4 ms·
Referring to AI alignment and specification of an objective that does not include the prioritization of our genetics somehow.
by janto 5y ago
Referring to AI alignment and specification of an objective that does not include the prioritization of our genetics somehow.
- catlifeonmars 5y agoOk I’ll bite. I’ll assume for a moment that, as you postulate, morality is purely a function of genetic fitness. Even if that was true, genetic adaptation occurs at evolutionary time scales. In a rapidly changing environment, genetics will always be out of phase with what is optimal. In other words, any sense of what is “good” or “evil” will always be outdated by hundreds of thousands of years at the least.
- janto 5y agoI guess I'd say that our genetically encoded behavior has a good track record over many millions of years of latching onto what "the right thing" is. That's quite a few chunks of 100k years that it was "outdated" yet still did ok given the conditions. I agree our slow-to-respond genetically encoded behavior/values might not be fully optimal for survival in the modern world. And that there are now threats ("evils" maybe) that our emotions fail to correctly respond to.