For a while now, it’d seemed to me that a lot of AI Alignment researchers had attitudes towards the problem that were, somehow, painfully counter-productive. This post puts it into words perfectly.
Some more ways in which fear is a worse motivation than excitement:
It’s less sustainable. Working under fear is unpleasant, which means you’ll do it less, and you’ll burn out faster. An excitement-based motivation, on the other hand, is self-fueling.
It’s less robust. If you run into some novel complication, or the problem becomes more challenging, a fearful motivation will become even more terrified, or despair and give up. An excitement-based one, on the other hand, will consider the new difficulty a stumbling block at worst, and may even be delighted by the higher challenge (and the greater glory to be gained by surmounting it).
It invites unclear thinking and rationalizations. You know (2); you know that if the problem becomes too challenging, you may give up altogether. Which means that if you run into a novel difficulty, you’ll be looking for ways to explain it away. In the doing, you may assume your way straight out of reality, and try to solve imaginary easier versions of the problem.
It’s damaging in the long-term and at macro-scale. If you conceptualize yourself as constantly operating in fear of something, that cultivates a self-image of being small, weak, helpless — and that will gradually lead to you semi-consciously compromising your own efforts, so that they’re in-line with that image. Being motivated by excitement, on the other hand, cultivates a self-image of someone who is up to the challenge — which, while it won’t magically improve your capabilities, will at least ensure you’re employing them fully.
The focus above is only on the negative effects of fear and the positve effects of being excited, on the contrary:
Sustainability: Humans reproduce fast compared to other mammals with a similiar size. This describes human condition within the last 50.000 years. Humans died early and often.
Robustness: Fear is an awesome motivator. Fear can optimize situations and can help overcome stumbling blocks, produce heightend situational awareness over long periods of time and is a motivation to not fail.
Unclear or over thinking: Shock or surprise (elements of fear) can make you just act and stop rationalizing.
Long term: Being excited in a care free enviroment that optimized scarsity causes https://en.wikipedia.org/wiki/Lifestyle_disease and may cause the use of fear to be politically and culturally manipulated b/c there is very little to be afraid of.
Indeed, a specific mix of fear and excitement is probably much better for certain problems than either of the two. Fear to ground you and remind you to be sober about the problem, excitement to provide a lasting and powerful motivation.
But empirically, it seems to me that a lot of people are operating on fear/excitement mixes that are skewed too much towards “fear”, unproductively so.
For a while now, it’d seemed to me that a lot of AI Alignment researchers had attitudes towards the problem that were, somehow, painfully counter-productive. This post puts it into words perfectly.
Some more ways in which fear is a worse motivation than excitement:
It’s less sustainable. Working under fear is unpleasant, which means you’ll do it less, and you’ll burn out faster. An excitement-based motivation, on the other hand, is self-fueling.
It’s less robust. If you run into some novel complication, or the problem becomes more challenging, a fearful motivation will become even more terrified, or despair and give up. An excitement-based one, on the other hand, will consider the new difficulty a stumbling block at worst, and may even be delighted by the higher challenge (and the greater glory to be gained by surmounting it).
It invites unclear thinking and rationalizations. You know (2); you know that if the problem becomes too challenging, you may give up altogether. Which means that if you run into a novel difficulty, you’ll be looking for ways to explain it away. In the doing, you may assume your way straight out of reality, and try to solve imaginary easier versions of the problem.
It’s damaging in the long-term and at macro-scale. If you conceptualize yourself as constantly operating in fear of something, that cultivates a self-image of being small, weak, helpless — and that will gradually lead to you semi-consciously compromising your own efforts, so that they’re in-line with that image. Being motivated by excitement, on the other hand, cultivates a self-image of someone who is up to the challenge — which, while it won’t magically improve your capabilities, will at least ensure you’re employing them fully.
This post also seems relevant here.
The focus above is only on the negative effects of fear and the positve effects of being excited, on the contrary:
Sustainability: Humans reproduce fast compared to other mammals with a similiar size. This describes human condition within the last 50.000 years. Humans died early and often.
Robustness: Fear is an awesome motivator. Fear can optimize situations and can help overcome stumbling blocks, produce heightend situational awareness over long periods of time and is a motivation to not fail.
Unclear or over thinking: Shock or surprise (elements of fear) can make you just act and stop rationalizing.
Long term: Being excited in a care free enviroment that optimized scarsity causes https://en.wikipedia.org/wiki/Lifestyle_disease and may cause the use of fear to be politically and culturally manipulated b/c there is very little to be afraid of.
Oh, I agree that some amount of fear can be useful in some circumstances. As Richard noted, the law of equal and opposite advice still applies here.
Indeed, a specific mix of fear and excitement is probably much better for certain problems than either of the two. Fear to ground you and remind you to be sober about the problem, excitement to provide a lasting and powerful motivation.
But empirically, it seems to me that a lot of people are operating on fear/excitement mixes that are skewed too much towards “fear”, unproductively so.