It is more obvious when taken to extreme: With the current feedforward transformer architectures, there is a fixed amount of compute per token. Imagine asking a very hard question with a yes/no answer to an LLM. There are infinite number of cases where the compute available to the calculation of the next token is not enough to definitively solve that problem, even given "perfect" training.
You can increase the compute for allowing more tokens for it to use as a "scratch pad" so the total compute available will be num_tokens * ops_per_token but there still are infinite amount of problems you can ask that will not be computable within that constraint.
But, you can offload computation by asking for the description of the computation, instead of asking for the LLM to compute it. I'm no mathematician but I would not be surprised to learn that the above limit applies here as well in some sense (maybe there are solutions to problems that can't be represented in a reasonable number of symbols given our constraints - Kolmogorov Complexity and all that), but still for most practical (and beyond) purposes this is a huge improvement and should be enough for most things we care about. Just letting the system describe the computation steps to solve a problem and executing that computation separately offline (then feeding it back if necessary) is a necessary component if we want to do more useful things.
Not only that but also LLMs "think" in a latent representation that is several layers deep. Sure, the first and last layers make it look like it is doing token wrangling, but what is happening in the middle layers is mostly a mystery. First layer deals directly with the tokens because that is the data we are observing (a "shadow" of the world) and last layer also deals with tokens because we want to understand what the network is "thinking" so it is a human specific lossy decoder (we can and do remove that translator and plug the latent representations to other networks to train them in tandem). There is no reason to believe that the other layers are "thinking in language".
Haxe is amazing, has macros etc. A force multiplier if you are a solo developer for sure. But damn, you feel kinda alone if you are using it. Not everything is an npm install away which negates your velocity gains from using a saner language.
I RARELY/NEVER have to build an app so fast that i just fart out whatever broken code as fast as my fingers can type. IF i get a project like this with a deadline of "yesterday" i politely just refuse. I will be wasting my personal time, and the clients time. And the result will be a broken mess that will eventually take more time to fix, than it would have if i in fact did it "correct" from the get go.
That said Haxe has externs, enabling you to target JS/PHP and use the rich ecosystem both langauges have. The best part of externs is that IF i only use 4 things from given package, i statically KNOW i only use these 4 things, and can refactor more easily, or even build the thing i need myself.
When I'm talking about velocity, I'm not talking about coding fast, but being able to write DRY, flexible yet easily maintainable code that can weather future requirements / refactorings easily. Personally, I'm also talking in the context of my own projects so nobody is breathing down my neck or pressuring me with time. I just want to write good code that is a joy to maintain for years to come.
yes but still, a local model, a lightning in a bottle that is between GPT3.5 and GPT4 (closer to 4), yours forever, for about that price is pretty good deal today. probably won't be a good deal in a couple years but for the value, it is not that unsettling. When ChatGPT first launched 2 years ago we all wondered what it would take to have something close to that locally with no strings attached, and turns out it is "a couple years and about $10k" (all due to open weights provided by some companies, training such a model still costs millions) which is neat. It will never be more expensive.
I only ever used cache to find what google thought was in the site (at the time of crawling) as these days it is common to not find that info in the updated page. For everything else, there is the Internet Archive.
I can't believe people are willingly installing rootkits of dubious origin on their computers to play games (and paying for the privilege). I know cheating is an issue, but still.
It's getting more and more ridiculous like controlling what software your computer can have installed. Pretty sure some even require TPM and that you are booting into a "trusted environment". The only one who suffers is the genuine gamer as cheaters find their way around, either by directly reading and writing from/to the RAM chip directly or using another PC running machine vision and feeding inputs as a emulated USB hid devices.
I don't think cheating is something that can be solved by a technology in FPS games, you can do some server side checks but it gets really difficult unless you stream the whole game as a video feed from a remote server. Cheating is a social issue, and we had working solution for it.. dedicated community servers which were community moderated.
It's ridiculous that gamers are willing to accept installing a rootkit in order to play a game. And it's ridiculous that they even defend the practice by crying about cheaters. And it's ridiculous that game developers reach for rootkits instead of securing their own games better against cheating and/or designing their games in such a way that computer-assistance gives no benefit. It's just ridiculous all the way down.
It's ridiculous that you think developers aren't doing other things apart from kernel level anti-cheat to secure their games. There has been an enormous amount of engineering around combatting cheaters.
Why doesn't Microsoft just build a more secure operating system than reaching for Microsoft Defender and other kernel level protection mechanisms. Well because they are but they're not good enough which is exactly the point.
Gamers who use their PC to play games, will install what they can to play such games its a simple as that. Not everyone has the same tolerance to risk.
Oh that's just the start of it. I recently learned there are cheating methods that use sniffing and live modification of system RAM via PCI Express DMA transfers. As usual, physical access to hardware trumps all.
This is an excellent mini documentary explaining the various techniques to avoid kernel level anti-cheat, including pre-boot EFI-based cheats, PCIe DMA sniffing, video over serial auto clickers, and the state of the art motion detection over streaming video with MITM USB host controller.
Yup. AI hacks are now the norm. They’re Pretty much undetectable too. I use aimmy for shooters, which amusingly claims it’s an accessibility app for the physically disabled.
I love it because suddenly I can actually compete with the sweatlords who spend every waking moment optimizing their play. I don’t have time for that shit, but I want to have fun.
Downvote away, but game cheating is subversive in the best way possible, and the haters can simply deal with it. I think it’s straight up good for society as it encourages a technological race to the top from others who want to stay competitive. There’s more value in learning how to cheat in a game with respect to learning about computers than there is in playing legitimately. Time spent getting good at a random video game is simply a waste of our precious limited lifespans.
I argue we should legalize doping in sports too. Daniel tosh argued for this famously, pointing out that a bunch of roided beefcakes playing sports is far more entertaining for the average viewer since it improves the quality of play.
> Aimmy was designed for Gamers who are at a severe disadvantage over normal gamers.
> This includes but is not limited to:
> - Gamers who are mentally challenged
Hope it was worth your dignity.
Seriously though, the whole point of ranked matchmaking is to give you a fair game with people who are similar in skill. If you compete without cheating, you will get matched up with other casual players instead of "sweatlords".
I hope every person who’s a victim of my AI hacks reacts at least as poorly as you do. The salt, rage, anger, etc of gamers is the funniest stuff ever.
What, with mild sarcasm? The description of the tool is hilarious.
I just legitimately don't see the appeal. I installed a CoD aimbot way back when, and it just ruined the experience. There is no challenge or satisfaction in winning when it's the tool doing all the work.
Cheaters in online competitive games will always be a huge laughing stock for me.
Can you imagine being so silly to go and cheat on a game? Ha!
I used to play a couple of specific competitive games and get the occasional cheater.
It was always a riot to run circles around these people and turn their cheats against them.
We had ways of both making the player know that we knew they were cheating and ruin their game, sometimes for hours.
I went out of my way to hold these players captive with me so they could not go ruin a new player experience.
A service to the gaming society, you see. The more upset and irritated the better.
I didn't have to win, I just had to prevent them from winning. It was such fun.
Still have a bunch of hatemail from these people and cherish every bit of it.
If I didn't get hatemail, I wasn't doing it right.
Why not play and cheat in solo games if you are bad at multiplayer? Only having fun when others are not is a bit disturbing. You could create an AI to help you beat Dark Souls or something, then you would respect both computer learning and human beings...
It’s up to the game devs to do whatever they want. It’s up to me to figure out how to subvert their intentions.
It doesn’t matter about what I am “okay with”, what matters is my power and my capabilities. See the philosophy of max stirner for a deeper explanation.
So I will do what I can to be a game cheater without being detected so that I preserve my advantage and don’t get lumped with other cheaters.
So you’re no different to anyone else who cheats at anything. You’ve come up with this whole philosophy and self aggrandising narrative about why you do what you do. But really you’re likely just covering for some weakness or feelings of impotence that are arising from elsewhere in your life and this is how you cope. Unless of course you’re doing it for monetary gain, but I haven’t seen that in your boasts so far.
If you’re really powerful, dominate people on a level playing field. I’m thinking you can’t and you’re too weak of mind, will or ability to do so.
Highly unpopular opinion: cheating is a social issue and the only future-proof long-term solution is… acceptance and adaptation.
Technology is here to stay. A machine will always outperform unaided humans at some tasks. Don’t make that the point of the competition. The genie is out of the bottle, and save for an apocalyptic event, it won’t ever go back.
Do the contrary. Give every player a state-of-art machine copilot, and let them bring their own improvements to it. This is the only way to make the field truly level again. If your game mechanics is ruined… I’m sorry, but then that - probably - wasn’t a sustainable idea.
People who are called “cheaters” are different. Some exploit bugs and just want to watch the world burn - no sympathy for those folks, fuck them. Some want to trample on everyone without doing anything - no competition here, I don’t get those people (can’t say “fuck them” though - maybe it’s some kind of a trauma they have, so they need that feeling of fake “victory”?). But some want to win, but feel that cannot do so with their bare hands and eyes. So they do what humanity always did - improve by using technology. If they genuinely want to become better - how about we just don’t hate them for this? Heck, the desire to improve through tech is the very foundation of this civilization. (Yes, even if one just buys a cheat program - it still makes sense in any society that had invented money.)
Just rank such players accordingly to their machine-assisted skill. Here, problem solved, and as a bonus you’ll get your next OpenAI Five paper in no time.
I know it’s very controversial. I know some game genres won’t survive (not complex games like LoL or Dota, though). Most likely a lot of MMOs (and most mobile casino junk) will suffer, as a lot of their mechanics is based on boring grind (that’s how they earn money, hah). I know the industry is doing the exact opposite, trying to shove the issue under the rug with bans and memeing super hard that “cheaters” (a derogatory term) are vile scum. And I can see why people are buying it - if the developers say it’s against the rules, no surprise a slightest trace of automation (like a programmable mouse) feels unfair. Reading some Reddit threads I sometimes wonder how those people don’t say that wearing glasses is cheating too. I see that as a conservative approach, and - as anything that merely tries to uphold the status quo - I honestly believe it’s not gonna work in the long term.
No trolling, I honestly believe in what I wrote. And, no, I don’t “cheat” (although I’ve experimented with some basic game hacking, of course - because reverse engineering is fun)
And, uh, yes, I think the same should apply to non-e-sports. The logic is a bit different, of course. But the value of medical breakthroughs drastically overweights the fictional “purity” in my perception of values. I don’t really care if some athlete can do something (doubly so because I don’t have a nationality I can root for; personal achievements are cool but there’s zero benefit for me or society besides the economic value of the competition event), but if some athlete can do twice as much because of some tech (drug or implant), that may be beneficial for me as well. And yes, I’ve seen that standup/meme about dope Olympics - it’s fun but it doesn’t really invalidate my views.
This is a very far fetched tin foil kind of response here
If I am playing a game with other humans and one of those humans doesn't feel like improving or cannot stand being beaten and decides to use technology that is not allowed to win at the expense of others. That's not a game people are going to want to play.
There is room for people that want to experiment with cheating and using technology in game to aid themselves and that's in a completely separate game that encourages that behavior.
If they genuinely want to become better, they need to apply concepts that allow them to improve which is training what you're not good at and accepting you won't always or ever be the best. Not using an aid that goes against the rules of the game and gives them an unfair advantage.
It's a similar principle I would apply from the world of sport and doping. Just because it might be partially a social problem, the solution isn't just to let it happen.
It’s not about not being able to stand being beaten and not being the best. I’m truly sorry that a lot of people who cheat are also toxic.
I’m not advocating for unfairness by just letting people use automation to trample upon those who don’t. Like I’ve said, a machine will always win an unaided human in some tasks - so there is no competition here. Rather, I propose to accept automation and start giving it to everyone, so games will be fair for everyone. If some mechanic relies on imperfect mechanical skills - that’s a bad mechanics that never will be fair despite any wishes to make it so.
Technology is the greatest equalizer. Don’t automatically blindly hate people who want to make bots or copilots. That was the whole point.
Firstly, the decision needs to be made whether we allow people to use automation and these games have already said no, which is the agreed upon consensus for players. We don't want to allow cheats or people to automate their gameplay.
We want the developers to provide the constraints so that there is a fair playground, they do this via the game mechanics and preventing outside automation. We just want to play games using the mechanics provided to us, we don't want to focus on building automation to outsmart the competition, we want to outsmart the competition by being a better player when everything is considered equal.
I am automatically against people who make bots or copilots in games that explicitly say you can't use them, in games where the players don't want them and don't want to up against them. That is the whole point.
These anti-cheat solutions are combatting people making or using copilots in games where people don't want them and where the rules say you can't use them. That is the whole point.
So all competitive action games (for example) should go away and be replaced by competitive coding or scripting of action-game-playing bots?
I'm hearing that cheating is a social issue and we're dumb for using technology to address it. But the social need to compete, here's a whole masturbatory fantasy of how coding is the last battlefield we'll ever need for that. The cunning and slandered cheaters have shown us a better way.
Let's rewrite the whole activity of gaming so that people who are bad at aiming guns but do know how to install a script can get a trophy.
Your comment was dead for some reason. I’ve vouched for it.
I believe that games that are purely competition of reflexes are inherently unfair. Bots are merely exposing this unfairness because they can reliably beat all kind of animals, but it exists in humans as well.
But here’s the thing - I don’t understand cheating in this kind of games (besides people being willingly toxic). It’s no fun at all, you just win and that’s it. There is zero competition. So, yea, you’re right it doesn’t make sense to give bot copilots in such games. Their time will come to an end when humans will learn to enhance their reflexes, until they they’ll be probably the last kind of games to keep the “no bots” flame alive.
Please notice a difference between toxic people who want to trump everyone, and normal people who want to improve their lives by using technology because they have skills for that (money is a tricky thing - I need to think about it more). Today they’re all pushed into the same ostracized group, called “cheaters”.
> Let’s rewrite the whole activity of gaming
Yes - why not rewrite? Humanity had rewritten various principles quite a number of times, and continues to do so. The time for games will come soon (not exactly yet, but it’s rapidly approaching and we can see the first signs).
A bot is just a machine (we don’t have GAI and it’s still an unanswered question if we will anytime soon). If the game is truly complex in a good way (LoL surely is not too bad in this regard) - it won’t auto-win it for a human, it will merely remove the routine and grind, exposing the true mechanics, exposing the ultimate meta.
The thing with community curated servers is that there very well can be a server which encourages cheating. And I'm not opposed to giving cheaters a place to be in. Even some games that do not have community servers, do not actually ban the cheaters but instead move them to their own server where they will be playing with other cheaters.
I wouldn't say these anti-cheat tools that come with Valorant or LoL are from dubious origins. eSports has grown, real money is on the table and cheating is rampant in many games and it's not always obvious cheating, it's often slight improvements.
People complain, but there are a ton of games that use a kernel based anti cheat system that people haven't been flapping their arms around. Valve anti-cheat, EasyCheat etc.
It's not really clear it has, or is. [0] People, at least in the West, don't seem to think watching video games be played is equivalent to watching sport, which is honestly the least surprising thing ever to me, but I guess I'm not a bombastic YouTube 'content creator'.
I think the main problem has nothing to do with eSports, it's simply that average people don't enjoy games that are rigged against them. Ergo, a profit incentive arises to prevent cheating in multiplayer games. And since a separate profit incentive dictates that all games must be online these days, we end up in a situation where I can't play a single-player game of Madden on Linux because it explicitly blocks Wine, [1] presumably to protect the fairness of multiplayer. (Or more cynically, the integrity of the Madden MUT loot pack casino. Profit motives everywhere!)
>I wouldn't say these anti-cheat tools that come with Valorant or LoL are from dubious origins.
You are talking about Vanguard, developed by Riot which is owned by Tencent which is a Chinese company? China has demonstrated enough grip on companies over the years that anything China can directly influence is dubious. China can decide to do something with the installations and has the power to do it in secret. It is significantly harder in western democracies. The relationship between the state and the private enterprises is entirely different.
Valve's VAC is not in the same league with Vanguard from a technical point of view (it is not even kernel based if I'm not mistaken). I would not let that slide either if it was though.
At least in terms of video games investment, Tencent is known to be very hands-off of the western companies it owns and is often seen to be a better option vs some western conglomerates. Ubisoft turning to Tencent to help it fight off a takeover attempt from Vivendi is a pretty good example.
Vanguard itself is developed by a US-based team and they've stated that they have very little to do with Tencent's anti-cheat team (Tencent uses its own solution for the CN region, including for Riots game) beyond sharing cheat samples.
How often have games coopted end-user computers for nefarious purposes? Yes, the anti-cheat software may often be crap, yet this next step is not a common phenomenon.
I want to make a joke but... the kids might actually believe the world didn't have to worry about malware until their computers were networked. Holy moly. I can't attack this without visual aids of some kind.
As someone who plays a competitive FPS game 5-6 nights - I’m all for anti-cheat that actually works. Online games are plagued with cheaters and it really ruins the experience. If rootkit-style anticheat helps the situation at all I’m personally willing to make that trade off.
That's what the cheaters are doing. It's a cat and mouse game and it's hard to detect cheats that are running in kernel space when your anti-cheat is running in user space.
Given that the average computer has hundreds of pieces of software of 'dubious origin' installed on it, in everywhere from the OS to the driver to the application layer, I can't believe you are seriously baffled that most people don't think twice about adding another one... That's sourced from a reputable vendor.
Reputable, in the sense that you're already willing to install and execute a closed-source binary blob (the game itself) on your PC.
I would not trust Kazakhstan to honor the TLD registrations if this took off and made some noise. Reminds me of Libya taking ownership of all those trendy .ly domains claiming you have to obey Libyan laws and regulations to operate them. Still a fun idea taken quite far!
>claiming you have to obey Libyan laws and regulations to operate them
It being a CCTLD, this is a true claim. At a basic level, these tlds belong to the country, and that country sets the rules. Libya reclaiming from ICANN was a jerk move but their claims are absolutely right. (Most cctlds already have this requirement.)
"claiming you have to obey Libyan laws and regulations"
I always smile when bosses want everything to be GDPR compliant. I am not sure why these laws are more important than the laws from the Chilean Navy. Why are we clicking on cookie popups? We think the EU is smarter than the PII laws from Cameroon? Elitism I say. My websites follow strict guidelines set by proper Constitution of Cameroon doctrines. Every fourth visit to my site we dump all contents in html form (obviously).
> Why are we clicking on cookie popups?
Because people want to track us to make money from invading our privacy?
You don't need a cookie consent banner if your cookies are needed to serve the client with your service. You can do analytics without cookies.
So to answer your question - Why are we clicking on cookie popups?
Because website owners don't want to stop selling your privacy and now have to inform you about it.
> GDPR compliant. I am not sure why these laws are more important than the laws from the Chilean Navy.
Purely market size. Europe is a large market. Same reason that just about every product is labelled with 'known to the state of California to cause cancer' - California is a large market.
Not purely market size, though it's a very important part for sure.
The other part is how likely a country is to try to enforce their laws, and what ability they will have to do so.
Even if a hypothetical US company had an equal number of customers & revenue in Chile as in the EU, if either the Chilean law being broken is one that Chile never bothers to prosecute, or if the worst thing they could do should they find out about the law breaking is to block the service at a national firewall level but not levy any punishments (say, if the US company has no staff or assets in Chile, and the crime has no possibility for extradition or other international collaboration to punish) then the company would be a lot less likely to comply than they are with GDPR. Because most US companies aren't able/willing to serve EU customers without having servers, employees, and revenue, physically in the EU; therefore the worst case for getting caught breaking GDPR is considerably more worth avoiding than if it would just be the EU blocking access to your servers.
I really don't understand how so many people on HN can complain about centralized control, but then so many (other) people are completely against Web3, solutions like Unstoppable Domains are able to let you own a domain and only transfer it if you sign with your key. Why don't more browsers read a Web3-based domain system like Freenames, Unstoppable Domains, ENS, or Filecoin Name Service?
DNS is a federated database, but it is subject to domain seizure etc. at multiple levels. I've seen people complain that their domain operator can just "steal" their domain!
If browsers won't do it, can't someone start a CCTLD (it's only $250K) and then read the blockchain to resolve the DNS records? I realize that this "someone" would be a central point of failure, but alas, that's how the Web currently still works. The best you can do is some sort of "DNS multicast" I think, but it would still be under the control of one company, sadly.
Personally, I'm a bit surprised why the Web hasn't standardized onion links / magnet links / hashes of content / cids / whatever you want to call them. Tor and Beaker Browser have had it for a long time, and Brave too I think. DNS then becomes just a glorified search engine for a small subset of URLs (the ones without a long path / querystring).
No, they can't. ccTLDs are associated with countries. There's no process for creating one that doesn't involve having IANA recognize you as a country.
You're probably thinking of the new gTLD process, which has only been open for applications once, for a brief period in 2012. It's not open to new applications, and the process for applicants was much more involved than a single payment.
Thanks for that link, and to (not-)answer GP's question on price:
> "While the application fee has not been determined, it will be set on a cost-recovery basis. The fee will ensure that the next round of the New gTLD Program is fully funded and does not require funds from ICANN's operating budget. As a point of reference, the application fee for the 2012 round of the New gTLD Program was US$185,000."
because like most things blockchains (cl)aim to solve (primarily money and its transfer, but in this case ownership of domain names), those things do not really really have a centralization problem. 99.999% of the people do not, and do not need to worry about getting their domain name seized. cryptobros like to pose centralization as a huge problem where it really isn't so that they can peddle you scamcoins to pump and to feed their gambling addiction. "web3 based domain system" solves something that is at most a nuisance (and at best a necessary evil) by introducing massive problems into the equation, all to do something that isn't really a problem in practice (and it doesn't even do it, you admit there is still centralisation, so what did we gain by introducing all those problems, really?)
See, this is just dogma that gets repeated on HN. It's obvious to any person who honestly thinks about it, that having a third party in control of your DNS (i.e. what IP addresses it resolves to) means that your entire site can be rugpulled from under you. If it becomes big enough.
Now, you can say, "most people don't care, they just have a small-time operation, just find a reputable domain operator who doesn't have a history of screwing people over." But that's exactly the use case for Web3 and blockchains in general. Why do you have to be forced to trust SOMEONE, with something as important as your brand / identity of your entire organization? And, for that matter, why should an entire community have to trust one guy who can change up the site at any time? That's not very secure, and many of you vehemently insist that no alternative should be made available, to anyone, "because scam"! You don't want browsers to even support it!
As your site gets larger and more people rely on it, you don't want to have that major point of failure at any point. I know that some people on HN go so far as to say that banks freezing your money, and ICE seizing your domain name, are very desirable features of the Internet. So, then don't complain about censorship and deplatforming. You can't have it both ways!
Your domain name and the site it points to is not the end goal, what the site represents is.
Your domain name is one of the means to get more people to know about you or to deliver your product. Decreasingly relevant, note, as no one types domain names usually (people search).
As more people know about you via various channels (most centralized one way or the other: curated lists, social platforms, search), takeover of your domain name (or any other channel) becomes less of a risk. If you take Coca-Cola’s or Apple’s or Basecamp’s domain, they will barely feel it. Perhaps Basecamp could feel it, as it probably plays a bigger role in delivery, but I am sure they would have a procedure specifically to manage that risk.
99% of the time, if you run an ordinary %product%, should you worry that it will be you vs. the world and all of your channels are taken over? Currently, I’d say not. I could be wrong.
You have to do a lot of mental gymnastics to justify why web3 is not needed. Here you literally argue that one’s brand name recognition is irrelevant, and you can be constantly moving domain names with no impact to your bottom line or your community.
That requires more than just an assertion. Extraordinary claims require extraordinary evidence.
> Here you literally argue that one’s brand name recognition is irrelevant
You argued that.
I argue the opposite: brand name recognition is what matters.
(And believe me, I am not the only one who makes that point and I certainly am not even remotely smart enough to have come up with it first.)
A domain name is merely one of the many things that may help you reach that recognition. These things evolve; domain names are less meaningful these days—an Instagram username runs circles around one—and all of those things are less critical the more recognition you achieve.
“If Coca-Cola were to lose all of its production-related assets in a disaster, the company would survive. By contrast, if all consumers were to have a sudden lapse of memory and forget everything related to Coca-Cola, the company would go out of business.” If you have your shiny domain name, but no one knows about you, you are as good as dead. If everyone knows about you, and your domain name gets taken over, you can’t really care less.
Obviously the domain is a large part of the brand.
It's like saying "1800 flowers" can rename itself. Or Jacoby and Myers had a phone number that was all 8's. And then when they split up, they got other numbers like 800-800-8000 but it wasn't the same.
People type in the domain name when they think of chess.com or whatever. You're using examples which are the most ubiquitous companies in the world that spend the most on brand recognition. That's not a great way to support your point!
Consider "basecamp.com" or "hey.com" -- would they do just as well if they had to switch every month to basecamp.nl and basecamp.io ? Probably not. And why should they?
Domain name is completely distinct from public awareness about you. They exist on completely different conceptual levels. It is a key distinction I suspect you are incapable of seeing.
A domain name is one of the means that together can help achieve that awareness and/or deliver your product.
It is like saying an airplane is “a large part of being in New York”. It is useful if you want to fly there, but once you’re there you don’t need it much. You can also drive.
> Consider "basecamp.com" or "hey.com" -- would they do just as well if they had to switch every month to basecamp.nl and basecamp.io ? Probably not.
Again, if you are specifically in the business of subverting the law and expect the world to turn against you and your domain is the main means of delivery then it may be wise for you to do something like this (or simply be a Tor hidden service with the same outcome). For any normal product this does not matter.
Incidentally, public awareness about the pirate bay did not really go down since their domain seizures.
You are way off.
What do you think people do exactly?
> Again, if you are specifically in the business of subverting the law and expect the world to turn against you and your domain is the main means of delivery then it may be wise for you to do something like this (or simply be a Tor hidden service with the same outcome). For any normal product this does not matter.
Ah, that old chestnut! Yeah and you don’t really need end-to-end encryption, unless you’re a criminal who doesn’t want the government to find things out. Let them have the certificate so they can solve crimes easier and keep you safe! Same logic.
I type domain names sometimes, but generally I estimate 99% of people tap a link 99%+ of the time.
> Yeah and you don’t really need end-to-end encryption
I didn’t say you don’t need privacy, you are putting words in my mouth.
> Ah, that old chestnut!
The real chestnut are people who think the only way to go is to abolish institutions, grab a piece of land and guard it with shotguns and rocket launchers. A quick thought experiment would show that this is a dead end.
Many of the same mindsets and people who deride blockchain as “you dont need it if you are not doing anything bad” are also going to do away with end-to-end encryption under the slogan “you don’t need it if you have nothing to hide”.
After all, it can be used to hide ANYTHING, including 2 billion dollar transfers, tax evasion, money laundering and of course supporting terrorism. At least the blockchain is public! The sentiment you express comes on one side of the freedom/security spectrum.
If you’re arguing in good faith, then you’ll have to think deeply why oppose blockchain but others shouldnt oppose end to end encryption for the same reasons of “nothing to hide”. Even I come down on the side of “if you are reduced to sneaking around, then your society is already in bad shape” and consider end-to-end encryption to be a bandaid that makes people complacent. But the war on end-to-end encryption is actually far more prevalent than that around the world, and far bigger than your silly war on blockchains and mere cryptographic signatures (which governments don’t oppose nearly as much):
Read it! And no, the strawman is that I’m talking about shotguns. I’m talking about open software and protocols eating the world if capitalistic for-profit corporations, just as they disrupted the Big Telcos, and then AOL/MSN etc. So will blockchain be the value layer and IPFS/Autonomi be the storage layer etc. And the Web will be increasingly outdated.
End-to-end encryption can actually facilitate privacy. Blockchain is at best orthogonal to privacy, a public ledger generally undermines it.
As an aside, I don’t really get how a domain name connects to privacy. When I use encryption (including HTTPS) to communicate, the goal is to stay hidden and unknown. When I set up a domain name for my business, the goal is directly the opposite. Not to say there can be no reasons to advertise while remaining anonymous, just not sure privacy is a great parallel to draw.
Come on, more strawmen? First talking about shotguns, now privacy.
I didn't talk about facilitating privacy. You brought this up, in order to switch the subject to something end-to-end encryption can facilitate.
Blockchain isn't about privacy, it's about making sure that no one can control or man-in-the-middle-attack the network. People don't have to trust the middleman anymore.
I wasn’t talking about privacy at all. I was saying that your approach of “you have nothing to worry about if you arent doing anything wrong” is not very great, and it is exactly what is used by governments to fight against something you probably like more than blockchain and consider necessary, even though it can be used to facilitate terrorism. So you should take a look at whether you have double standards with regard to things you dont like vs things you like and think we need.
You keep saying "your website" but any successful website will be "our website". There'll be an organization, company, community behind it.
And now "the person(s) with the private keys" can rugpull it. Or do whatever they want.
Yes. Maybe A DAO could solve that. But that means everything, including domain names is in there from the get-go. Which isn't how this works on practice.
Blockchain technology is great for valuable assets owned by individuals. But much less so for groups and organizations that own valuable assets. And valuable domains almost exclusively fall under the latter.
The whole point is to NOT HAVE “a person with the private keys” to the entire database.
Each participant should be able to take only the actions as themselves, and affect a small part of the network. In aggregate they together effectuate the evolution of the network.
That is exactly the point — that we need blockchain software for entire communities rather than individuals!
How does that help the organisation (community) manage their single domain?
For example: who controls the intercoin.org domain? I'm quite sure it's a combination of trust and hierarchy and as fallback a society with laws and lawyers and law-enforcement.
Which, IMO is "good enough" for nearly all situations.
As I have already said, with the current DNS system, that’s how it works. You have to trust that your provider won’t screw you over. It’s held together by duct tape and spit.
You know, in every OTHER technology, that’s how it was before we automated things that humans previously did. You may as well have said this about telephone switchboard operators, or tying up the line, until VoIP brought the costs to zero. “Who connects your calls? I’m quite sure paying $1 a minute was good enough for nearly all situations.” Except, when it all got automated and the providers turned into dumb hubs because open protocols eliminated the middleman. Where are these phone providers today? They provide the infrastructure only, and we route around problems. Same with blockchain.
Paying someone exorbitant amounts to “maintain your domain” because it is famous, and a hosting company to “handle spikes in traffic” etc. All that results in the need to extract rents from the ecosystem in ever-more-toxic ways, using toxic forms of capitalism:
But it gets worse than that. The externalities to our society of the profit protive and private ownership of public forums are immense, including widespread depression, tribalism evho cham and national anger reaching a fever pitch. All predictable. Take a look at exactly how it works:
Again, Web3 and blockchain is one possible way to do it but the keys to all the solutions are decentralization and open protocols! They remove the middlemen and oligopolies (like phone companies used to be, or the original AOL/MSN walled gardens) by making an alternative system not owned by anybody and with no single points of failure or control by a few people:
You have still not answered the simple question of who, within this web3 org, controls the (.org, or even the .eth) domain.
This is also a single person. Even on ENS, whom you have to trust. By moving everything onto ENS, you've, practically, not solved the issue of who controls your domain.
Sure. It mightn't be the ICAN. Instead, now, it's Jeffrey who has the private keys on his ledger. Congrats!
That's like asking "who, within Bitcoin, controls the Bitcoin network".
Each person with a wallet and private keys contains neither more nor less than what they are entitled to control. Smart contracts manage collective decision making.
You should learn about "abstract accounts", i.e. smart contracts on the blockchain being the owners of different things, and acting on behalf of multiple people. This is far more secure, there is no one private key that can leak, but rather the smart contract has business logic, that everyone knows what the rules are. It's like a constitution of a country.
If you wanted to point the DNS to another IP address, for instance, you could have a rule that requires a proposal to be made, and for people to have a chance to vote on it during a voting period, with vote weights being equal or proportional to how much of a token people hold. Votes could be delegated. I wrote about all this on CoinDesk in 2020: https://www.coindesk.com/tech/2020/03/12/in-defense-of-block...
In fact, as more things become decentralized, the need to host a website at a particular IP address will go away, too. All of these Web 2.0 things are too centralized and prone to be rugpulled and changed, and the idea that someone must pay all the hosting costs is stupid, when even in 2004 BitTorrent participants also had to "seed" the same files they were "leeching". The reliability is actually necessary to the MEMBERS of the community who use it every day (exactly who you're talking about), rather than the LEADERS.
You can host static web sites on IPFS, for instance, and use smart contracts on the blockchain for business logic. That's what happens with NFTs, for instance, but that's just a first-generation technology, like the games Space Invaders and Pong.
For the existing models is beam-search like methods hopeless due to combinatorial explosion? Are there no smart ways to improve it? Evaluating multiple futures will be slow but if it means that the model can give vastly better output, it might be a worthwhile trade-off in some cases. I feel like our standard way of sampling the output of the LLMs is a bit too simplistic and my hunch is that it should be possible to get a lot more out of them even if it means losing speed.
People are considering that sort of beam-search approach - this is what they call "tree of thoughts" - generate a branching tree of alternate continuations, then pick the best one based on some criteria.
This doesn't seem an ideal approach though, since it amounts to generating a bunch of shallow responses and picking the best, rather than the preferred thinking more deeply before generating. It's not the same as a computer chess program considering N-moves ahead where you are guaranteed that one of those move sequences really is the best one (as long as you don't accidentally prune it out). In contrast, if you generate all possible "shallow" N-token responses (bunch of monkeys gibbering), there is no guarantee any of those will be the high quality response you are hoping for.
Really planning ahead - reasoning deeply before speaking - would seem harder to implement though, since it'd involve applying a variable number of reasoning steps (maybe looping), then determining when to stop. This also seems different from the proposed insertion of "reasoning tokens" since those are shallow reasoning steps (normal single pass through transformer's layers), when it seems what is really needed is more depth of reasoning ("more layers"), perhaps coupled with some working memory/tokens. Both schemes (more tokens vs more depth) are also related to the wish to use a variable amount of compute for different tasks/inputs - less compute for simple tasks, more for hard ones.
Ah yes, I totally agree. I was inspecting the method as a stopgap solution (especially because it does not require retraining or any other special tricks) until researchers figure out "planning" in a broader sense. It is very inefficient otherwise, but in the meantime, is just simple sampling with a couple parameters to tune from the output softmax the best we can do? is there no low hanging fruit there?
I suppose the closest alternative to planning ahead (considering alternatives before taking any action - in this case generating tokens) is getting it right the first time, which is only really possible in cases of highly constrained circumstances (prompts) where the model saw enough similar examples to predict the same correct/preferred response. So, to that extent, I suppose better prediction - bigger model, more/better training, etc, reduces the need for planning a bit. Architectural changes, such as adding working memory, that boost predictive power, would also help.
But, yeah, hard to see too many alternatives.
1) Get it right first time (not always possible)
2) Don't plan, but at least consider a bunch of poor alternatives - tree of thoughts
