."

Tom Ough observes that the AI safety movement has proactively raised concerns about rapid AI advancement, citing the period around GPT-4's release as an example of fears that life would soon become unrecognizable. He explains that this accelerated change has not yet materialized due to challenges in surpassing human performance across all tasks and limitations in energy-efficient AI training.