LM Studio looks cool, but I wonder, why their GUI app isn't open-source? Also their site has careers section, where do they get money to operate like that? Couldn't find anything about their monetization model.
Also, lazier. I'm more likely to stick with information from the first 1-3 search results I decided to click, while AI will parse and summarize dozens in fraction of time I spend reading just one.
In this study they asked to replicate 1:1 headline publisher and date. So for example if AI rephrased headline as something synonymous it would be considered at least partially incorrect. Summarization doesn't require accurate citation, so it needs a separate study.
I use it instead of search most of the time nowadays. Why? Because it does proceed to google it for me, parse search results, read the pages behind those links, summarize everything from there, present it to me in short condensed form and also provide the links where it got the info from. This feature been here for a while.
A bit offtopic, but why would anyone want to keep their instance in line with local laws? Aren't internet sites operating under jurisdiction of where they are hosted? Or is it just some coincidence that those people decided to host their stuff at datacenters at their local proximity? When I'm choosing hosting the first thing I think about: "hmmm I shouldn't host in country where I live because I don't want to ever have any problem with local authorities, and if I host elsewhere authorities there won't be able to reach me physically so the worst thing that could happen is the site gets shut down".
I also asked ChatGPT itself, and it listed a number of approaches, and one that sounded good to me is to pin layers to GPUs, for example we have 500 GPUs: cards 1-100 have permanently loaded layers 1-30 of AI, cards 101-200 have permanently loaded layers 31-60 and so on, this way no need to frequently load huge matrices itself as they stay in GPUs permanently, just basically pipeline user prompt through appropriate sequence of GPUs.
That’s how llms work. When they say 175 billion parameters, it means at least that many calculations per token it generates
I don't get it, how is it possible that so many people all over the world use this concurrently, doing all kinds of lengthy chats, problem solving, codegeneration, image generation and so on?
It is what it feels like, but it's not really 100% this way (yet). It is a bad self-reinforcing cognitive bias: we think "forums are dead, that's why we stick to the sitename" instead of actually finding dozens of still alive forums and going there, in turn sitename gets more populated while forums feel more dead. But there are still plenty alive. Also, there are relatively new kinds of forums which sometimes work very well for their niche, like Discourse communities for example.
That's really sad. Can we at least have their stuff now? I mean it would be a waste to not open-source all the code and assets they've developed and make them available under MIT, CC and similar licenses.
Cool idea, I almost forgot this feature even exists. I think I dismissed it the past when I realized it's probably not going to be easy to switch VPN servers this way.
It does hurt, your VPN should support proper port-forwarding for soulseek to work well. In most cases, you will only be able to download files, but your shares will be inaccessible. It doesn't seem to work with ProtonVPN for example, even when you built-in port-forwarding feature. And even if it did work, you would need to reconfigure and restart soulseek every time you reconnect the VPN, because their port-forwarding is randomizing the ports and there's no way to turn that off.
Let me explain how Honkai Star Rail handles gearing. Every single character has six relic slots: head, hands, body, feet, planar orb, and planar ornament. These relics go from level 0 to level 15, and four of them have a randomized primary stat. They all feature four randomized secondary stats, and every three levels a random one of those secondary stats gets a bonus. Each relic also belongs to a set of relics, and characters benefit from having two or four pieces of a given relic set. That means for every character in your party, you need to get the right items at the maximum rarity, the right primary stats, the right secondary stats, and the right level-ups for those secondary stats.
This is min-maxer mindset and I would hope randomized systems like this will prevent it but unfortunately no: even here some people think they actually need to roll every dice exactly the right way. I don't think it's true that this is really necessary. And no, it is not necessary to do top 10 world parses; you can just beat endgame content on modest, casual difficulty and call it a day, rather than try hard to set a record.
Even co-op in gacha games doesn't qualify as MMO, because for that you need hundreds or thousands of players being simultaneously in the same persistent world. This is the same reason why games like Dota, League of Legends or Counter Strike aren't considered MMO.
Yes, the cheapest ones might have some risks, I mostly presented it as an example of what the opposite extremity looks like. There is a lot in-between, something a bit more expensive is even more guaranteed win. For example last time I used Hetzner, I had a server with 64gb RAM, 2TB SSD, and 16 cores Ryzen for something like €34/month. Hetzner support is very decent and they're very well known, have decent reputation and been providing their services for a long time.
Maybe the problem is that they are using ridiculously overpriced enterprise services like AWS or Azure, which provide their own solutions for a lot of common things like backups, replicas, logging, etc, but cost 100x more than what you can get with DIY on some cheap VPS if you're fine with spending 1.25x more time.
Also, given that the instance is called “infosec.exchange”, you can be sure that he is not running this on some cheap VPS.
LM Studio looks cool, but I wonder, why their GUI app isn't open-source? Also their site has careers section, where do they get money to operate like that? Couldn't find anything about their monetization model.