(Coming from thebrowser.com newsletter where this was curated)
The opsec and entropy points sound right to me. One thing I’ll add—microwaves are Faraday cages. If you’re really anxious about security you can put all your devices in a microwave while talking to your lawyer/therapist/journalist. (Speakers, headphones, mice etc. can turn into mics too)
On a topic this serious it’s worth interviewing people who’ve blown the whistle, and the lawyers who’ve represented them—you can’t figure the enforcement reality ex-ante first-principles style, some real stories would make this document more robust. A couple hours of consultation would make a big difference and I’m sure you can get it funded.
In particular, it sounds weird to me to believe lawyers risk jail, for therapists to be untrustworthy (you don’t need to tell them the details in any case), or to think Russia is safe (I really don’t think you want the attention of Russian authorities. I met a journalist whose colleague was kidnapped by the FSB once. Your case will be dramatically complicated by adding a rogue state into the mix.) I am not a lawyer, I am entirely guessing here, but please consider talking to one (especially if you believe whistleblowers can’t safely talk to lawyers, doing it for them is a public service).
And one thing you might find is that psychology is a major hurdle. I filed an SEC whistleblower complaint once (different story, the government encourages cooperation) and the biggest barrier for me was processing the grief of seeing people I trusted commit a crime, so that I could feel safe taking such a drastic feeling step.
And: is the implication that it’s an unconfirmed open secret the AI labs are doing really bad stuff? Has it gotten to that yet? I must say I am deeply skeptical about the AI lab whistleblower types so far—Kokotajlo and Aschenbrenner give me bad vibes, and both have made some very bad predictions.
(Not endorsing breaking any laws, not legal advice, etc. etc. I basically don’t think you should blow the whistle in this way, and I’m not a huge fan of the famous whistleblowers personally, they were reckless and put lives at risk.)
I will have to check more on microwaves, I’m not comfortable recommending them right now.
Regarding the trusted circle part
Daniel Ellsberg’s psychiatrist’s house was broken into by CIA agent Howard Hunt.
Assange’s lawyers had to communicate via encrypted comms because they realised they were being spied on. There is a long list of cases of lawyers being intimidated and surveilled.
You can just go look at the websites of legal resources on this topic, none of them are willing to even hint at the notion that they might support classified leaks. If you actually schedule calls with them, you’ll realise the same thing.
Regarding psychology, I basically agree and I don’t think I have solved this problem yet. Hence I didn’t share much about it. I’d love if you could write more about it about your personal experience (either in public, or we could have a private chat) as it is valuable info.
Regarding Russia, my only consideration is probability of not going to prison (or worse). Do you propose any other country with a lower probability of going to prison?
is the implication that it’s an unconfirmed open secret the AI labs are doing really bad stuff?
Almost certainly yes as of today, and everything builds a larger picture. For instance if some of the leaders are proven to be sex offenders like Altman has been accused of (idk if the accusations are true), or are proven to be successionists like Rich Sutton is, it strengthens the political movement against building ASI.
I expect my guide to be even more valuable a couple years from now, than it is today.
I’m not a huge fan of the famous whistleblowers personally, they were reckless and put lives at risk.
I can maybe see why you have that opinion, but I think the criteria here should be something closer to—who used less violence—the US govt or the people exposing them? Since I expect ASI to lead to human extinction or a permanent dictatorship, I am highly open to solutions that involve some violence or collateral damage, but are not as bad as the ASI outcome itself.
(Coming from thebrowser.com newsletter where this was curated)
The opsec and entropy points sound right to me. One thing I’ll add—microwaves are Faraday cages. If you’re really anxious about security you can put all your devices in a microwave while talking to your lawyer/therapist/journalist. (Speakers, headphones, mice etc. can turn into mics too)
On a topic this serious it’s worth interviewing people who’ve blown the whistle, and the lawyers who’ve represented them—you can’t figure the enforcement reality ex-ante first-principles style, some real stories would make this document more robust. A couple hours of consultation would make a big difference and I’m sure you can get it funded.
In particular, it sounds weird to me to believe lawyers risk jail, for therapists to be untrustworthy (you don’t need to tell them the details in any case), or to think Russia is safe (I really don’t think you want the attention of Russian authorities. I met a journalist whose colleague was kidnapped by the FSB once. Your case will be dramatically complicated by adding a rogue state into the mix.) I am not a lawyer, I am entirely guessing here, but please consider talking to one (especially if you believe whistleblowers can’t safely talk to lawyers, doing it for them is a public service).
And one thing you might find is that psychology is a major hurdle. I filed an SEC whistleblower complaint once (different story, the government encourages cooperation) and the biggest barrier for me was processing the grief of seeing people I trusted commit a crime, so that I could feel safe taking such a drastic feeling step.
And: is the implication that it’s an unconfirmed open secret the AI labs are doing really bad stuff? Has it gotten to that yet? I must say I am deeply skeptical about the AI lab whistleblower types so far—Kokotajlo and Aschenbrenner give me bad vibes, and both have made some very bad predictions.
(Not endorsing breaking any laws, not legal advice, etc. etc. I basically don’t think you should blow the whistle in this way, and I’m not a huge fan of the famous whistleblowers personally, they were reckless and put lives at risk.)
Thanks for the reply!
I will have to check more on microwaves, I’m not comfortable recommending them right now.
Regarding the trusted circle part
Daniel Ellsberg’s psychiatrist’s house was broken into by CIA agent Howard Hunt.
Assange’s lawyers had to communicate via encrypted comms because they realised they were being spied on. There is a long list of cases of lawyers being intimidated and surveilled.
You can just go look at the websites of legal resources on this topic, none of them are willing to even hint at the notion that they might support classified leaks. If you actually schedule calls with them, you’ll realise the same thing.
Regarding psychology, I basically agree and I don’t think I have solved this problem yet. Hence I didn’t share much about it. I’d love if you could write more about it about your personal experience (either in public, or we could have a private chat) as it is valuable info.
Regarding Russia, my only consideration is probability of not going to prison (or worse). Do you propose any other country with a lower probability of going to prison?
Almost certainly yes as of today, and everything builds a larger picture. For instance if some of the leaders are proven to be sex offenders like Altman has been accused of (idk if the accusations are true), or are proven to be successionists like Rich Sutton is, it strengthens the political movement against building ASI.
I expect my guide to be even more valuable a couple years from now, than it is today.
I can maybe see why you have that opinion, but I think the criteria here should be something closer to—who used less violence—the US govt or the people exposing them? Since I expect ASI to lead to human extinction or a permanent dictatorship, I am highly open to solutions that involve some violence or collateral damage, but are not as bad as the ASI outcome itself.