Posts
Comments
I haven't seen Debate on Instrumental Convergence between LeCun, Russell, Bengio, Zador, and More. Why did it get universally negative votes?
That's the idea behind the post, yeah. I am referring more to the general culture of the site, since it is relevant here.
I find it strange that our response to "politics is the mindkiller" has been less "how can we think more rationally about politics?" and more "let's avoid politics". If feasible, the former would pay off long-term.
Of course, a lot of more general ideas pertaining to rationality can be applied to politics too. But if politics is still the mindkiller, this may not be enough -- more techniques may be needed to deal with the affective override that politics can cause.
Listeners are probably not assuming that the person they are listening to is being honest.
Seconded.
Interesting, thanks for the reply. I agree that it could develop superhuman ability in some domains, even if that ability doesn't manifest in the model's output, so that seems promising (although not very scaleable). I haven't read on mesa optimizers yet.
I have very little knowledge of AI or the mechanics behind GPT, so this is more of a question than criticism:
If a scaled up GPT-N is trained on human-generated data, how would it ever become more intelligent than the people whose data it is trained on?
Or maybe good enough is the enemy of better. Regardless, the point's been made
Perfect is the enemy of good; good enough is also the enemy of good.
In my case, I probably wouldn't give my life for less than lives of a billion strangers, so that ratio would have to be extremely high, to the point where it's probably incalculable.
Why?
I am unfamiliar with the science here -- what is the difference between a "reversed-effect stimulant" and a depressant?
We wax poetic about both because we like doing it. I don't think that is affected by whether it's a fluke of evolution
There are some acronyms that are already in common usage.
Suggestion: continue to use qualifiers, but encourage writing them as acronyms so that they take up less space. Same meaning but quicker and less annoying.
"Wrong" as in "less likely to match reality." Not very much is certain, but that doesn't mean we are forbidden from talking about certainty.
It seems like you are interpreting a discussion that doesn't effortlessly concede to your point of view as a discussion inherently biased against your point of view.
Pledging now to join if at least 8 do.
Same here.
People do have different metabolism, though. And they do to such an extent that some struggle to maintain an ideal weight more than others. Do you disagree?
- Use some of this ten years of energy to break the door.
- Or the walls. If there isn't a door. I think ten years is enough for either.
- Or fuck it, the ground. I dig my way out with my superhumanly powerful hands.
- The ceiling! I can claw into the walls to climb up.
- Painstakingly carve a key out of one of the parts in my phone.
- Wait, but I like my phone! I painstakingly carve a key out of a metal piece of my clothing.
- Send a friend my location, so that they can help me escape.
- Or I just call 911.
- Host a "Babble challenge" on LessWrong to crowdsource ideas.
- Get media attention.
- Ask the room politely to let me out.
- Ask the room aggressively to let me out.
- Bash a hole in the wall with my phone.
- ...or the door. Or any other border between the room and the everywhere else.
- Call angrily whoever put me here.
- Call nicely whoever put me here, then convince them that there is a second person here. Your experiment is ruined! Surely, you should momentarily open the door so that this other person may leave.
- Wake up.
- Determine my address, order something on Amazon, then play an audio file containing repeated screams for help as the Amazon worker arrives.
- Or just do the screaming myself. But that's kind of tiring.
- Hire someone to go to the address I am at, stare at the door and scream for help...?
- Call a friend asking them to go to the ad........this seems overly convoluted
- Make a list of 50 strategies to get out. Choose the best one. Proceed.
- The same, but with 5,000 strategies. I have a little over ten years, don't I?
- The same, but I try each in order. One is bound to work.
- Call a friend, asking them for strategies. Ignore their repeated offers to help, refusing to reveal that the only reason for this is "that only gets to happen in universe #7!"
- Call 911, asking them for strategies. Repeat until an officer is sent to my location to charge me for false alarm or something.
- Actually, this officer thing is a good idea. Loudly threaten to commit various crimes online. Each post contains a signature; it is my current address in ROT13. They'll never find me!
- Loudly threaten to commit various crimes unless someone goes to my current address (not in ROT13).
- Offer to donate $10,000 to a charity of the address-goer's choice. Omit the fact that I don't have $10,000.
- Offer to give $10,000 to the address-goer. Omit the fact that I don't have $10,000, then lock the address-goer in the room (without a phone!) on my way out so as to avoid the consequences of failing to deliver on my promise.
- Announce that someone other than me has died at my current address. Wait for people to arrive there and collect the body.
- Announce that I have died at my current address. Wait for people to arrive there and collect my body. Surprise!
- Announce, as myself, that I have died at my current address. State that I will be ghost-stabbing my adversaries (I'm a ghost, so it won't do any real harm). Wait for people to arrive and deliver me to a mental institution.
- Announce, as myself, that someone else died at my current address. State how proud I am of having killed them. State that I will be residing here and feasting on the body of my victim for the next two weeks.
- Make my address a geocache. Of course, the box must be found by opening that locked door right there. The box is kind of fleshy and weirdly shaped and please don't open it
- Eat the door, which is made of some edible substance.
- Eat the door, which is not made of an edible substance. Successfully leave the room, then die shortly afterwards.
- Announce, on question 38 of a coincidentally similar LessWrong "Babble challenge," that I am in the exact situation described by the post, hoping that some LessWronger will save me. Please. help
- Go to sleep. Wake up in another parallel universe, never remembering my situation the day before. This is our reality, and we'll never know.
- Punch the window.
- Throw my phone at the window.
- Open the window. Maybe only one part of it is locked.
- No, all of it is locked. I ask the person sitting under the window to open it from the other side.
- Wait out my ten years. Once I'm dead, I find there is an afterlife!
- Try to eat the door. Die before I leave. There is an afterlife!
- Become better at figuring out how to leave locked, empty rooms.
- Cut whatever is causing the door to be locked with a sharp component of my phone.
- Cut whatever is causing the door to be locked with a sharp metal element of my clothing!
- Hire a consultant on escaping locked, empty rooms.
- By pure luck, make a string of discoveries that allow me to make a superintelligent AI with the purpose of escaping me from this room.
This took me ~35 mins.
I would suggest about three, four, or five main rationalist hubs as opposed to one. This could be a compromise between total dispersion/decentralization and lack of respect for the differing preferences of rationalists.