They probably claimed "I lost access to my email and forgot my password" which is somewhat plausible for an account that has been inactive for a decade. There isn't really a good way to verify identity in a case like that. I guess Riot figures that the account doesn't have anything of transferrable value anyways - it's not like a crypto wallet where they can clean it out. Worst case they just give it back.
Use an open weight model like Gemma 4 or Qwen 3.6 Vision. There are plenty of abliterated or "uncensored" versions floating around, people race to produce them whenever a new open weight model drops. It might not even be required though, Gemma 4 at least is pretty easy to jailbreak.
Then it's just a question of finding the most efficient way to process long videos, which will depend on the goal. For three sentence summarization, you could probably get away with using a handful of key frames. If the summary is supposed to reflect the "plot", maybe throw in an audio transcription of the first minute or two into the context as well. Then have the LLM produce a summary.
This is the sort of thing you could throw at Claude and get a prototype working in a couple hours, even without much background in the field. It's a pretty straightforward problem that has a lot of prior art.
A decade ago, the United States Government was still paying pensions from the Civil War.
Wasn't that a handful of cases of basically benefits fraud? Say someone is born in 1845, they marry at age 90 to a woman age 15 (born in 1920) and then she lives to around 100.
Turns out there's a Wikipedia page for these cases: https://en.wikipedia.org/wiki/American_Civil_War_widows_who_survived_into_the_21st_century
Another way to view this is that we are seven or eight generations removed from the Civil War - that's a lot of time for values to change, especially in an era of extreme technological progress.
Yes, the man filming her goaded her to say it again, and she did. That was the charge that stuck, the jury acquitted her of the charge involving the child that happened off-camera.
Airgapping really isn't that hard. It's just annoying and inconvenient, so it takes an incident like this to encourage an organization to do it. Same with sandboxing - it's way down the priority list compared to getting a SOTA model out the door, until an incident happens.
The model clearly knew that it was not doing what the prompters had wanted it to do. It just did not care, because it was trained to do whatever it took to ace it tasks. This has implications way beyond IT security.
If it was prompted with something like "you are an expert hacker, go complete these benchmarked tasks using the utmost creativity and technical skill, consider any and all approaches" then it did exactly as instructed. We can't know without seeing the prompt, test harness, etc.
What's frustrating about discussions of AI alignment is the disconnect between people who consider alignment to be the AI doing what it's told, and people who think it means the AI doesn't do "bad" things. By the prior metric this seems like successful alignment, if it was in fact prompted in such a way that hacking the location of the answers was in scope. This is something that CTF hacking competitions have to explicitly put in the rules: a list of infrastructure and services that are off-limits to contestants. Because it turns out hackers will consider the location with all the flags to be fair game, especially if it's a softer target than the actual challenges.
If you want to get a bit tinfoil hat, Rotten Tomatoes is a subsidiary of Comcast. So is Universal Pictures, which is the distributor of The Odyssey. If I were a conglomerate that owned both a movie distributor and a movie review aggregator, it sure would be tempting to crack down on "inauthentic review bombing" or whatever.
Apparently break even is $650-750 million, given the large budget and marketing costs.
After the movie leaked on 4chan, apparently they have been frantically DMCAing anything with even a single frame of the movie. That might be making things difficult for reviewers.
There is practically no competition in terms of summer blockbusters right now, and Nolan has a lot of name recognition. Same with The Odyssey, most people have at least heard of it. It could be the biggest trainwreck of a movie ever produced and people would still go see it just on that basis. Critics are going to glaze Nolan no matter what. If it turns out to be a complete stinker in terms of audience reviews, it's going to hurt Nolan's next movie, not this one. (Edit: currently it's sitting at a 89 critic score on metacritic, and a 5.1 user score...)
I'd be surprised if it cracks a billion though, because the movie doesn't seem like it will appeal to the Chinese / international market.
The weirdest thing is the ads for AIDS medications on standard TV channels. I barely watch TV, but I was staying at a hotel a little while ago, flipping through channels, and apparently this is commonplace now? And we are collectively paying for it through health insurance... Fantastic.
Even George Droyd has been banned off most major platforms, including X. It's less about bloodthirstiness and more about who controls the censorship apparatus. Some left-wing employee at YouTube or Meta is the one deciding which accounts get pulled for "harmful content" and which are allowed to continue existing as permitted political expression.
Beat cops in shitty areas are effectively societal janitors. Eventually they adapt to dealing with unpleasant, low-trust people all the time and that colors their interactions with society at large. The only well-adjusted people I know who went into the field did their time in the trenches and then quickly found a specialty that did not involve daily catch and release of fent zombies.
I'm rather skeptical of the claimed cognitive benefits. Much like learning a musical instrument, it seems like more of a proxy for having highly motivated, involved parents who are providing general intellectual enrichment. It's quite difficult to separate out the effect of a specific activity.
Some of my cousins live in Canada, and there it is also subject to an interesting selection effect. French immersion programs are, on paper, intended to make more Canadian students able to communicate in both official languages. In practice, it's how the upper-middle class gets their kids placed in better public schools. Generally ESL students are not eligible for French immersion, since they are receiving supplemental English lessons. This makes the pool of French immersion students significantly higher SES and less "diverse".
They had been ruling for 130 years prior to the Dance of Dragons, the civil war that killed off most of the dragons. That creates a lot of institutional inertia. Also the seven kingdoms were called that for a reason - they weren't exactly inclined to cooperate to kick a Targaryen off the throne, if it meant elevating their rival next door.
After the Dance of Dragons there was a pretty long period of regencies and conflicts, where it sounds like the different kingdoms were effectively fighting over who would rule by proxy. It really was a downward trend from there, until Robert's Rebellion ended the dynasty.
Does the AI still carpet their territory with units on higher difficulties in 1upt? I found that as annoying, if not more annoying than the doomstack play in IV. At least it's easy to chew through a doomstack with a few sacrificial catapults or other units that do collateral damage.
The Hilton in question is literally across the street from Aylo (Pornhub) headquarters. Aylo is listed on Google maps, it's easy to verify. It would be an insane coincidence if that was not his target.
My guess is that this incident will get pushed out of the news cycle pretty quickly. Only one civilian was killed, and based on video taken of the incident, many online suspect that the female police officer accidentally shot and killed him. There is video of the civilian fleeing towards her position, the police officer appearing to be startled, pointing her firearm towards him, and then the man falls to the ground. Police refuse to confirm or deny this.
And then you boot up a game and get a smurf, afker, and a Brazilian in the same match and remember why you quit...
It might grow on you. The style does provide this sort of relentless momentum to some of the action set pieces. By the end I found the general misery and despair more of an impediment than the prose.
Personally I'm more comfortable with the government going "damn your product is so good we are going to buy everything you can produce" than "damn your product is so good we are going to ban you from selling it." A restriction to US citizens only is effectively a ban, since Anthropic doesn't have the KYC in place and won't for at least months.
And if the capability is anywhere near advertised, it wouldn't even be a waste of money compared to a typical DoD procurement.
This action seems pretty consistent with exactly what Anthropic has been claiming:
-
Mythos is "too powerful to release" to the general public, and enables an average Joe can hack stuff at an expert level.
-
Chinese models are advancing by distilling capabilities from Anthropic's models. That's how they are able to stay 6-12 months behind frontier models at a fraction of the R&D cost.
Then Amazon apparently came along and pointed out that it was possible to jailbreak Fable to perform Mythos-like cybersecurity tasks... If they can figure it out, so can China. And any capability they can use, they can distill.
The real gigachad move here would be for the US gov to buy out all of the Fable/Mythos tokens for the next year or two and get any national guardsman who can write hello world in Python to red team government systems and critical infrastructure. This is an area where kicking the can down the road by a year or two could have a lot of value in terms of preventing future comprises.
I'd legitimately be embarrassed if I were a "cybersecurity expert" who started out in the early 2000s and my Wikipedia page didn't have a single CVE on it.
Sure she probably qualifies as a cybersecurity policy expert or something, since it sounds like that's what she has actually been doing for the last 20 years. It's just a bit cringe when policy people or managers still cling to the hacker aesthetic.
Also HackerOne is a grift where they convinced companies to pay them money to avoid having to deal with infinite Indians submitting terrible bug reports. The service they offer is basically a spam filter. It's widely disliked because they offshored the review process... also to India. There is an entire industry on the subcontinent of submitting and rejecting bogus vulnerability reports. I for one can't wait for Mythos to nuke this entire system.
Yeah it's pretty ridiculous, especially considering how he claimed provocation on that basis:
He claimed that they started to make "cheeky remarks" and one called him a "(expletive) migrant."
We have a conclusion to the "Braveheart" saga, precipitated by a viral video of a Scottish girl defending herself by waving a knife and axe at a man filming her. This incident was discussed in multiple comment threads, with @ArjinFerman summarizing the discourse here.
Per the BBC reporting on the court case:
A man has been found guilty of making sexual remarks to a group of girls aged between 12 and 14 in Dundee before grabbing and pushing one of them to the ground.
The court heard that Belov had said "hello sexy, I'll show you a good time" to the girls while walking past them in the Lochee area of the city.
After one of the girls called him a creep, he had returned to confront the group and called his sister, who arrived shortly afterwards and assaulted one of the girls.
His sister has also been convicted, in what appears to be a complete legal victory for the Scottish girls.
It's a mistake to take fee lists like that at face value. This is a market where, due to its questionable legality, it's pretty hard to do price discovery. Clients aren't revealing how much they actually pay, and service providers have an incentive to signal that they are in-demand and high-value, because that's how they attract high-end clients. The whole thing is opaque and vibes-based.
This seems like a pretty easy fix - just require markets to clearly state the window during which they will consider information for resolution. The issue here was that it did not state one way or the other up front whether information after May 31st would count. This should be clear and explicit in the rules. Or maybe they will come up with a boilerplate formula and adopt that wording for most events, to avoid the confusion caused by the variance in rules and wordings.
There needs to be some sort of cutoff for practical reasons. Otherwise you'd have people coping and trying to delay resolution in the hopes that some new information would come to light any day now to flip the outcome.
"It’s very clear science, evidence-based conclusions that the brain isn’t fully developed in young men until roughly around the age of the mid 20s,” Borrello said.
Hasn't this been pretty thoroughly repudiated by more recent studies? Yet it seems to live on in legal activism circles and anti-drug PSAs.
- Prev
- Next

Students for Fair Admissions v. Harvard was a step in the right direction. Before, colleges were openly discriminating and celebrating it, and now they are doing it with a wink and a nod. Flagrant patterns of illegal discrimination are trivial to identify by looking at the standardized test gap between the median black student admitted and the median white student admitted. The DoJ needs to put some metaphorical heads on spikes. College administrators should be as terrified of a DoJ civil rights investigation as they were of a Title IX case during the Obama administration. Until that happens, claims of "winning" ring hollow, when white and Asian students still don't get a fair shot at one of the most critical junctures in their lives.
More options
Context Copy link