Estimates of the expected utility gain of AI Safety Research
When thinking about AI risk, I often wonder how materially impactful each hour of my time is, and I think that this may be useful for other people to know as well, so I spent a couple of hours making a couple of estimates. I basically expect that a tonne of people have put a bunch more time into this than me, but this is nice to have as a rough sketch to point people to. I'm going to make 3 estimates: an underestimate, my best-guess estimate and (what I think is) an overestimate. Starting facts [1] : Currently 8.3 Billion people on planet earth Current median age: 31.1 years Current life expectancy: 73.8 years I am going to commit statistical murder and assume this means that everyone on the planet lives ~42.7 years from this point onwards. Underestimate: 40 years of life left/person Median: 42.7 years + ~15 years' increase in life expectancy (20 years' growth in the past 60 years) = about 60 years of life left Overestimate: Everyone gets life extension and lives to heat death of universe: 10^100 years Since the population is growing, we should take that into account: Underestimate: We only care about the lives of people currently alive Median: We keep growing at current ~1% growth rate per year Overestimate: Population growth of 2% per year until the heat death of the universe Given these parameters, we can figure out the total expected years of life we care about for each scenario: Under:
When considering the impact of AI safety research, many people wonder how much each hour of their time contributes to mitigating potential risks. To address this, the author has created three estimates—an underestimate, a best-guess estimate, and an overestimate—to provide a rough framework for understanding the material impact of time spent on this critical field.
Starting with the current facts, there are 8.3 billion people on Earth, with a median age of 31.1 years and a life expectancy of 73.8 years. The author simplifies these figures by assuming that everyone on the planet will live approximately 42.7 years from the present point onwards. This simplification, known as "statistical murder," is a common practice in such calculations.
The three estimates are based on different assumptions about life expectancy and population growth. The underestimate assumes 40 years of life left per person, the median estimate accounts for a 15-year increase in life expectancy (based on historical growth rates), resulting in about 60 years of life left, and the overestimate considers life extension to the heat death of the universe, which is approximately 10^100 years.
Population growth is also factored into these estimates. The underestimate focuses only on the lives of people currently alive, the median estimate assumes a 1% annual growth rate, and the overestimate considers a 2% annual growth rate until the heat death of the universe.
Using these parameters, the total expected years of life for each scenario can be calculated. For the underestimate, this results in 40 years multiplied by 8.3 billion, equating to 332 billion years. The median estimate calculates the current population's life span as 60 years multiplied by 8.3 billion, with an additional population growth component. The overestimate, however, leads to an impractical calculation due to the exponential growth of population and years, making it best to skip this scenario.
For the underestimate, the author assumes 20 years of research to achieve a 1% chance of a 1% decrease in the final risk for the entire field. In this scenario, extinction is projected to occur 30 years from now. The best-guess estimate and the overestimate provide a range of possibilities, allowing individuals to gauge the potential impact of their efforts in the field of AI safety research.
These estimates serve as a rough framework for understanding the material impact of time spent on AI safety research. While the overestimate may seem extreme, it highlights the potential for significant benefits if AI risks are mitigated effectively. The underestimate and best-guess estimate offer more realistic scenarios, emphasizing the importance of continued research and collaboration in this critical area.
