Thanks for publishing it :)
There is no direct analogy to recombination in gradient descent.
In a eukaryotic organism, only the germline will continue indefinitely– all other (somatic) cell lines’ days are numbered.
The germline is strictly separated from the soma only in non-sponge animals.
One idea is that accessory genomes can serve as a scratch space while the core genome is more conservative.
Super interesting …
I only took a brief look at the post so perhaps you’re covering some of that in the text but, quoting from the end of George Stiffman’s article America Doesn’t Know Tofu in Asterisk:
The ability of Chinese craftspeople and chefs to turn humble plant-based ingredients into dazzling culinary experiences is on par with the highest gastronomy in the West. But to the creators, these foods are rarely seen as “art.” They are subsistence. To consumers, these foods are not pride and treasure. They are relics of poverty, discardable afterthoughts en route to modernization.
This trend might appear to affirm a doctrine of economic development: that rising income increases demand for meat. But I wonder if this is the wrong lesson to draw. Chinese people don’t reject common vegetarian foods because there is something _fundamentally_ more valuable about meat. They do so because of _perceived_ value — associations of plants with poverty and meat with prosperity.
I think this fact is lost on many animal advocates in the West. Over the last few decades, investors have poured billions of dollars into companies attempting to replicate the experience of eating meat, dairy, and eggs. These products won’t succeed, however, on cost, taste, and convenience; they need to win on perceived value.
Also from MIRI: Abram Demski has lots of interesting ideas
Connor Leahy—CEO of Conjecture, an AI Safety startup. You’re gonna get a lot of valuable hot takes from him.
Rob Miles—AI Safety YouTuber, although recently he’s been busy with other stuff.
Dawn Drescher—CEO of ImpactMarkets, so knowledgeable about funding in the EA/AI Safety space, but also great person to talk to about a lot of stuff.
Oliver Habryka—LW Admin
Perrin Walker—aka SolenoidEntity, the voice behind SSC/ACX podcast and many other projects
Karl von Wendt aka Karl Olsberg—German sci-fi writer, educating the German public about AI X-risks
Kelsey Piper—no intro needed
Geoffrey Hinton ans/or Yoshua Bengio—likewise
Makes sense, thanks
Possibly naive question: if Non-Linear have material that undeniably rebuts these accusations and they only need to sort it out/organize for presentation, why not publish it in a disorganized/scrambled format, sort it out later and then publish clean/sorted out version? In this way, they will at least show that they’re not scheming anything and are honest about why they asked Ben to delay the post.
What am I missing?
Hi, Aline! We had some technical issues with Airtable and had to create a new one: https://airtable.com/app1JWwSscnpSUgNp/pag1IXp0QTiU5iWrB/form (I also replaced the hyperlinks in the article).
Thanks for pointing this out. I went through all the postings yesterday to replace the link but for some reason EA Forum escaped me.