I think this is a highly inaccurate model, and not useful. By adding risk and reward to determine outcome value, Tao is claiming that outcome value is increased by riskiness. Similarly, high risk situations with low rewards have relatively no downsides (assuming there are no external shocks).
This is nonsensical.
When people talk about high risk situations, they tend to conflate two things: the magnitude of potential downsides and the likelihood of success. To make his model meaningful, Tao would need to incorporate that from the start.
I strongly recommend Fooled By Randomness as primer for thinking about this.
There exists some (often non-Gaussian) curve for describing risk vs reward. I don’t know how you simplify that… Maybe take for number, one for the average negative value outcome and one for the average positive value outcome, and their associated probabilities. And then just add those together? I’m not sure.
I do not know who he is beyond this Mastodon post, and I do not believe I can “dismiss” him as a person. But I can recognize the relative usefulness (or lack thereof) of the idea on its merits. Being able to do so, with humility, is an important life skill.
> By adding risk and reward to determine outcome value, Tao is claiming that outcome value is increased by riskiness.
Tao talks about minimising "value at risk" which he defines as risk MINUS reward. If you prefer to switch signs and talk about maximising "outcome value" then that would be reward MINUS risk.
“For instance, a "safe" action might have a reward of 5 and a risk of 3, which I would represent as 5±3 in this model, the return would typically range anywhere from 2 to 8 units of net benefit.”
He is literally adding risk and reward to determine the upper net benefit bound.
I wonder if you missed the three follow-up posts? It's really in those that his analysis takes shape. What you call the "upper net benefit bound" plays no part in it.
This is nonsensical.
When people talk about high risk situations, they tend to conflate two things: the magnitude of potential downsides and the likelihood of success. To make his model meaningful, Tao would need to incorporate that from the start.
I strongly recommend Fooled By Randomness as primer for thinking about this.
There exists some (often non-Gaussian) curve for describing risk vs reward. I don’t know how you simplify that… Maybe take for number, one for the average negative value outcome and one for the average positive value outcome, and their associated probabilities. And then just add those together? I’m not sure.