Maia does this reasonably well! You can play against it on Lichess. I have gotten a few "feels like a human" moments when playing against it - for example, getting it to fall into a trap that could trick a human but would easily be seen by a traditional search algorithm. It's not adjustable but there are a few different versions with different ratings (although it's not a very wide range).
Piggy-backing off this - does anyone know of a quick way to evaluate the maia weights from python or js for a single board state? I'm trying to hack something together with my own search func intended for human play and I can't quite figure it out from the cpp in Lc0.
It is a bit roundabout, since it involves converting maia models to onnx before loading into pytorch and some outdated versions of libraries (maia/lc0 are a little old). We were using this for transfer learning for a competition, so we needed some flexibility that we didn't know how to do quickly/easily in TF.
Hope this helps.
------------------
Personal note: given your interest in chess ai and your starcraft username, I think we would have a lot of shared interests. Feel free to reach out (info is in my profile).
Not to be a party pooper, but posting detailed financial analysis of the exact sales data of a multi-million dollar business using numbers obtained through an obviously overlooked backdoor seems like a very bad idea. Haven't people have gone to jail for less? (iirc "but it was an insecure API" has not held up in court in the past)
On a more positive note, I've used a QR menu recently and it really is a game changer. Scanned a code, pressed a few buttons, and my food was there in minutes! Looking forward to seeing it more often, especially in places where you're not looking for stellar service.
I'm interested to know what the correct way to report this would have been? Specifically in this case. And what would one expect after reporting it? I've found many things like this and I only reported two (Genius, they said thanks) and Amazon (they replied but ultimately ignored it, and the issue is still there today)
First thing I would do is look for a security.txt file or search to see if they operate some kind of bug bounty. Failing that, I would browse their website or search for contact details (or even just a contact form). WHOIS can be useful for this. Ideally you'd want some kind of security contact, or a technical contact, but other times you have to make do with the general contact email/form.
In this specific case, they have a general email address at the bottom of their privacy policy, so that's what I'd use.
I'd send them an email along the lines of "I found a security issue with your website; how would you like me to report it to you?". Then they'll hopefully put me in touch with the right person.
In terms of what I'd expect… If they operate a bug bounty (which they don't in this case) then I'd expect what's on offer. If not, it would depend. I often don't expect anything. There have been businesses I've disclosed security vulnerabilities to that are shady enough that I've refused the reward they offered. Sometimes I don't want anything to do with them.
> Looking forward to seeing it more often, especially in places where you're not looking for stellar service.
I loathe them perhaps even more than I loathe the order-kiosks that McDonald's has rolled out. My phone is smaller than the folded napkin, I would rather not have to scroll to examine a menu.
Regardless, a restaurant should think twice about outsourcing this kind of thing to a 3rd party that now has all of your (and your competitors) financials. Even if the API is better vetted, why would you trust this faceless, profit-motivated site with your data?
"Convenience" seems to be the way they market "getting rid of employees" these days — from self-service gas, self-checkout lanes, etc.
That sign in particular says "Martial Arts Hall" ("budokan" / 武道館) from right to left.
Imagine an alternate universe where English could also be read left-to-right or right-to-left. If you were to see a sign saying "Hall Arts Martial", you'd immediately know the right way to read it.
Recently when I buy books on Amazon, they're often cheap versions printed in India. Sometimes there's even a "Only for sale in the Indian subcontinent" label on it. The quality is much worse - flimsy paper, faded ink, rough edges, and often quite dirty despite being new. Does anyone know of a more reliable place to buy new books?
Silurian hypothesis is specific to Earth’s geologic record which we’ve sampled all over the world millions of times. It’s less ridiculous to posit the same happening on Mars where we’ve only sent awkward robotic explorers. We haven’t explored any of the lava tubes or done ground penetrating radar studies or cut thousands of core samples all over the planet so we really have no idea what’s buried there like we do on Earth.
It's always fun to mention that many of the foundational ideas in the modern wave of machine learning & neural networks stem from work done in the 50s and 60s: perceptrons, backpropagation, stochastic approximation, etc. were all explored in depth back then.
It was only after compute power scaled up enough to apply these techniques practically that they became revolutionary. Really makes you wonder what things people are working on right now that will also need to wait 30 years.
"We fired the CEO for being too profit-driven" is a terrible message to send to your employees when you've lured them in with $1m comp packages that mostly consist of "Profit Participation Units" that are only worth anything when you actually make money.
Somewhat hidden beneath the huge headline of Altman being kicked out is that Brockman (chairman) is also out. Which could indicate something more systemically wrong than just a typical "CEO did something bad" situation.
> As a part of this transition, Greg Brockman will be stepping down as chairman of the board and will remain in his role at the company, reporting to the CEO.
Remember that Greg Brockman is a co-founder of OpenAI, and like Sam Altman, he is a main driving force behind the scene. Now both are gone. There must be something really really seriously wrong.
A coup wouldn't have him immediately fired. Instead he'd be placed in some advisory position while they transition in a new CEO. The immediate firing means scandal of some sort.
Brockman is off the board but not fired. Which is weird right? You'd think if he was involved in whatever the really bad thing is then he would be fired.
No, that sort of thing isn't that weird, in relatively young companies. Think of when Eric Schmidt was CEO of Google. Larry Page and Sergei Brin reported to him as employees of Google, and he (as CEO of Google) reported to himself-and-also-them (as the board), and all of them (as the board) reported to Larry and Sergei (as majority owners).
For another example, imagine if OpenAI had never been a non-profit, and look at the board yesterday. You'd have had Ilya reporting to Sam (as employees), while Sam reports to Ilya (with Ilya as one member of the board, and probably a major stakeholder).
Now, when it gets hostile, those loops might get pretty weird. When things get hostile, you maybe modify reporting structures so the loops go away, so that people can maintain sane boundaries and still get work done (or gracefully exit, who knows).
Turns out, there's no such thing as an LLM, it's all been a hustle with a low-paid army of writers in Kenya that Sama and gdb have been giving iv meth to.
So, at ~$23 a day, basically the same price as a capsule hotel in Tokyo with pretty much the same features? Not a bad deal given the differences in average rent between Tokyo and SF.
you and most of you who think this is a good thing are gonna be changing their tune the second the dude in the bunk above you or next to you starts jacking off.
> If you downloaded a book from that website, you would be sued and found guilty of infringement.
How often does this actually happen? You might get handed an infringement notice, and your ISP might terminate your service if you're really egregious about it, but I haven't ever heard of someone actually being sued for downloading something.
Whether or not it's enforced, it's illegal and copyright holders are within their rights to sue you. This is piratebay levels of piracy but because it's done by a large company and is sufficiently obfuscated behind tech, people don't see it the same way.
In Germany if you torrent stuff (without a VPN), you're very likely to get a letter from a law firm on behalf of the copyright holders saying that they'll sue you unless you pay them a nice flat fee of around 1000 Euro.
It's no idle threat, and they will win if it goes to court.
That's because, when torrenting, you're typically also seeding a copy of it, i.e. you're distributing your local copy to other devices, and thus you're directly aiding in piracy. Simply downloading content from a centralized server, as explained above, is different.
Although, one could argue what OpenAI & Meta are doing is closer to the torrent definition than the "simply downloading" definition, given that they're using that to redistribute information to others. It'll be an interesting case.
Honestly don't think our current laws are even good for this case.
This clearly needs some sort of regulation or policy.
It's clearly pretty bullshit if you ask chatgpt for a joke and it repeats a Sarah Silverman joke to you, while they charge you a subscription for it and she gets none of that sub money.
If books aren't under copyright protection and they're entirely legal to download, I agree that this lawsuit has no merit.
If that's not what you're saying, I don't understand your point. Is it the difference between the phrases "would be" and "could be," or even "should be"?
Exactly, never happens. It's a threat parents and teachers tell school children to try to spook them from pirating but it isn't financially worth it for an author or publishing company to try to sue an individual over some books or music downloads. The only cases are business to business over mass downloads where it could make financial sense to pay for lawyers to sue.
Actually no- downloading copyright infringing material is legal as far as I can tell but uploading it isn’t. The illegal part of torrenting copyrighted material is the uploading that the protocol requires you to do. Your ISP will send you an infringement notice because they want you to stop doing illegal things on their network
He hacked into a server to release a database of paywalled studies to the public. Not only is it not the same but it was the hacking that brought charges upon him.
It's been quite a few years, but AFAIK he didn't hack JSTOR. He downloaded papers en mass using a guest MIT account that had legal access to JSTOR. Maybe he violated their terms, but that is not illegal. He did illegally trespass an unlocked MIT switch closet to do this. They blocked several IPs but his script would rotate to continue. The downloading was over a week or two, enough for security to set up a camera in the closet to catch him retrieving the laptop.
I believe JSTOR sued him to prevent him from releasing the downloaded materials, worried he had offloaded the papers separately from the laptop. The final blow was an outrageous set of charges by the federal government. I also recall several prominent leaders in the open source movement calling it out for what it was, a power trip to make an example of a "digital terrorist". Such a shame.
This just all depends on what circle you're using the word "hacking" in. While technical circles are mostly concerned about the technical details on how someone might exploit a system, legal circles don't really care. They usually avoid using the word "hacking" anyway.
The relevant law here, the CFAA, is often referred to as the US law that criminalizes "hacking", but what it specifically does is criminalize anyone who "intentionally accesses a computer without authorization or exceeds authorized access" which is much more broad than how technical disciplines might use the word.
So yes, stealing a password off a friend's post-it note and Hasselhoffing their instagram might not be considered "hacking" if you're hanging out at Defcon, this would be considered "hacking" in legal or colloquial terms.
> AFAIK he didn't hack JSTOR. He downloaded papers en mass using a guest MIT account that had legal access to JSTOR.
Potato, potahto. Or, like kids these days say it, "corporate wants you to find differences between these two pictures...".
Fact is, from the POV of the legal system, "using a guest account that had legal access to" a system, but to which (the account) you didn't have legal access, would typically be seen as hacking. So is running curl in a loop, if it results in you getting sued for it. So is just guessing the URL (e.g. incrementing a user ID in a GET query param), if it lets you access things you shouldn't be able to.
Yes, it's not aligned with how technology works. But it is aligned with expectations of behavior, which is what the law is really about.
> Fact is, from the POV of the legal system, "using a guest account that had legal access to" a system, but to which (the account) you didn't have legal access
I don't think that is accurate.
He had legal access to the account. The account had legal access to the service.
The argument was that downloading articles en masse was an _abuse_ of the service, which was a violation of the Terms of Service and therefore a CRIMINAL ACT.
https://www.maiachess.com/
https://lichess.org/@/maia1