Wait, really? Are you highly confident you're thinking of this video, and not the similarly-styled Beautiful Tomorrow video?As far as I knew, this video didn't play for anybody -- I don't think it loaded for any of the organizers.
I usually assume things like this are at least partially fake; an army of low-wage employees is both cheaper and more convincing than the current state of the art in AI. That said, I see that this was originally a Microsoft project, and the Wikipedia article quotes Bill Gates (who I wouldn't expect to directly lie about this intentionally.) Now that it's spun off, it's in other hands, though. I also wouldn't expect that a fake AI would have had to be censored to keep it from talking about politically sensitive topics, which apparently the real one did. (I assume low-wage employees pretending to be an AI would automatically know what topics to stay away from, assuming they are from the same country.) So I'm not sure what to believe.
I have used Immersed for screensharing movies in VR. (We're both mac users, so nothing other than Immersed is possible currently.) It's a bit finicky but you can definitely make it work. I had to lower the resolution of the shared screen until the streaming framerate rose to acceptable levels, but then it was generally great.The main advantage of this over the more typical approach with a video call is that you can get more of a sense of 'presence' -- you can't see the other person's face, but an avatar can in some ways feel more expressive (you can see their head movements and hand movements.)However, at least when we tried this, the Immersed screensharing worked much better than the Immersed multiplayer avatars, so we ended up going back to video chat. I expect it's improved since then, though. (The BigScreen avatars are super cool, an incredibly strong sense of presence, but we can't screenshare on BigScreen because that feature is Windows-only. Very frustrating.)(Disclosure: I made a small investment in Immersed, because I think it's super cool.)
I think you are misunderstanding what 'crazy distortions' paul is referring to (that's on the tax end when you are taking money away, which mainly affects the rich; not the UBI end when you are giving it out, which mainly affects the poor.)On the UBI end, you should expect to see that creating a UBI will, in equilibrium, cause the cost of living to rise by an amount less than the amount of the UBI. If the cost of producing the goods required to feed everbody requires nearly the entire productive output of society -- as in the medieval scenario you are pointing at -- I believe you should expect a small UBI to have almost no net effect in equilibrium, because (as you observe) it will be almost completely absorbed by rising prices.It is a mistake to see rising prices as "an indicator of damage". They are an indicator of damage that would be done if all that production ceased, sure. But you're miscounting if you call them an indicator of damage that is done, because the whole value of the price mechanism is that it is not done, because prices rise until production is again adequate to meet demand.The rise in prices means that the net effect of a UBI is reduced relative to the gross effect, but that's not an "inefficiency". It's not causing any value to be destroyed, it's just a change of accounting. You get some more money, you spend some more money (but less than you get). The real net redistributive effect of the UBI is in the excess you get beyond the increase in what you spend. (And as you observe, if there is not much spare productive capacity, that amount could be small.)
It seems like something went wrong with the post mirroring here. It's got a bit of it, and then cuts off without any indication that there's more.
Coauthor here: FWIW I also favor eventually switching to the (more reasonable IMO) streaming approach. But this does require a lot more complexity and state on the server side, so I have not yet attempted to implement it to see how much of an improvement it is. Right now the server is an extremely dumb single-threaded Python program with nginx in front of it, which is performant enough to scale to at least 200 clients. (This is using larger than 200 ms windows.) Switching to a websocket (or even webrtc) approach will add probably an order of magnitude in complexity on the server end. (For webRTC, maybe closer to two orders, from my experiments so far.)
Do you have examples of that kind of output for comparison? (Is it reproducing formatting from an actual forum of some kind, or the additional "abstraction headroom" over GPT-2 allowing GPT-3 to output a forum-type structure without having matching examples in the training set?)
I like this a lot. I would also like to hear a post-mortem from the winner in a lot of cases, although of course it's kind of silly to impose it. But I do sometimes see the winner and the loser agree that the bet turned out to be operationalized wrong -- that they didn't end up betting on the thing they thought they were betting on. I'd like to know whether the winner thinks they won the spirit of the bet, as well as the letter.
I am still skeptical of the strength of "MNM" effects. Control systems with huge lag times are infamously unstable. Are most people really able to judge whether they should be scared or not based on the R value from a week or two ago, which they don't even know but have to eyeball from the trend in cases?
Right, I was thinking the same thing -- not just a person, but medical personnel. So you're going from patient 1, to someone's hands, who is then directly touching patient 2, plausibly even patient 2's mucous membranes. That's much more direct than a typical fomite contact, which is more like face-hands-fomite-hands-face (or if you sneeze on a doorknob, face-fomite-hands-face.)