Rank: #10 out of 4859 in peer accuracy at Metaculus for the time period of 2016-2020.
ChristianKl
Karma: 36,259
This document basically says that Anthropic gives the military exceptions where they can use Claude in ways that violate the standard ToS. Then it gives one example of those exceptions.
When defending against employees and external parties having concerns about the deal with the military, this allowed Anthropic to point at the one exception while ignoring the fact that the policy allows for making other nonpublic exceptions that are made in a classified setting.
What is the basis of that claim?
Anthropic was not publicly assuring people that they had enforcement mechanism that prevented their software from being used in ways that their employees didn’t like. Especially, if you care about alignment, thinking about working mechanism would have been important.
I think it’s quite poor for the alignment community to have let Anthropic get away with that at a time.
I’d be interested to hear your opinion on the degree to which this is a move toward soft nationalization — my sense is that you at least partially disagree?
This news story is a sign of friction between Anthropic and the government which is a bit the opposite of nationalizing Anthropic. I think when the deal between the military and Anthropic was first made and the expectations document was published that the most likely future would be one where the military would sooner or later to whatever it wants with the software.
Given that the Trump administration is accused of doing plenty that isn’t exactly “lawful”, calling for deals that allow all lawful usage is not the maximum demand that Hegseth could make.
I’m just being snarky on that one, I figured I could safely bury the snark in a footnote to a footnote but I underestimated the LW readership.
When Anthropic made the deal, there were plenty of people who thought that the military wouldn’t do things outside of the agreement and this is why Anthropic was okay to make the deal.
Unless I’m badly mistaken, these are limitations from Anthropic’s standard terms of service, not something recently introduced.
When Anthropic did their deal with the military they granted that there are explicit expectations where the military can violate the terms of service and Anthropic can give the military additional expectations that are given in classified documents so that the Anthropic can tell no one including most of their employees who don’t have security clearances about the new expectations.
The deal also did not contain the right of Anthropic to have any knowledge about how it’s software will be used and no enforcement mechanism.
The deal probably looked to the military like it was structured to allow the military to do what it wants while Anthropic can pretend to it’s employees without security clearance and the public that they do things to limit the usage which is fine for the military.
Pete Hegseth dislikes don’t ask don’t tell arrangements. He’s the secretary of war and not the secretary of defense.
There’s also the chance that the military was asking Claude to do something and Claude refused. Then when the military asked Anthropic to make Claude comply, Anthropic refused so Hegseth made up the story about the Anthropic asking about usage in the Maduro raid.
WSJ: ‘Some administration officials were frustrated that the company was dictating how its technology could be used’ have they literally not encountered terms of service before lol
They probably have but are in the habit of just ignoring terms of service when it comes to classified matters.
I feel like your post focuses too much on the present memes about what people in the 21st century believe about medieval believes about witches instead of what people back then believed. The Survey of Scottish Witchcraft for example provides a lot of information about witch trials and from them I don’t get the impression that it was about a very narrow class of people.
Asking “Why do companies care about higher education?” predisposes that companies are the agent under concern. Decisions about hiring decisions are however made by humans, managers and HR personal. Managers that have a college degree have a strong incentive to steer their company into valuing college degrees. Managers who went to Ivy league universities have a strong incentive to steer company policy into rewarding people with Ivy league degrees.
How much do you think the skill you used is the basic superforcaster skill? Did you do Metaculus or GJOpen and think people with similar forcasting skills are likely also going to be good at investing or do you think you had different skill that go beyond that?
That’s great regarding the allergy concern.
When it comes to messaging a key question seems to be about whether to present the goal as increasing safety or about presenting it as reducing vaccine hesitancy. If you present it as something that’s about increasing safety that might help to get support from the MAHA crowd but at the same time it can alienate mainstream researchers.
One idea that comes to my mind that you could do without alienating too many people would be to run a petition with Jay Bhattacharya as the target to change NIH policy on self experimentation. The administration wants to the amount of science that gets prevented by Institutional Review Boards, so it would be good fit into what they want to do anyway and might be a relatively easy win that might expose the project in a favorable light to the people at the top of the current HHS.
We are talking about a community that has community evaluation of technology and only uses technology when the community evaluates the tech to be good for the social fabric of the community.
I’m not sure whether “out of conviction” is a very good category. https://globalnews.ca/news/11214661/ont-measles-mennonite/ does suggest that historic distrust in authorities is a key factor.It’s not that they distrust mainstream medicine in particular but that general distrust of authorities leads to distrust of vaccination as well.
While it that case it’s more historical distrust, it’s probably similar to the distrust you get prevent churches from gathering during lockdowns because of bad public health policy and afterwards discover that the congregation doesn’t trust you anymore on vaccines.
As far as the page goes, I feel like it’s a worse than the old one. While it’s a bit more beautiful it’s a lot less functional. Trading functionality for beauty seems to go against the core way of LessWrong’s usual design that’s quite minimalistic. It would probably be best to revert to the old one.
In particular:
The new one doesn’t show the biography at the top but to the right bottom where it’s a lot less central.
It makes Sequences very hard to find, which reduces the agency of users to present their work by presenting.
If you are seeking a post by someone where you don’t know the exact title but have some idea about the karma that the post had, the new placement of the karma makes it harder to scan for the post. Maybe there should both be list like the old UI and an expanded option.
Having to click on the dial to see which sort mode is currently selected seems to add unnecessary friction.
Currently, the user statistics don’t seem to be shown at all if there’s a biography. That’s probably more of a bug than a design decision.
A lot of the goal of the redesign is actually to make it so that you have more control which content you want to represent you on LW. You can now select which posts you want at the top of your profile, and which one you want to give the most prominence, which you didn’t have previously.
Previously, you could group the post you consider to be thematically together into sequences that would be shown at the top of the page.
I think it’s natural that user first think that having Top posts which the hover text “based on karma”, actually shows the posts with the most karma even when it sometimes doesn’t if you configure that on the profile.
One way to communicate the fact that those are switchable would be to show a button for switching when hovering over them on your own profile. Maybe renaming “top posts” into featured posts would also be helpful while changing the tooltip to mention that it also shows user selected posts.
While it’s true that cancers today are more manageable if caught early plenty of what gets diagnosed as cancer disappears without intervention and gets very harmful surgical interventions like amputations if diagnosed. During the Obama administration they purposefully reduced cancer by policy testing to reduce the amount of false positives.
Colonoscopy is not a risk-less procedure. There’s no evidence that the advice of getting a colonoscopy if you are over 45 reduces all-cause-mortality. The biggest modern randomized colonoscopy trial (NordICC) enrolled ages 55–64 found no reduction in all-cause-mortality.
For health-conscious people “I do only things that are proven to help with the goal of not dying” is a perfectly reasonable way to relate to the topic of these kinds of tests.
I think it is, why are we comparing burglaries to digital crimes when the latter is likely far more common?
Because Meta shares a huge responsibility for making the digital crimes easy to do. According to their own analysts their platforms are third of involved in a third of all successful scams in the U.S.
This isn’t just about ads but also about other communication, but it should be Meta’s responsibility to provide an environment for their users that doesn’t make them prime targets for crime.
Digital crime proliferation is a sign of big tech failing customers by not adequately protecting them.
If you argue that the likely fine you have to pay is lower then the profit you are making and thus you don’t need to engage in strong measures to reduce fraud, I do see that as sign of intent.
When Meta shows it’s users an ad that it believes to with 90% probability from a scammer, it should at least tell the user about the ad likely being a scam. Withholding that information when especially older users probably thinks that Meta goes through some effort about not just presenting the user with scams seems clearly intentional as it would be easy to show the user a warning that Meta thinks that the ad is more likely than not a scam.
The article you links essentially argues a strawman:
Most people named in the Epstein files are not being prosecuted for the simple reason that what appears there does not meet anything like the legal standards required for prosecution, let alone conviction. Being mentioned in an email, a contact list, or a flight log may be morally damning and emotionally enraging, but it’s not evidence of a crime in the way the criminal justice system is actually supposed to require.
You wouldn’t persecute people who are just mentioned in an email, contact list or flight log. You would persecute those accused by the women who told the police they are victims. According to the lawyer for the victims there are at least twenty men against whom victims gave testimony.
Given victim testimony and the files we have I think it’s also would make sense to say that Epstein was running a criminal enterprise that’s subject to the RICO act. That means that plenty of employees like the pilot that was trafficking the girls to the island likely committed crimes. If you start you RICO proceedings you can offer lower level employees immunity for providing more evidence.
What scenario are you imagining where the pizza joint facilities fraud and people get scammed because of the pizza joint that otherwise wouldn’t be reached by the scam?
If you have some mafia element that runs a casino as their main profit source that technically legal and makes 90% of their profit through crime, would you say that’s a criminal enterprise?
The pizza joint does not provide a way to target specific demographics who are likely vulnerable to scam attempts and run complex machine learning algorithms that optimize the effectiveness of the scam and auction of the marks to different scammers.
Comparing the money made by meta to the amount of value stolen via burglaries is not a vibe based argument.
Some bigger spenders – known as “High Value Accounts” – could accrue more than 500 strikes without Meta shutting them down, other documents say.
Fraudulent ad campaigns can reach massive size: Four removed by Meta earlier this year were responsible for $67 million in monthly advertising revenue, a document reviewed by Reuters shows.
To draw attention to the company’s perceived failures, an employee earlier this year began issuing reports highlighting that week’s “Scammiest Scammer.” The report profiled whichever advertiser had earned the most user complaints about scams in the past week.
Colleagues praised the initiative. But being name-checked in the report wasn’t always enough for such accounts to get shut down.
The right action for ads that are more likely than not fraudulent is to put them in a queue to be reviewed by human moderators and probably tell the police about fraud attempts that human moderators consider to be relatively certain to be fraud.
When ten percent of their revenue is facilitating fraud than getting rid of those ten percent of their revenue hits “specific revenue guardrails” even when it doesn’t impact legitimate users at all. It’s quite obvious that removing 25% of the profits would result in some revenue guardrails being violated.
Increasing the price for fraudulent ads is a way to keep revenue high while reducing the amount of fraud.
The company’s internal estimate that it would earn 10.1% of its 2024 revenue from scams and other prohibited ads was “rough and overly-inclusive,” Stone said. [...] Is this reasonable or a greedy corporation trying to cover its ass?
If Meta would believe that the correct number is substantially lower in a way that would motivate people to be less angry, they would probably have shared with it, so what we take from that statement is that Meta believes that the correct number is so high that it’s embarrassing to them.
Generally, do you believe that if corporate accountants have the job to estimate a number that’s bad for the corporation and could possibly surface in lawsuits or government investigations are they more likely to over- or underestimate it?
I would also note that the corporate statement includes any sign of Meta investing resources. It does not say things like “Because we care about our users not getting scammed we spent 100 million on investigators to remove fraud from our platform.”
For Elon motivating SpaceX employees is important, so he needs to tell the public a story about how SpaceX isn’t just about building AI datacenters even if he thinks building superintelligence is the main goal of SpaceX and it’s important to build AI datacenters for that and he can’t build enough of them on earth.
As far as the AI datacenters in orbit go, it’s not clear to me that the thermodynamics work out for that project and it’s easy enough to lose the heat that the datacenters produce.
I think the idea is that >5k karma users have karma to lose to punish them for posting low-quality content and it’s better to have humans make the judgement about what’s low-quality than AI.
Murder is about intent. I think Dario believes that his actions reduce the chance of human extinction due to AI because Anthropic is doing a better job then competitors.
When it comes to Sam Altman, I don’t think he believes that OpenAI is likely going to kill humanity.
Facebook on the other hand, is intentionally and knowingly facilitating fraud because they think that the government is unlikely to punish them for it and try to make as much money as they think they can get away with without the government punishing them.
If it looks like the van in the picture, the antennas might be needed for surveillance functions. If you have visible antennas claiming it to be a normal police van or “Robinson’s Garden Equipment” would not do the job.