mruwnik

Karma: 208

mruwnik 22 Mar 2023 17:45 UTC
1 point
0
on: “Publish or Perish” (a quick note on why you should try to make your work legible to existing academic communities)
This sounds like there’s a niche for technical writers who format/edit AF posts as proper ML articles for credits, points and glory

mruwnik 21 Mar 2023 20:55 UTC
1 point
0
in reply to: Ishan’s comment on: Is AI Safety dropping the ball on privacy?
There are extensions like adnauseum which try to poison your data trace. Though it’s dubious whether they help much. You could have some kind of crawler thingy which would pretend to be like 100 normal users so you get lost in the noise. But even that could probably be filtered out if someone really wanted to—it would be hard to accurately simulate a human (I also dimly recall reading an article about it?). Maybe something that records other peoples sessions and plays them back? Or a LLM doing it (hehe)? But even that wouldn’t help in the case of being logged in to various services, and I’m guessing that most people don’t automatically log out of gmail whenever they change tabs?
One source of hope is that data gets stale quickly. People can change their minds (even if they don’t), so just because you know what I thought a year ago doesn’t mean that you know what I think now. Then again, most people don’t care either way and it would be pretty simple to remove the small number of outliers who suddenly go dark. One possible way of cleaning up would be to spend a couple of months posting more and more radically strange posts (e.g. going all in on flat earthism) before going private in order to mislead any analysis. This is hard and requires passing an ITT.
Tor + cleaning cookies + logging out of everything after using them + separate user profiles goes a long way. But it’s very inconvenient.

mruwnik 10 Mar 2023 18:55 UTC
7 points
3
in reply to: Amalthea’s comment on: Speed running everyone through the bad alignement bingo. $5k bounty for a LW conversational agent
Check https://aisafety.info/ - any feedback (or even better—answers) is useful

mruwnik 10 Mar 2023 18:54 UTC
6 points
0
in reply to: trevor’s comment on: Speed running everyone through the bad alignement bingo. $5k bounty for a LW conversational agent
The chatbot should get better over time, but for now the website works quite well as a collection of questions and answers, where the answers go through an editing process. New questions can be easily added, and if you can’t find an answer, but there is a similar question waiting to be answered, you can bump its priority a bit, which will hopefully result in someone working on it sooner.

mruwnik 9 Feb 2023 9:05 UTC
4 points
0
on: Why Are Bacteria So Simple?
One thing worth mentioning is that proto-eukaryotic cells originally got the energy to focus on larger scale things by capturing other prokaryotic cells. Mitochondria and chloroplasts look a lot like simplified bacteria that specialise in producing ATP and glucose+oxygen, respectively, both of which give the cell more energy. They even have their own, specific DNA inside them.
To get really speculative, this is quite similar to how agriculture and the resulting centralisation+specialisation allowed for the explosion of civilization, where the last ~10k years have overwhelmingly more complex stuff than the preceding 100k years.

mruwnik 2 Feb 2023 21:28 UTC
3 points
0
on: Can we “cure” cancer?
You should probably differentiate more between curing and preventing. Most of the infection gains were in preventive measures—both in avoiding infections and boosting the body’s systems. And this is where massive anticancer gains have been possible (e.g. banning asbestos and dissuading smoking).
The low hanging gains in actual curing were generally mechanisms that make use of differences between different kingdoms, e.g. antibiotics targeting bacteria cell walls, since they have a different structure than eukaryotic cells. The problem with cancers is that they’re your own cells that are going wild, for a whole host of reasons. This means that any successful anti-cancer thingy will either have to specifically target cancer cells (which is generally hard) or be area spells, that will also kill off healthy cells. Pretty much what DirectedEvolution wrote.

mruwnik 2 Feb 2023 10:41 UTC
1 point
0
in reply to: Jay Bailey’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
It can be rephrased as a variation of the off button, where rather than just turning itself off, it runs NOPs, and rather than getting pushed manually, it’s triggered by escaping (however that could be defined). A lot of the problems raised in the original paper should also apply to honeypots.

mruwnik 1 Feb 2023 22:17 UTC
1 point
0
in reply to: Chris_Leong’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
Check out corrigibility (someone really should write that tag...), starting from this paper.

mruwnik 1 Feb 2023 22:09 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
It could be somewhat helpful, for sure. And certainly better than nothing (unless it creates a false sense of security). Though I doubt it would be adopted, because of how much it would slow things down.

mruwnik 1 Feb 2023 11:23 UTC
1 point
0
in reply to: PotteryBarn’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
It would depend a lot on the implementation details. Also on whether it’s possible. The basic wireheader seems simpler, in that it doesn’t need to factor in humanities stupid ideas.
Adding in extra details, like that it has to pretty much make a virtual heaven for people, makes it a lot harder than just pumping them with morphine.
maximizing pleasure is very close in idea space to max suffering, because both do pretty much the same, but with opposite signs (yes, this is a travesty of an oversimplification, but you get the idea). The most common raised problems are mainly Goodhart issues. How would you specify that the AI maximise pleasure?
You seem to have a slightly idiosyncratic interpretation of wireheading, in that it’s usually described as blissing out on drugs, rather than being simulated in an enriching and fun environment. A lot more people would be fine with the second option than would agree to the first one.
As to the simple, conscious blissers, they seem empty to me. I expect this is a difference in basic morality/worldview though, in that I don’t find hedonism attractive. And boredom is very useful.

mruwnik 1 Feb 2023 11:08 UTC
1 point
0
in reply to: LVSN’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
So you’re saying that it would be good for it to have general skills, but better? Or that it would be good for it to have skills that humans don’t have? Or that an ensemble of narrow skills that can complement each other, but in isolation won’t be dangerous, would be good?

mruwnik 1 Feb 2023 11:00 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
The answer to all of these is a combination of “dunno” and “it depends”, in that implementation details would be critical.
In general, you shouldn’t read too much into the paperclip maximiser, or rather shouldn’t go too deep into its specifics. Mainly because it doesn’t exist. It’s fun to think about, but always remember that each additional detail makes the overall scenario less likely.
I was unclear about why I asked for clarification of “make as many paperclips as possible in the next hour”. My point there was that you should assume that whatever is not specified should be interpreted in whatever way is most likely to blow up in your face.

mruwnik 1 Feb 2023 10:49 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
Also check out the AI boxing tag

mruwnik 1 Feb 2023 10:49 UTC
1 point
0
in reply to: clearthis’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
Do you mean as a checkpoint/breaker? Or more in the sense of RLHF?
The problem with these is that the limiting factor is human attention. You can’t have your best and brightest focusing full time on what the AI is outputting, as quality attention is a scarce resource. So you do some combination of the following:
- slow down the AI to a level that is manageable (which will greatly limit its usefulness),
- discard ideas that are too strange (which will greatly limit its usefulness)
- limit its intelligence (which will greatly limit its usefulness)
- don’t check so thoroughly (which will make it more dangerous)
- use less clever checkers (which will greatly limit its usefulness and/or make it more dangerous)
- check it reeeaaaaaaallllly carefully during testing and then hope for the best (which is reeeeaaaallly dangerous)
You also need to be sure that it can’t outsmart the humans in the loop, which pretty much comes back to boxing it in.

mruwnik 1 Feb 2023 10:40 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
Depends on who’s predicting, but usually yes. Although the resulting predictions are much more fuzzy, since you have to estimate how funding will change and how far ahead are the known secret labs, and how many really secret labs are in existence. Then you also have to factor in foreign advances, e.g. China.

mruwnik 1 Feb 2023 10:36 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
You could make it more convenient by it asking its operators before doing anything—the result would be pretty much the same.
The problem with this is that you assume that humans will be able to:
- notice what the AI is planning
- understand what the AI is planning
- foresee the consequences of whatever it is planning
which greatly limits the usefulness of the system, in that it won’t be able to suggest anything radically different (or radically faster) from what a human would suggest—human oversight is the limiting factor of the whole system. It also encourages the AI to hide what it’s doing.
There is also the problem that given enough intelligence (computing speed, whatever think-oomph), it’ll still be able to think circles around its operators.

mruwnik 1 Feb 2023 10:27 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
Making a wireheading AGI probably would be easier than getting a properly aligned one, because maximisers are generally simpler than properly aligned AGIS, since they have fewer things to do correctly (I’m being very vague here—sorry).
That being said, having a coherent target is a different problem than being able to aim it in the first place. Both are very important, but it seems that being able to tell an AI to do something and being quite confident in it doing so (with the ability to correct it in case of problems).
I’m cynical, but I reckon that giving a goal like “make people happy” is less likely than “make me rich” or “make me powerful”.

mruwnik 31 Jan 2023 22:41 UTC
1 point
0
in reply to: LVSN’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
What do complementary skills look like?

mruwnik 31 Jan 2023 21:40 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
Good questions.
I think the assumption is that unless it’s specified, then without limit. Like if you said “make me as much money as you can”—you probably don’t want it stopping any time soon. The same would apply to the colour of the paperclips—seeing as you didn’t say they should be red, you shouldn’t assume they will be.
The issue with maximisers is precisely that they maximise. They were introduced to illustrate the problems with just trying to get a number as high as possible. At some point they’ll sacrifice something else of value, just to get a tiny advantage. You could try to provide a perfect utility function, which always gives exactly correct weights for every possible action, but then you’d have solved alignment.
Does “make as many paperclips as possible in the next hour” mean “undergo such actions that in one hours time will result in as many paperclips as possible” or “for the next hour do whatever will result in the most paperclips overall, including in the far future”?

mruwnik 31 Jan 2023 21:06 UTC
1 point
0
in reply to: Droopyhammock’s comment on: All AGI Safety questions welcome (especially basic ones) [~monthly thread]
Do you mean something more specific than general wireheading?
I’d reckon it a lot less likely, just because it’s a lot simpler to kill everyone than to keep them alive and hooked up to a machine. That is, there are lots and lots of scenarios where everyone is dead, but a lot fewer where people end up wireheaded.
Wireheading is often given as a specific example of outer misalignment, where the agent was told to make people happy and does it in a very unorthodox manner.