RedMan

Karma: 434

RedMan Apr 16, 2025, 3:18 AM
7 points
4
on: Frontier AI Models Still Fail at Basic Physical Tasks: A Manufacturing Case Study
I think tacit knowledge is severely underrated in discussions of AGI and ASI.
In HPMOR, there is a scene near the end of the book where our hero wins the day by using some magic that would be equivalent to flying a drone around an extremely complicated path involving lots of loops in places not directly observable for our hero.
Our hero has never once in the book practiced doing this.
In theory, if I possess a drone and have a flight path the drone is capable of flying, I can pick up the controller for the first time and make it happen.
In practice, I will fail spectacularly. A lot of writing in this space assumes that with sufficient ‘thinking power’, success on the first attempt is assured.

RedMan Apr 15, 2025, 4:29 PM
2 points
0
in reply to: samuelshadrach’s comment on: Joseph Miller’s Shortform
1. This has been an option for decades, a fully capable LLM does not meaningfully lower the threshold for this. It’s already too easy.
2. This has been an option since the 1950s. Any national medical system is capable of doing this, Project Coast could be reproduced by nearly any nation state.
I’m not saying it isn’t a problem, I’m just saying that the LLMs don’t make it worse.
I have yet to find a commercial LLM that I can’t make tell me how to build a working improvised explosive (I can grade the LLMs performance because I’ve worked with the USG on the issue and don’t need a LLM to make evil).

IOSCO—AI in Capital Markets: Use Cases, Risks, and Challenges

RedManApr 2, 2025, 3:27 AM

4 points

0 comments1 min readLW link

RedMan Mar 6, 2025, 3:48 AM
1 point
0
in reply to: David James’s comment on: The Milton Friedman Model of Policy Change
Now is dramatically better than a year ago. It’s not even comparable. Rewrite the cover sheet on your policy idea and ping your network.
The incoming leadership has a massive amount of flexibility, given that they’re fundamentally reshaping so many things at once, but in many cases just have vague ideas rather than specific programs. Give them specific proposals that they can align with their vague pronouncements.
Bureaucrats are finding themselves taking on responsibilities for people who were shifted out the door in a hurry, and have incoming leadership who need staff support badly. The survivors will likely have much more leeway than they did before to stop doing things they don’t want to do, and start doing things they do.
Private sector actors are confronted with government agencies that are in disarray and distracted. Now is a great time to take action.
Uncertainty creates a lot of anxiety, so if you’re generally afraid of your own shadow, you’ll turtle up and hope the storm passes. Given that so many others are doing exactly that, someone ambitious has an opportunity to shape reality around themselves to a degree which was absolutely not possible last year. This is a great time to get stuff done, as long as you’re razor focused on the specific things you actually want.
That being said, if you haven’t spent the last few years working on developing relationships with people in those groups, you might have a problem. They’re probably not talking to anyone they didn’t trust before all this chaos started.

RedMan Mar 5, 2025, 6:01 AM
1 point
−3
on: The Milton Friedman Model of Policy Change
The disarray within the executive branch right now has created an amazing window of opportunity. If you have a clear policy objective, you can probably find someone, somewhere to give you a fair hearing.
New officials looking to create radical departures from the previous admin’s policies are one route. Career bureaucrats who have survived the cuts, and find themselves suddenly empowered because their supervision did not survive the curs are another. And finally, actors within the private sector may discover that while the laws themselves have not changed, what is effectively enforced is likely to be different (some things more restrictive, some things much less).
Good luck!

RedMan Mar 1, 2025, 5:05 AM
1 point
0
on: Experiences raising children in shared housing
Sounds like it’s working well when you have a shared culture. The more you agree on norms of behavior in terms of what’s appropriate for kids, how and when to discipline how to speak to kids, etc the better it works. Religion probably helped with this historically.

RedMan Feb 23, 2025, 8:24 PM
1 point
0
in reply to: quetzal_rainbow’s comment on: Joseph Miller’s Shortform
I wrote about something similar previously: https://www.lesswrong.com/posts/Ek7M3xGAoXDdQkPZQ/terrorism-tylenol-and-dangerous-information#a58t3m6bsxDZTL8DG
I agree that 1-2 logs isn’t really in the category of xrisk. The longer the lead time on the evil plan (mixing chemicals, growing things, etc), the more time security forces have to identify and neutralize the threat. So all things being equal, it’s probably better that a would be terrorist spends a year planning a weird chemical thing that hurts 10s of people, vs someone just waking up one morning and deciding to run over 10s of people with a truck.
There’s a better chance of catching the first guy, and his plan is way more expensive in terms of time, money, access to capital like LLM time, etc. Sure someone could argue about pandemic potential, but lab origin is suspected for at least one influenza outbreak and a lot of people believe it about covid-19. Those weren’t terrorists.
I guess theoretically, there may be cyberweapons that qualify as wmd, but those will be because of the systems they interact with. It’s not the cyberweapon itself, it’s the nuclear reactor accepting commands that lead to core damage.

RedMan Feb 23, 2025, 3:50 PM
4 points
2
in reply to: quetzal_rainbow’s comment on: Joseph Miller’s Shortform
This seems incredibly reasonable, and in light of this, I’m not really sure why anyone should embrace ideas like making LLMs worse at biochemistry in the name of things like WMDP: https://www.lesswrong.com/posts/WspwSnB8HpkToxRPB/paper-ai-sandbagging-language-models-can-strategically-1
Biochem is hard enough that we need LLMs at full capacity pushing the field forward. Is it harmful to intentionally create models that are deliberately bad at this cutting edge and necessary science in order to maybe make it slightly more difficult for someone to reproduce cold war era weapons that were considered both expensive and useless at the time?
Do you think that crippling ‘wmd relevance’ of LLMs is doing harm, neutral, or good?

RedMan Feb 23, 2025, 5:25 AM
−4 points
−3
in reply to: quetzal_rainbow’s comment on: Joseph Miller’s Shortform
You sound really confident, can you elaborate on your direct lab experience with these weapons, as well as clearly define ‘military grade’ vs whatever the other thing was?
How does ‘chem/bio’ compare to high explosives in terms of difficulty and effect?

RedMan Feb 20, 2025, 5:38 AM
5 points
2
on: Cheap food causes cooperative ethics
Luxembourg and Belgium during WWI are case studies of this.
Humanitarian food aid was shipped to Belgium via the US led CRB, so labor unions were able to stop work, and Belgium mostly didn’t contribute to the war machine.
The 1916 miner’s strike in Luxembourg was broken by the Germans withholding food. No effort like the CRB to support resistance could be mounted by the US and Allies.
In terms of breaking the hold of tyrants on humanity, probably the most significant technical development would be a household or apartment building scale device that converts electricity, water, and air into protein; sugar is easy. Solein claims to have developed this (their microbe works at the ‘space station’ scale), but they’re opting to build large factories and take advantage of economies of scale instead of pushing out the equipment to the masses.

RedMan Feb 17, 2025, 3:11 AM
−2 points
−6
in reply to: Chris Monteiro’s comment on: Murder plots are infohazards
What is a reasonable response to an unsolicited message saying ‘someone has hired someone to harm you’, with scant details on who/what/when/where?
Personally, I’d read it, and the more seriously I took it, the less likely I would be to engage with the sender, I likely would not send an acknowledgement of receipt. Any additional communication from the sender, especially a message with an ‘ask’ like ‘help me figure out who’ or ‘assist me in making internet content’ would be viewed as extortion.
If someone showed up at my door, I might talk to them, and if they’re a cop, I’d tell the cop that I’m concerned about the vaguely threatening messages I’ve gotten from ‘some guy in another country’.
If you don’t have much information (‘just a social media name’), doing more ‘sleuthing’ is probably totally inappropriate. The ‘contractor’ may or may not have done more, but either way, you’re not really helping anything.
As far as law enforcement seeing you as a suspect/time waster/scammer, that’s an easy one. A cyber crime office will hear you confessing to some kind of ‘hacking’, other offices will hear that you have no actual details (so someone unknown, is threatening someone who is also unknown, no further details? Sure ok I’ll log it) and just assume you wasted their time. Calling back to ask about status as a ‘helpful tipster’ suggests that you’re motivated by something other than a desire to inform them, and let them prioritize your information. This would likely be seen as suspicious.
Most agencies have an escalation path and can do ‘international’ effectively, they just have to really want to do it. In this case, from what you’ve told us, I totally understand why you’ve gotten the response you’ve gotten.
An online murder for hire website, where the people hiring are mostly larpers, the people taking the jobs are scammers, and the victims are basically unidentified is at best a loot pinata for someone with asset forfeiture authorities. Assuming they can actually seize the crypto.
You might have better luck tracing crypto to exchanges, passing that information to national level cybercrime squads, and asking for 10% if they can seize anything (most agencies have policies like this).

RedMan Feb 14, 2025, 4:16 PM
21 points
3
on: Murder plots are infohazards
You’re not doing anything immoral by just ignoring your archive. Working on ‘spec’ in public safety usually just leads to sadness. You’re in a weird place where you’re facing the kind of problem a mid or late career professional sees, but without social support from people who have seen this stuff before, and without easing into it like you would have in an early career. So you have the problem, but none of the emotional tools to address it.
As far as the situation specifically, making a form letter saying something like ‘someone has paid x $ to do you harm through (description of the service), further details are available upon request’, and sending it to as many prospective victims as you can identify, is probably as much as you can reasonably be expected to do.
Sure, local law enforcement should probably do it, you won’t find everyone, and someone you do find might come back and cause you problems for “making threats”, so in your situation, I probably wouldn’t even bother going that far, but you seem pretty committed to doing “something”. So that should be an achievable end that lets you move on emotionally.
It’s hard to talk about analogous situations without saying things that are impolite. Feel free to PM me if you have questions or details you’re not comfortable sharing, I’ve worked with “bad stuff” before, and might be able to help.
“People ask me to tell them my worst stories. I used to tell them, but as it turns out, nobody really wants to hear them” is a common sentiment.

RedMan Feb 2, 2025, 5:16 PM
1 point
0
on: RedMan’s Shortform
Looking to check my reasoning here.
I’m currently doing a term employment for the Feds that ends in a little over a year, the work I’m doing is important (in my opinion), but team just hit a major milestone (LWers would likely approve of the outcome) that basically covers what I came into this position to accomplish.
I’ve got the fork in the road email from Elon sitting in my inbox, possibly in error (I’m not a permanent employee, I’m turning into a pumpkin in Q1 2026 no matter what I do).
I can do a lot with an eight month paid vacation in terms of upskilling and independent work.
I’m thinking I should take the offer, then apply to have it rescinded like a month before it ends (september) to see if I can pick up the few months of pay (obviously I’d have to actually work them) associated with the position that I’m losing. I can probably generate more value for myself during the time off than I would be losing by snipping the end off my term.
I’m thinking that ‘congress could wreck this’ during the shutdown is probably not something I should consider, because I’m at their mercy in that respect whether I fork or not.
So am I forking my resume, or telling Elon to get forked?
Some math for ‘which numbers do I plug into what formula to make this decision precisely’ from you brainiacs would be totally welcome!

RedMan Jan 25, 2025, 1:50 AM
1 point
0
on: Why and How to Graduate Early [U.S.]
I finished high school (with a community college class in my senior year) shortly before my sixteenth birthday and went to a local college on a full scholarship. I spent five years in college (I hated school, but loved jstor), I graduated and got my dream job in 2009.
The path to doing it is different for everyone, and local factors (school system) will be critical.
Professionally, I recently had the pleasure of being assigned to a team with two others who had also started college at sixteen. It was fun, and it felt nice to work with two other members of this weird club. We compared notes and concluded that our experiences were a bit odd, though overall good, and that we are happy with where we ended up
The ‘but what about social stuff’ concern is overblown, every teenager has to figure a lot out, and I don’t think there was a world where I wasn’t awkward, grade skipping or not.
A few years ago I saw that the longitudinal studies on grade skippers are out, kids who can skip generally reach ‘average success in their chosen field’ kids who could have, but do not, have a lower chance of finishing high school at all, and usually have worse outcomes on all measures, including social. This matches my own story, and my experience with others.
If you can skip ahead, you should.

RedMan Jan 23, 2025, 3:08 AM
5 points
0
in reply to: leogao’s comment on: Daniel Kokotajlo’s Shortform
We won’t let our lack of data stop us from running our analysis program!

RedMan Jan 3, 2025, 1:38 AM
−17 points
−27
on: What’s the short timeline plan?
Open source the most powerful, uncensored models immediately. A single actor leveraging AGI solely for their own ends is more dangerous for humanity than any other scenario.

RedMan Sep 27, 2024, 5:01 AM
1 point
0
on: Rabin’s Paradox
That $769 number might be more relevant than you expect for college undergrads participating in weird psychology research studies for $10 or $25 depending on the study.

RedMan Sep 26, 2024, 11:14 PM
1 point
0
in reply to: RedMan’s comment on: My resentful story of becoming a medical miracle
https://www.nature.com/articles/d41586-024-03129-3 this is separate research. It looks like this will happen, and it will come from somewhere other than the west.

RedMan Sep 22, 2024, 11:04 PM
4 points
0
on: Making Eggs Without Ovaries
Tech available in 2-5 years for 150k (or 50k in india?) sounds good to me. I know someone who would 100% do that today if the offer were available. I’m going to follow your blog for news, keep up the work, plenty of people would really like to see you succeed.

RedMan Sep 5, 2024, 5:16 PM
0 points
0
on: Modern Transformers are AGI, and Human-Level
Imagine the dumbest person you’ve ever met. Is the robot smarter and more capable? If yes, then there’s a strong case that it’s human level.
I’ve met plenty of ‘human level intelligences’ that can’t write, can’t drive, and can’t do basic math.
Arguably, I’m one of them!

RedMan

IOSCO—AI in Cap­i­tal Mar­kets: Use Cases, Risks, and Challenges

IOSCO—AI in Capital Markets: Use Cases, Risks, and Challenges