jamilton

23 ms·

301.
by jamilton 3y ago
People love "numbers go up" games. There's a fair amount of additional complexity, but the sell isn't immersion, it's optimization ("how do I make the number go up faster ?").
302.
by jamilton 3y ago
People have said they're going to stop support the Java edition since the day Microsoft bought Mojang, but I really doubt it's in their current plans. If they were planning to, they'd just stop updating it, there would be no
303.
by jamilton 3y ago
That was my thought too. The previous wave of articles on this said part of their contractors were paid $2/hour: https://time.com/6247678/openai-chatgpt-kenya-workers/
304.
by jamilton 3y ago
Right, so the problem is the entire economy, not this job existing and paying $15 and hour.
305.
by jamilton 3y ago
Neat. Not sure what negatives people are seeing? Looks like people communicating to me. Mostly English, but some Japanese, Arabic, Chinese, Korean, and Portuguese too, that's interesting.
306.
by jamilton 3y ago
The models seem charmingly low-poly from the gifs.
307.
by jamilton 3y ago
I agree re: sci-fi or serious. I could see "intentionally intermediate most communication through AI" being a niche subculture some years from now, but I wouldn't actually expect it to be a panacea to communication.
308.
by jamilton 3y ago
You can do the same thing on vast.ai too. It's a little inconvenient to use non-base models and plugins this way (you pay extra for more storage), but it's definitely an easy way to use the full power of SD.
309.
by jamilton 3y ago
There's a higher "skill ceiling" with SD. You can install different models for different styles or subjects, use ControlNet for composition, and use plugins to do things you can't easily do with MJ.
310.
by jamilton 3y ago
ControlNet helps a lot with composition and lighting ( https://sandner.art/create-atmospheric-effects-in-stable-dif... ). It requires more work than just entering a prompt, but probably less work than doing it manually once y
311.
by jamilton 3y ago
The image quality of DeepFloyd is much lower than Stable Diffusion 1.5 though, it's a pretty major tradeoff. Can definitely be part of the workflow since it really is good at composition, but right now it's not a replacement.
312.
by jamilton 3y ago
It'd have to be a pretty big object, I think. Your reasoning alone is not enough for me to consider it not-weird. If there was evidence that there was a windstorm in the area at the time and/or that a big piece of sheet metal fell
313.
by jamilton 3y ago
If you calculate and come up with the answer that it's really likely we should be seeing life, is it more likely that life is hiding or that one of your assumptions is wrong? Maybe try adding confidence levels to your estimates or chan
314.
by jamilton 3y ago
But they wouldn't go a little further to avoid being seen at all?
315.
by jamilton 3y ago
You can share passwords easily, I don't think you can trivially share passkeys.
316.
by jamilton 3y ago
Misconceptions. There's no inherent reason a false statement would have lower probability than a true one. To be clear, I'm referring to things like GPT-3.5 reportedly consistently messing up on statements like "what's h
317.
by jamilton 3y ago
I'm saying getting banned from the top subreddits deprives you of very little value.
318.
by jamilton 3y ago
That framing fails to describe the case where the model is confident in a response (at the token-level), and is wrong, which I think is still considered hallucinating.
319.
by jamilton 3y ago
The other million subreddits are most of the good part of reddit, IMO.
320.
by jamilton 3y ago
Aseprite calls itself "source available" now. Their license has a "no redistribute" clause, so if want to use it and can't compile it yourself, you have to buy it. I spent about an hour trying to compile it on Windo
321.
by jamilton 3y ago
Yeah, I think this is an experience some people seek out today.
322.
by jamilton 3y ago
If AIs will be “corporations but faster and/or smarter”, and you accept that corporations behave unethically, AIs seem plainly worse. I don’t think “AI = corporation” is a useful framework, current AI doesn’t particularly resemble cor
323.
by jamilton 3y ago
I think Sydney is worrisome in that it shows that even Microsoft, spending tons of money, can’t (or doesn’t want to) stop LLMs from going “off-script”. If your concern is “will Microsoft et. al be careful with deploying new tech”, Sydney w
324.
by jamilton 3y ago
A Minetest game I recommend is Nodecore: https://content.minetest.net/packages/Warr1024/nodecore/ Description: "A surreal dream-world where everything is possible but nothing is easy. Peaceful and Zen.
325.
by jamilton 3y ago
Who said food shouldn't be a right?
326.
by jamilton 3y ago
It's not for everyone, and that includes a lot of landlords.
327.
by jamilton 3y ago
That pre-supposes that it's creators have read everything useful that GPT-4 has read and put it into practice. It's likely it wouldn't have any useful improvements for the overall architecture because it's novel relative
328.
by jamilton 3y ago
Wow, that's funny, that's my high score too. I guess there's a limited amount of possibilities, so it's probably not that unlikely. Seed was 5209605, if anyone wants to try it.
329.
by jamilton 3y ago
I don't think they're saying that's a goal, I think they're curious if it is the case. LLMs are bad at arithmetic, this uses a LLM to process the prompt, that class of result seems plausible.
330.
by jamilton 3y ago
What do you mean by true unemployment?

More ›