• 0 Posts
  • 26 Comments
Joined 1 year ago
cake
Cake day: July 13th, 2023

help-circle




  • Japan is probably the highest trust society in the planet. You regularly walk into Lawson’s or 7-11 in tokyo that have no employees visible and very few cameras if any. You self check out and are on your way.

    Meanwhile in the bay area more and more regular goods in supermarkets and drug stores are locked in boxes you need to find an employee for to unlock.

    Trust in society makes everything more convenient and easier, it’s an invisible tax multiplied into every daily transaction and the US is in the “close to critical failure” state with almost nobody trusting anyone. No experts, no government, just snakecharmers like Don who change trust for blind faith. And make no mistake, people who have to question every interaction every day eventually are so worn down, the snakecharmers can pick them up by selling hope that you only trust them and all will be ok.

    This is no coincidence. Common narratives in the west, through the GDP lens declare Japan a depression ridden shithole doomed by demographic decline - But the GDP lens misses an entire variable.

    Quality.

    In the west, we have sold quality to growth. A phone that breaks every year sells a new phone. A dish made with lower quality ingredients makes more profit. Quality of life and convenience in Japan is incredibly high. Stuff works. Reliably. Your train is never late. Our escalator doesn’t break down. Your power doesn’t go out. Your food is generally high quality. There a small convenience store every 50 meters - no trips to large walmart megastores requires.

    Yes Japan has issues, this isn’t about otaku fawning here - this is about the fact that we sell quality in the west for profit which does not fly in Japan and if you visit Tokyo in 2023 from SF or Seattle, … the contrast is stark. Somehow every car got replaced with electric or hybrid versions. The trains running on Yamanote are new. The connectivity 5G. Shit just works. The food is highly affordable and it’s quality hasn’t declined.

    Of course people visiting from low trust societies without social compact to not fuck shit up are going to behave like barbarians. Especially when this behavior is incentivized by views / engagement or monetisation.

    I’m with Singapore on this one. Someone who uploads a video like that should be caned and their social media accounts force wiped to start over to disincentive any possible gains






  • That’s a fundamental misunderstanding of how diffusion models work. These models extract concepts and can effortlessly combine them to new images.

    If it learns woman + crown = queen

    and queen - woman + man = king

    it is able to combine any such concept together

    As Stability has noted. any model that has the concept of naked and the concept of child in it can be used like this. They tried to remove naked for Stable Diffusion 2 and nobody used it.

    Nobody trained these models on CSAM and the problem is a dilemma in the same way a knife is a dilemma. We all know a malicious person can use a knife for murder, including of children Yet society has decided that knives sufficient other uses that we still allow their sale pretty much everywhere.




  • Game industry professional here: We know Riccitello. He presided over EA at critical transition periods and failed them. Under his tenure, Steam won total supremacy because he was trying to shift people to pay per install / slide your credit card to reload your gun. Yes his predecessor jumped the shark by publishing the Orange Box, but Riccitellos greed sealed the total failure of the largest company to deal with digital distribution by ignoring that gamers loved collecting boxes (something Valve understood and eventually turned into the massive Sale business where people buy many more games than they consume)

    He presided over EA earlier than that too, and failed.

    Both of times, he ended up getting sacked after the stock reached a record low. But personally he made out like a bandit selling EA his own investment in VG Holdings (Bioware/Pandemic) after becoming their CEO.

    He’s the kind of CEO a board of directors would appoint to loot a company.

    At unity, he invested into ads heavily and gambled on being able to become another landlord. He also probably paid good money on reputation management (search for Riccitello or even his full name on google and marvel at the results) after certain accusations were made.




  • None of that really works anymore in the age of AI inpainting. Hashes / Perceptual worked well before but the people doing this are specifically interested in causing destruction and chaos with this content. they don’t need it to be authentic to do that.

    It’s a problem that requires AI on the defensive side but even that is just going to be eternal arms race. This problem cannot be solved with technology, only mitigated.

    The ability to exchange hashes on moderation actions against content may offer a way out, but it will change the decentralized nature of everything - basically bringing us back to the early days of the usenet, Usenet Death Penaty, etc.




  • I think at this point we are arguing belief.

    I actually work with this stuff daily and there is a number of 30B models that are exceeding chatGPT for specific tasks such as coding or content generation, especially when enhanced with a lora.

    airoboros-33b1gpt4-1.4.SuperHOT-8k for example comfortably outputs > 10 tokens/s on a 3090 and beats GPT-3.5 on writing stories, probably because it’s uncensored. It’s also got 8k context instead of 4.

    Several recent LLama 2 based models exceed chatgpt on coding and classification tasks and are approaching GPT4 territory. Google bard has already been clobbered into a pulp.

    The speed of advances is stunning.

    M- architecture macs can run large LLMs via llama.cpp because of unified memory interface - in fact a recent macbook air with 64GB can comfortably run most models just fine. Even notebook AMD GPUs with shared memory have started running generative AI in the last week.

    You can follow along at chat.lmsys.org. Open source LLMs are only a few months but have started encroaching on the proprietary leaders who have years of headstart