id
stringlengths
17
17
body
stringlengths
0
19.5k
posted_at
stringlengths
24
24
karma
int64
-6
185
parent_comment_id
stringlengths
17
17
post_id
stringlengths
17
17
post_title
stringlengths
2
127
post_slug
stringlengths
2
61
post_url
stringlengths
20
146
post_author
stringclasses
214 values
post_posted_at
stringlengths
24
24
JcdH7uMcdbicL6bFf
It doesn't say "during 2040", it says "before 2040".
2024-09-12T00:37:26.454Z
7
GvDQFbzeigsbYS9sJ
4kuXNhPf9FBwok7tK
AI forecasting bots incoming
ai-forecasting-bots-incoming
https://www.safe.ai/blog/forecasting
Dan H
2024-09-09T19:14:31.050Z
qLccD6pLhQnq9zeXZ
I'd also add that a high fraction of these costs won't be increased if you improve cyber crime productivity (by e.g. 10%). As in, maybe a high fraction of the costs are due to the possiblity of very low effort cyber crime (analogous to the cashier case). And Fabien's original motivation was more closely related to thi...
2024-09-27T20:32:02.081Z
10
MmzibToAQaFHN6EjW
KLDc5JuQZLiimB58b
Is cybercrime really costing trillions per year?
is-cybercrime-really-costing-trillions-per-year
https://www.lesswrong.com/posts/KLDc5JuQZLiimB58b/is-cybercrime-really-costing-trillions-per-year
Fabien Roger
2024-09-27T08:44:07.621Z
sGJfhhWgdDYw48qbC
I don't love "smooth" vs "sharp" because these words don't naturally point at what seem to me to be the key concept: the duration from the first AI capable of being [transformatively useful](https://www.lesswrong.com/posts/kcKrE9mzEHrdqtDpE/the-case-for-ensuring-that-powerful-ais-are-controlled#Control_is_likely_achiev...
2024-09-29T02:36:02.441Z
46
null
6svEwNBhokQ83qMBz
"Slow" takeoff is a terrible term for "maybe even faster takeoff, actually"
slow-takeoff-is-a-terrible-term-for-maybe-even-faster
https://www.lesswrong.com/posts/6svEwNBhokQ83qMBz/slow-takeoff-is-a-terrible-term-for-maybe-even-faster
Raemon
2024-09-28T23:38:25.512Z
9BijEkmHFLjpKoNHx
Long duration/short duration takeoff
2024-09-29T02:41:00.314Z
2
rhoFrebAgbd2XzYv5
6svEwNBhokQ83qMBz
"Slow" takeoff is a terrible term for "maybe even faster takeoff, actually"
slow-takeoff-is-a-terrible-term-for-maybe-even-faster
https://www.lesswrong.com/posts/6svEwNBhokQ83qMBz/slow-takeoff-is-a-terrible-term-for-maybe-even-faster
Raemon
2024-09-28T23:38:25.512Z
sSBStJps5K5MAChSy
> But, I also think that variable is captured by saying "smooth takeoff/long timelines?" (which is approximately what people are currently saying? You can have smooth and short takeoff with long timelines. E.g., imagine that scaling works all the way to ASI, but requires a ton of baremetal flop (e.g. 1e34) implying lo...
2024-09-29T03:38:27.311Z
6
Dri5YtMsE5c8AitX6
6svEwNBhokQ83qMBz
"Slow" takeoff is a terrible term for "maybe even faster takeoff, actually"
slow-takeoff-is-a-terrible-term-for-maybe-even-faster
https://www.lesswrong.com/posts/6svEwNBhokQ83qMBz/slow-takeoff-is-a-terrible-term-for-maybe-even-faster
Raemon
2024-09-28T23:38:25.512Z
3jpcMofEkSXpakB4b
I think we're pretty likely to see a smooth and short takeoff with ASI prior to 2029. Now, imagine that you were making this exactly prediction up through 2029 in 2000. From the perspective in 2000, you are exactly predicting smooth and short takeoff with long timelines! So, I think this is actually a pretty natural p...
2024-09-29T03:40:56.386Z
5
sSBStJps5K5MAChSy
6svEwNBhokQ83qMBz
"Slow" takeoff is a terrible term for "maybe even faster takeoff, actually"
slow-takeoff-is-a-terrible-term-for-maybe-even-faster
https://www.lesswrong.com/posts/6svEwNBhokQ83qMBz/slow-takeoff-is-a-terrible-term-for-maybe-even-faster
Raemon
2024-09-28T23:38:25.512Z
nTPkKDoAAHskJmviD
Long takeoff and short takeoff sound strange to me. Maybe because they are too close to long timelines and short timelines.
2024-09-29T03:45:55.070Z
4
Dri5YtMsE5c8AitX6
6svEwNBhokQ83qMBz
"Slow" takeoff is a terrible term for "maybe even faster takeoff, actually"
slow-takeoff-is-a-terrible-term-for-maybe-even-faster
https://www.lesswrong.com/posts/6svEwNBhokQ83qMBz/slow-takeoff-is-a-terrible-term-for-maybe-even-faster
Raemon
2024-09-28T23:38:25.512Z
uHQSNqRSJfcWnTFKm
> 'like the industrial revolution but 100x faster' Actually, much more extreme than the industrial revolution because at the end of the singularity you'll probably discover a huge fraction of all technology and be able to quickly multiply energy production by a billion. That said, I think Paul expects a timeframe lon...
2024-09-29T16:40:30.980Z
5
7Fx4pLAsqo6A76pGf
6svEwNBhokQ83qMBz
"Slow" takeoff is a terrible term for "maybe even faster takeoff, actually"
slow-takeoff-is-a-terrible-term-for-maybe-even-faster
https://www.lesswrong.com/posts/6svEwNBhokQ83qMBz/slow-takeoff-is-a-terrible-term-for-maybe-even-faster
Raemon
2024-09-28T23:38:25.512Z
nYDH3LkyQgEtMpAJZ
I find this essay interesting as a case study in discourse and argumentation norms. Particularly as a case study of issues with discourse around AI risk. When I first skimmed this essay when it came out, I thought it was ok, but mostly uninteresting or obvious. Then, on reading the comments and looking back at the bod...
2024-09-29T19:03:40.680Z
57
null
LvKDMWQ3yLG9R3gHw
'Empiricism!' as Anti-Epistemology
empiricism-as-anti-epistemology
https://www.lesswrong.com/posts/LvKDMWQ3yLG9R3gHw/empiricism-as-anti-epistemology
Eliezer Yudkowsky
2024-03-14T02:02:59.723Z
2Q24vNtzyGkhBQYpP
What are the close-by arguments that are actually reasonable? Here is a list of close-by arguments (not necessarily endorsed by me!): - **On empirical updates from current systems**: If current AI systems are broadly pretty easy to steer and there is good generalization of this steering, that should serve as some evid...
2024-09-29T19:04:10.202Z
11
nYDH3LkyQgEtMpAJZ
LvKDMWQ3yLG9R3gHw
'Empiricism!' as Anti-Epistemology
empiricism-as-anti-epistemology
https://www.lesswrong.com/posts/LvKDMWQ3yLG9R3gHw/empiricism-as-anti-epistemology
Eliezer Yudkowsky
2024-03-14T02:02:59.723Z
XgmzbyHoeF4BYPvqL
I don't think "continuous" is self-evident or consistently used to refer to "a longer gap from human-expert level AI to very superhuman AI". For instance, in the very essay you link, Tom argues that "continuous" (and fairly predictable) doesn't imply that this gap is long!
2024-09-29T19:10:07.474Z
9
omSfZMziRwx3G8vYy
dHrTjywTTfKJZ5TQL
COT Scaling implies slower takeoff speeds
cot-scaling-implies-slower-takeoff-speeds
https://www.lesswrong.com/posts/dHrTjywTTfKJZ5TQL/cot-scaling-implies-slower-takeoff-speeds
Logan Zoellner
2024-09-28T16:20:00.320Z
xQWSbQ68Q75Mp9Edt
This is a great post on the topic which I ~~pretty much entirely~~ mostly agree with. Thanks for writing this so I didn't have to! > I think the argument presented in this post is a pretty strong case against "The AI will kill literally everyone with more than 80% probability", so I wish people either stopped saying t...
2024-09-29T20:25:10.469Z
16
null
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
8np8ySapyCjXSw8RB
I also don't think making any commitment is actually needed or important except under relatively narrow assumptions.
2024-09-29T20:27:54.208Z
10
6Lf3gkdpWnTNyPDAJ
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
ePmLCsrASz2K9S9zH
> we have the proposal of simulating smaller Universes and less coordinated humans, which makes the AI think that the simulators might be richer and have a better chance of solving alignment This only matters if the AIs are CDT or dumb about decision theory etc.
2024-09-30T00:18:29.857Z
2
qavmddaLzKCeevSJr
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
HETwZubqshjfEfFpL
> their force of course depends on the degree to which you think alignment is easy or hard. I think even if aliens similar to humans always fail at alignment, it's plausible that this type of scheme saves some humans because more competent aliens bail us out.[^before] This is even less good to depend on... [^before]...
2024-09-30T00:25:21.392Z
4
pvwtQTHNdmz5tpPTH
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
T8jLe3tsKeSmBtBwe
> acausal trade framework rest on the assumption that we are in a (quantum or Tegmark) multiverse I think the argument should also go through without simulations and without the multiverse so long as you are a UDT-ish agent with a reasonable prior.
2024-09-30T00:37:13.150Z
6
J3ubhnhe2dvZbZ8Y6
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
F7Ez3wseFNoktTBwF
To be clear, I think the exact scheme in [A proposal for humanity in the future](https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your#A_proposal_for_humanity_in_the_Future) probably doesn't work as described because the exact level of payment is wrong and more mi...
2024-09-30T01:01:25.954Z
4
xQWSbQ68Q75Mp9Edt
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
KzTnpADWW7fYmrCag
I agree that Nate's post makes good arguments against AIs spending a high fraction of resources on being nice or on stuff we like (and that this is an important question). And it also debunks some bad arguments against small fractions. But the post really seems to be trying to argue against small fractions in general: ...
2024-09-30T01:33:02.101Z
4
gSoPjSqvoJMfyxJrS
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
XXviraNeaa5v4BbyE
> However, do I still understand correctly that spinning the quantum wheel should just work, and it's not one branch of human civilization that needs to simulate all the possible AIs, right? This is my understanding.
2024-09-30T01:35:10.691Z
3
mpdufAmbRsEvq4Yt9
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
nmPft9JxugYkjfxBc
Another way to put this is that posts should often discuss their limitations, particular when debunking bad arguments that are similar to more reasonable arguments. I think discussing limitations clearly is a reasonable norm for scientific papers that reduces the extent to which people intentionally or unintentionally...
2024-09-30T01:37:44.604Z
15
nYDH3LkyQgEtMpAJZ
LvKDMWQ3yLG9R3gHw
'Empiricism!' as Anti-Epistemology
empiricism-as-anti-epistemology
https://www.lesswrong.com/posts/LvKDMWQ3yLG9R3gHw/empiricism-as-anti-epistemology
Eliezer Yudkowsky
2024-03-14T02:02:59.723Z
ugGBcrDvumZudBLEJ
I think we should have a norm that you should explain the limitations of the debunking when debunking bad arguments, particularly if there are stronger arguments that sound similar to the bad argument. A more basic norm is that you shouldn't claim or strongly imply that your post is strong evidence against something w...
2024-09-30T01:48:31.569Z
28
gSoPjSqvoJMfyxJrS
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
se2YKGAyhLcZHsFMK
Notably, David is proposing that AIs take different actions prior to making powerful sims: not kill all the humans.
2024-09-30T03:26:50.884Z
2
pwqybpihvvyRoKox6
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
Luug5CTZrFdG7AZAP
Yes, but most of the expected cost is in keeping the humans alive/happy prior to being really smart. This cost presumably goes way down if it kills everyone physically and scans their brains, but people obviously don't want this.
2024-09-30T15:14:56.612Z
2
FnnyDhaLhp7igaD7i
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
vysHsvNZLv9PRebva
Note that we don't want the AI to hand over the reins to humans, just to avoid killing humans when possible/cheap and we're willing to pay quite a bit for this (in the proposal, IDK if I personally think we should pay). So, it should look like a good offer for AIs who care about Tegmark IV (with a measure etc). So, i...
2024-09-30T15:30:05.698Z
12
mgxPyiFE8CqJHeph5
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
PddExgQEwFDLSfBuw
I agree that it is kind of insane for an AGI which cares about scope sensitive resources to treat sims in this way and thus we should expect a more sensible decision theory. > Introducing the option of creating lots of simulations of your adversary in the future where you win doesn’t seem like it’d change the result ...
2024-09-30T15:48:06.280Z
5
d46wNMrAHtqZQNxgj
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
F5bjJo4hApRKFPiQ4
This proposal doesn't depend on mugging the AI. The proposal actually gets the AI more resources in expectation due to a trade. I agree the post is a bit confusing and unclear about this. (And the proposal under "Can we get more than this" is wrong. At a minimum, such AIs will also be mugged by everyone else too meani...
2024-09-30T15:49:20.950Z
2
h57B9ufmJBMKFAXrX
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
z5fXrEADdPEmeTsBi
Let's conservatively say that evolved life gets around 1% of the multiverse/measure and that evolved life is willing to pay 1/million of its resources in expectation to save aliens from being killed (either "selfishly" to save their own civilization via UDT/FDT supposing that AIs are good enough predictors at the relev...
2024-09-30T16:01:25.648Z
2
djsDxRnx7cMoochKn
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
zN6BtY9Xsz3qcm7nn
Some more notes: - We shouldn't expect that we get a huge win from AIs which are anthropically muggable, as discussed in [Can we get more than this?](https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your?commentId=djsDxRnx7cMoochKn#Can_we_get_more_than_this_), b...
2024-09-30T16:17:09.203Z
4
xQWSbQ68Q75Mp9Edt
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
R4wQusHjbTwNijw6s
No, it is in the AIs best interest to keep humans alive because this gets it more stuff.
2024-09-30T16:18:04.614Z
2
sFC8Sgjpz2AYSGTsf
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
Fjf6Yhqwr3aqLBcdx
You said "shouldn't just do what's clearly in his best interests", I was responding to that.
2024-09-30T17:15:30.374Z
4
9gmJ3vujmDMkyzhNx
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
LQ9cg9vZKog9x5KmR
> in full generality, as opposed to (e.g.) focusing on their bretheren. (And I see no UDT/FDT justification for them to pay for even the particularly foolish and doomed aliens to be saved, and I'm not sure what you were aluding to there.) > > [...] > > rather than concentrating their resources on creatures more similar...
2024-09-30T18:08:27.644Z
6
DG5eksbSxRBnoGLD7
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
ef3qzHeAF2HCguSKd
> Towards one extreme, you have the literal people who literally died, before they have branched much; these branches need to happen close to the last minute. By "last minute", you mean "after I existed" right? So, e.g., if I care about genetic copies, that would be after I am born and if I care about contingent life...
2024-09-30T18:19:58.701Z
2
QddJtRioiWWt6J3jk
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
NwXEpdN6WHfMXLGFQ
> will the surviving branches of humanity (near or distant), or other kind civilizations throughout the multiverse, have enough resources on offer to pay for a human reserve here? Nate and I discuss this question [in this other thread](https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-una...
2024-09-30T18:24:28.856Z
4
GGXEb8cqouvjutKsn
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
kEqimyJD8qrpmvzXp
> But trouble arises when the competant ones can refine P(takeover) for the various planets by thinking a little further. Similar to how the trouble arises when you learn the result of the coin flip in a [counterfactual mugging](https://www.lesswrong.com/tag/counterfactual-mugging)? To make it exactly analogous, imagi...
2024-09-30T18:48:00.960Z
4
rBrcqRxNh4jtwvBJM
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
rCGpAZbxGTC8mjctn
Hmm, maybe I misunderstood your point. I thought you were talking about using simulations to anthropically capture AIs. As in, creating more observer moments where AIs take over less competent civilizations but are actually in a simulation run by us. If you're happy to replace "simulation" with "prediction in a way th...
2024-09-30T22:10:43.809Z
2
NaXt2fiu9HpDSJYeX
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
TLAiRvrT3FkauRcCr
I probably won't respond further than this. Some responses to your comment: --- I agree with your statements about the nature of UDT/FDT. I often talk about "things you would have commited to" because it is simpler to reason about and easier for people to understand (and I care about third parties understanding this...
2024-09-30T22:42:33.828Z
7
KbtGWj6euDdCvXRvN
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
BQjJ62RGXidyBCsSe
> I agree that "not dying in a base universe" is a more reasonable thing to care about than "proving people right that takeoff is slow" but I feel like both lines of argument that you bring up here are doing something where you take a perspective on the world that is very computationalist, unituitive and therefore take...
2024-10-01T00:26:14.415Z
4
6PegBSdcA3dz3MjeC
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
3dg3HsJKGNLfRy2rM
(I mostly care about long term future and scope sensitive resource use like habryka TBC.) Sure, we can amend to: "I believe that AI takeover would eliminate humanity's control over its future, has a high probability of killing billions, and should be strongly avoided." --- We could also say something like "AI takeo...
2024-10-01T01:24:55.526Z
4
8uzgPMh9MEE4dGodk
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
xffc8ffKXWTT4YF99
> And I think this is also true by the vast majority of common-sense ethical views. People care about the future of humanity. "Saving the world" is hugely more important than preventing the marginal atrocity. Outside of EA I have never actually met a welfarist who only cares about present humans. People of course thin...
2024-10-01T01:29:22.272Z
2
8uzgPMh9MEE4dGodk
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
HGvxDGXJWSx9se8h3
At a more basic level, I think the situation is just actually much more confusing than human extinction in a bunch of ways. (Separately, under my views misaligned AI takeover seems *worse* than human extinction due to (e.g.) biorisk. This is because primates or other closely related seem very likely to re-evolve into ...
2024-10-01T01:31:40.730Z
4
xffc8ffKXWTT4YF99
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
nyCh4o9y7GiMqxQxt
Thanks, this seems like a reasonable summary of the proposal and a reasonable place to wrap. I agree that kindness is more likely to buy human survival than something better described as trade/insurance schemes, though I think the insurance schemes are reasonably likely to matter. (That is, reasonably likely to matte...
2024-10-01T07:04:34.157Z
6
WhqrAXvmuD4YdNDtS
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
z3wdu7vGjTibsaWdz
Here is another more narrow way to put this argument: - Let's say Nate is 35 (arbitrary guess). - Let's say that branches which deviated 35 years ago would pay for our branch (and other branches in our reference class). The case for this is that many people are over 50 (thus existing in both branches), and care about ...
2024-10-01T19:00:41.400Z
4
DjbCDNaE8FohwWpFs
ZLAnH5epD8TmotZHj
You can, in fact, bamboozle an unaligned AI into sparing your life
you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
https://www.lesswrong.com/posts/ZLAnH5epD8TmotZHj/you-can-in-fact-bamboozle-an-unaligned-ai-into-sparing-your
David Matolcsi
2024-09-29T16:59:43.942Z
GiCyswWvngrsaHk2c
(Oops, slow reply) > If SLT were to say nontrivial things about what instruction fine-tuning and RLHF are doing to models, and those things were verified in experiments, would that shift your skepticism? If SLT results in interesting predictions in some case or was generally able to notably improve our ability to pr...
2024-10-03T23:31:27.224Z
14
a9jdQxyG4LQfrDZN5
CmcarN6fGgTGwGuFp
Dialogue introduction to Singular Learning Theory
dialogue-introduction-to-singular-learning-theory
https://www.lesswrong.com/posts/CmcarN6fGgTGwGuFp/dialogue-introduction-to-singular-learning-theory
Olli Järviniemi
2024-07-08T16:58:10.108Z
m8ScYhZmeenT6tc3y
The "Shoggoth-Face" idea seems pretty good[^name]. One issue is that you might still have selection pressure early in training on the Shoggoth to hide reasoning as the Face hasn't yet learned to robustly hide stuff which would be punished. One thing is that I might try to address this is to directly train the Face to...
2024-10-08T18:39:24.467Z
16
MyypbEafojmzW59Qp
TecsCZ7w8s4e2umm4
5 ways to improve CoT faithfulness
5-ways-to-improve-cot-faithfulness
https://www.lesswrong.com/posts/TecsCZ7w8s4e2umm4/5-ways-to-improve-cot-faithfulness
Caleb Biddulph
2024-10-05T20:17:12.637Z
3gczGJtmApfaEE6D8
In case Joshua Achiam ends up reading this post, my question for him is: My understanding is that you think P(misaligned AGI will kill all humans by 2032) is extremely low, like 1e-6. Is this because: - We won't have AGI by 2032? - Much-smarter-than-human AIs could takeover, but not AGI and much-smarter-than-human A...
2024-10-10T17:47:46.079Z
18
null
WavWheRLhxnofKHva
Joshua Achiam Public Statement Analysis
joshua-achiam-public-statement-analysis
https://www.lesswrong.com/posts/WavWheRLhxnofKHva/joshua-achiam-public-statement-analysis
Zvi
2024-10-10T12:50:06.285Z
6KhyMiL5YxkG9tna8
Eliezer did not call for enforcement via nuclear war. He said that: > Make it explicit in international diplomacy that preventing AI extinction scenarios is considered a priority above preventing a full nuclear exchange, and that allied nuclear countries are willing to run some risk of nuclear exchange if that’s what ...
2024-10-10T17:52:14.597Z
16
ZQ8LcPAQQgHz7o6cd
WavWheRLhxnofKHva
Joshua Achiam Public Statement Analysis
joshua-achiam-public-statement-analysis
https://www.lesswrong.com/posts/WavWheRLhxnofKHva/joshua-achiam-public-statement-analysis
Zvi
2024-10-10T12:50:06.285Z
iBqnnv7SdgafxZ5TW
> Violently enforcing certain particularly important principles on non-signatories is entirely within the norm True as stated, though I'm not aware of examples of this being enforced on non-signatories which are nuclear powers. This is just quantitatively riskier, not a notable change in norms. And I agree this seem...
2024-10-10T17:58:55.875Z
8
JZDJbFoRBvGQWLLQv
WavWheRLhxnofKHva
Joshua Achiam Public Statement Analysis
joshua-achiam-public-statement-analysis
https://www.lesswrong.com/posts/WavWheRLhxnofKHva/joshua-achiam-public-statement-analysis
Zvi
2024-10-10T12:50:06.285Z
iF8f8Fp38LAREs4mz
Sorry, I actually meant "not a notable change in norms". I agree that it is quantiatively much costlier from the perspective of the US.
2024-10-10T18:10:01.494Z
2
kTJcvM2mGHj7ivJeo
WavWheRLhxnofKHva
Joshua Achiam Public Statement Analysis
joshua-achiam-public-statement-analysis
https://www.lesswrong.com/posts/WavWheRLhxnofKHva/joshua-achiam-public-statement-analysis
Zvi
2024-10-10T12:50:06.285Z
xGEi4mJyiSJDHaeKR
I agree that it is norms violating for a country to respond to a conventional strike on their datacenter with a nuclear response. This is different from the statement that the conventional strike from the other country is norms violating. I don't think conventional strikes on military assets of nuclear power are that ...
2024-10-10T18:14:06.817Z
8
ftNgrYuBxAxW8QNvS
WavWheRLhxnofKHva
Joshua Achiam Public Statement Analysis
joshua-achiam-public-statement-analysis
https://www.lesswrong.com/posts/WavWheRLhxnofKHva/joshua-achiam-public-statement-analysis
Zvi
2024-10-10T12:50:06.285Z
kehHqCAwsFQYXhcya
This essay doesn't seem very good in terms of making accurate predictions about the future. I wish Dario answered questions like "if we let things go as fast as possible, what will the curve of energy production or industrial production look like" or "how plausible is it that AIs can quickly bootstrap to nanotech and ...
2024-10-12T00:00:16.768Z
123
null
DxvzLngnWHhvdCGLf
Dario Amodei — Machines of Loving Grace
dario-amodei-machines-of-loving-grace
https://darioamodei.com/machines-of-loving-grace#basic-assumptions-and-framework
Matrice Jacobine
2024-10-11T21:43:31.448Z
BYnJgaCHXX3FvtiGe
Actually >1e6 was my conservative guess even without reversible computing. Claude claimed that Landauer Limit is 2e8 times more efficient than current GPUs (A100). (I didn't check Claude, I just asked 3.5 sonnet "How far are modern GPUs from theoretical energy efficiency for non-reversible computing?" and got this ans...
2024-10-12T01:20:13.338Z
10
FtjkrFLHSLoaftaL2
DxvzLngnWHhvdCGLf
Dario Amodei — Machines of Loving Grace
dario-amodei-machines-of-loving-grace
https://darioamodei.com/machines-of-loving-grace#basic-assumptions-and-framework
Matrice Jacobine
2024-10-11T21:43:31.448Z
8qq8fHaL5Jk9wXsbQ
The setup here implies a empirical (but conceptually tricky) research direction: try to take two different AIs trained to both do the same prediction task (e.g. predict next tokens of webtext) and try to correspond their internal structure in some way. It's a bit unclear to me what the desiderata for this research sho...
2024-10-15T00:53:55.478Z
20
null
uJvKLDSSYA4y7vzyx
Minimal Motivation of Natural Latents
minimal-motivation-of-natural-latents
https://www.lesswrong.com/posts/uJvKLDSSYA4y7vzyx/minimal-motivation-of-natural-latents
johnswentworth
2024-10-14T22:51:58.125Z
FMbpFAgJtKYgShPiJ
I agree on abandoning various specifics, but I would note that the new standard is much more specific (less vague) on what needs to be defended against and what the validation process and threat modeling process should be. (E.g., rather than "non-state actors", the RSP more specifically says which groups are and aren'...
2024-10-15T22:05:08.478Z
11
DG93HAcFHsr7E3dpz
KrYNqLkaCnBdHpZAs
Anthropic rewrote its RSP
anthropic-rewrote-its-rsp
https://www.lesswrong.com/posts/KrYNqLkaCnBdHpZAs/anthropic-rewrote-its-rsp
Zach Stein-Perlman
2024-10-15T14:25:12.518Z
nMGtv4MQuLCdhGH8x
Hmm, I looked through the relevant text and I think Evan is basically right here? It's a bit confusing though. --- [The Anthropic RSP V1.0](https://www-cdn.anthropic.com/1adf000c8f675958c2ee23805d91aaade1cd4613/responsible-scaling-policy.pdf) says: > The model shows early signs of autonomous self-replication ability...
2024-10-16T04:17:32.443Z
12
3qpgpa3bhXygbeprc
KrYNqLkaCnBdHpZAs
Anthropic rewrote its RSP
anthropic-rewrote-its-rsp
https://www.lesswrong.com/posts/KrYNqLkaCnBdHpZAs/anthropic-rewrote-its-rsp
Zach Stein-Perlman
2024-10-15T14:25:12.518Z
X9MSL52jtKHoL3xmR
> humanity as a whole probably doesn't want things to get incredibly crazy so fast, and so we're likely to see something tamer Doesn't this require a pretty strong and unprecedented level of international coordination on stopping an obviously immediately extremely valuable and militarily relevent technology? I think a...
2024-10-18T16:07:33.910Z
21
oejpcMryqrJbymsJK
DxvzLngnWHhvdCGLf
Dario Amodei — Machines of Loving Grace
dario-amodei-machines-of-loving-grace
https://darioamodei.com/machines-of-loving-grace#basic-assumptions-and-framework
Matrice Jacobine
2024-10-11T21:43:31.448Z
jtw34wJQhJ5haJDNi
> My guess would be that they're not saying that well-designed control evaluations become untrustworthy It's a bit messy because we have some ability to check whether we should be able to evaluate things. So, there are really three relevant "failure" states for well done control: 1. We can't find countermeasures suc...
2024-10-21T04:55:52.777Z
6
vBeDEAgeB6Rpwqx87
nnvn6kaBLajiicH8e
Sabotage Evaluations for Frontier Models
sabotage-evaluations-for-frontier-models
https://assets.anthropic.com/m/377027d5b36ac1eb/original/Sabotage-Evaluations-for-Frontier-Models.pdf
David Duvenaud
2024-10-18T22:33:14.320Z
F22oqrkgm8a4TvXXc
I'm sympathetic to 'a high fraction of "alignment/safety" work done at AI companies is done due to commercial incentives and has negligible effect on AI takeover risk (or at least much smaller effects than work which isn't influenced by commercial incentives)'. I also think a decent number of ostensibly AI x-risk focu...
2024-10-21T16:18:05.327Z
26
null
h4wXMXneTPDEjJ7nv
A Rocket–Interpretability Analogy
a-rocket-interpretability-analogy
https://www.lesswrong.com/posts/h4wXMXneTPDEjJ7nv/a-rocket-interpretability-analogy
plex
2024-10-21T13:55:18.184Z
sctmLPCtwuiPA4i7e
> I think the primary commercial incentive on mechanistic interpretability research is that it's the alignment research that most provides training and education to become a standard ML engineer who can then contribute to commercial objectives. Is your claim here that a major factor in why Anthropic and GDM do mech in...
2024-10-21T17:53:42.735Z
11
md7QvniMyx3vYqeyD
h4wXMXneTPDEjJ7nv
A Rocket–Interpretability Analogy
a-rocket-interpretability-analogy
https://www.lesswrong.com/posts/h4wXMXneTPDEjJ7nv/a-rocket-interpretability-analogy
plex
2024-10-21T13:55:18.184Z
XhcuJn8FjgmJuTDBe
I'm pretty skeptical that Neel's MATS stream is partially supported/subsidized by GDM's desire to generally hire for capabilities . (And I certainly don't think they directly fund this.) Same for other mech interp hiring at GDM, I doubt that anyone is thinking "these mech interp employees might convert into employees f...
2024-10-21T19:27:28.319Z
10
GRJkrnLqpHAEgFiyi
h4wXMXneTPDEjJ7nv
A Rocket–Interpretability Analogy
a-rocket-interpretability-analogy
https://www.lesswrong.com/posts/h4wXMXneTPDEjJ7nv/a-rocket-interpretability-analogy
plex
2024-10-21T13:55:18.184Z
e5Ts2JM55nbLwN9uP
(I think random (non-leadership) GDM employees generally have a lot of freedom while employees of other companies have much less in-practice freedom (except for maybe longer time OpenAI employees who I think have a lot of freedom).)
2024-10-21T21:29:17.154Z
4
PG8QRiN7gLXMbLWvx
h4wXMXneTPDEjJ7nv
A Rocket–Interpretability Analogy
a-rocket-interpretability-analogy
https://www.lesswrong.com/posts/h4wXMXneTPDEjJ7nv/a-rocket-interpretability-analogy
plex
2024-10-21T13:55:18.184Z
aHMSvxDcQyNzvJQBH
(Huh, good to know this changed. I wasn't aware of this.)
2024-10-22T01:18:34.996Z
2
kmBSZ4aRQ3bHvxpTu
h4wXMXneTPDEjJ7nv
A Rocket–Interpretability Analogy
a-rocket-interpretability-analogy
https://www.lesswrong.com/posts/h4wXMXneTPDEjJ7nv/a-rocket-interpretability-analogy
plex
2024-10-21T13:55:18.184Z
9cmyvERuMgmmxnd74
Our evaluations indicate that we have very poor levels of safety (e.g. the AI would probably be able to escape if it wanted to) and we can't find countermeasures which suffice to ensure any level of safety (without basically giving up on using this model).
2024-10-22T17:11:55.225Z
3
GYf5qmCagNtRWyG4h
nnvn6kaBLajiicH8e
Sabotage Evaluations for Frontier Models
sabotage-evaluations-for-frontier-models
https://assets.anthropic.com/m/377027d5b36ac1eb/original/Sabotage-Evaluations-for-Frontier-Models.pdf
David Duvenaud
2024-10-18T22:33:14.320Z
fSduBe6fva93Ndepb
> I think your internal usage restrictions are wildly unaffordable Sure, but note that Evan says "Lifting those internal usage restrictions requires an affirmative safety case (as specified below) for safe internal usage." I don't think it is (clearly) wildly unaffordable to require an affirmative safety case prior t...
2024-10-23T15:51:14.441Z
9
EwPEMoYBjHorJsp3M
Loxiuqdj6u8muCe54
Catastrophic sabotage as a major threat model for human-level AI systems
catastrophic-sabotage-as-a-major-threat-model-for-human
https://www.lesswrong.com/posts/Loxiuqdj6u8muCe54/catastrophic-sabotage-as-a-major-threat-model-for-human
evhub
2024-10-22T20:57:11.395Z
7XmC9FvzFdJrKhprv
I would say that the "internal use restrictions" are just an ad hoc control argument and I'd prefer to explicitly think about it like that. But, it is worth noting that if you are sufficiently conservative in the deployment (as Evan describes), then I think it would be fine to make a very minimal safety case. Here is ...
2024-10-23T16:02:34.180Z
6
fSduBe6fva93Ndepb
Loxiuqdj6u8muCe54
Catastrophic sabotage as a major threat model for human-level AI systems
catastrophic-sabotage-as-a-major-threat-model-for-human
https://www.lesswrong.com/posts/Loxiuqdj6u8muCe54/catastrophic-sabotage-as-a-major-threat-model-for-human
evhub
2024-10-22T20:57:11.395Z
HH4rRoZdBAKmrcRdj
This paper seems like a great first stab at starting to answer [very important questions about scalable oversight sample efficiency](https://www.lesswrong.com/posts/6AT4vhYzww56CR6cm/scalable-oversight-as-a-quantitative-rather-than-qualitative) (at least for non-scheming models). Better understanding of sample effiicie...
2024-10-24T19:12:04.648Z
7
null
iA87ke4so2eqiJfsi
Balancing Label Quantity and Quality for Scalable Elicitation
balancing-label-quantity-and-quality-for-scalable
https://www.lesswrong.com/posts/iA87ke4so2eqiJfsi/balancing-label-quantity-and-quality-for-scalable
Alex Mallen
2024-10-24T16:49:00.939Z
2gBsST9mgg2zhFZ3g
Counting the number of papers isn't going to be a good strategy. I do think total research outside of labs looks competitive with research from labs and probably research done outside of labs has produced more ~~differential~~ safety progress in total. I also think open weight models are probably good so far in terms...
2024-10-24T23:54:22.508Z
10
prxdMM2LeEkZzKZkR
LmhpAyCn8Xpqfwvhb
IAPS: Mapping Technical Safety Research at AI Companies
iaps-mapping-technical-safety-research-at-ai-companies
https://www.iaps.ai/research/mapping-technical-safety-research-at-ai-companies
Zach Stein-Perlman
2024-10-24T20:30:41.159Z
hq9KDxo84ExHn9Ke8
I'm just talking about research intended to be safety/safety-adjacent. As in, of this research, what has the quality weighted differential safety progress been. Probably the word "differential" was just a mistake.
2024-10-25T18:12:19.514Z
5
LJ37iKrkKzvMteRcY
LmhpAyCn8Xpqfwvhb
IAPS: Mapping Technical Safety Research at AI Companies
iaps-mapping-technical-safety-research-at-ai-companies
https://www.iaps.ai/research/mapping-technical-safety-research-at-ai-companies
Zach Stein-Perlman
2024-10-24T20:30:41.159Z
hHi98JKBrpg4Y2CWW
It depends on the type of transparency. Requirements which just involve informing one part of the government (say US AISI) in ways which don't cost much personnel time mostly have the effect of potentially making the government much more aware at some point. I think this is probably good, but presumably labs would pre...
2024-10-26T20:36:43.381Z
6
f7YdvLvHYqgsHimYq
FDBrNMJ8fyx3TB9rt
What AI companies should do: Some rough ideas
what-ai-companies-should-do-some-rough-ideas
https://www.lesswrong.com/posts/FDBrNMJ8fyx3TB9rt/what-ai-companies-should-do-some-rough-ideas
Zach Stein-Perlman
2024-10-21T14:00:10.412Z
W6NwyRFrRCcEzx9vL
Somewhat off-topic, but isn't this a non-example: > We have strong evidence that interesting semantic features exist in superposition I think a more accurate statement would be "We have a strong evidence that neurons don't do a single 'thing' (either in the human ontology or in any other natural ontology)" combined w...
2024-10-27T01:01:36.795Z
8
zzg26HXcX7FN9nAzS
ztokaf9harKTmRcn4
A bird's eye view of ARC's research
a-bird-s-eye-view-of-arc-s-research
https://www.lesswrong.com/posts/ztokaf9harKTmRcn4/a-bird-s-eye-view-of-arc-s-research
Jacob_Hilton
2024-10-23T15:50:06.123Z
ByyCiDEmQAnkvArJp
> it turns out that enforcing sparsity (and doing an SAE) gives better interp scores than doing PCA It's not clear to me this is true exactly. As in, suppose I want to explain as much of what a transformer is doing as possible with some amount of time. Would I better off looking at PCA features vs SAE features? Yes, ...
2024-10-27T02:29:05.706Z
3
WnbNfsRxAJv4SNccs
ztokaf9harKTmRcn4
A bird's eye view of ARC's research
a-bird-s-eye-view-of-arc-s-research
https://www.lesswrong.com/posts/ztokaf9harKTmRcn4/a-bird-s-eye-view-of-arc-s-research
Jacob_Hilton
2024-10-23T15:50:06.123Z
q8MrZgadJnRJLLQRj
I do think this was reasonably though not totally predictable ex-ante, but I agree.
2024-10-27T02:32:23.958Z
2
fotqcRL3hDLMKHvpP
ztokaf9harKTmRcn4
A bird's eye view of ARC's research
a-bird-s-eye-view-of-arc-s-research
https://www.lesswrong.com/posts/ztokaf9harKTmRcn4/a-bird-s-eye-view-of-arc-s-research
Jacob_Hilton
2024-10-23T15:50:06.123Z
ixSnsM6cPZdp9maiW
Joe's argument here would actually be locally valid if we changed: > a sufficient number of IQ 100 agents with sufficient time can do anything that an IQ 101 agent can do to: >a sufficient number of IQ 100 agents with sufficient time can do anything that some number of IQ 101 agents can do eventually We can see why...
2024-10-30T00:46:18.593Z
10
uKsqfoGrLYAyYjRbn
iDRxuJyte6xvppCa3
How might we solve the alignment problem? (Part 1: Intro, summary, ontology)
how-might-we-solve-the-alignment-problem-part-1-intro
https://www.lesswrong.com/posts/iDRxuJyte6xvppCa3/how-might-we-solve-the-alignment-problem-part-1-intro
Joe Carlsmith
2024-10-28T21:57:12.063Z
pJ6Ynw7bXFZKwL8hf
All of these operationalizations are about exact notions from the training setup. Another important notion is revealed identity: - Does the AI think of the other AI as "itself" to the extent that it thinks about stuff like this at all? - Do the AIs cooperate in a way which is reasonably similar to cooperating with yo...
2024-11-04T19:17:17.040Z
4
null
4j6HJt8Exowmqp245
Different senses in which two AIs can be “the same”
different-senses-in-which-two-ais-can-be-the-same
https://www.lesswrong.com/posts/4j6HJt8Exowmqp245/different-senses-in-which-two-ais-can-be-the-same
Vivek Hebbar
2024-06-24T03:16:43.400Z
Y9Fv2HwFGgyJ8Nu7q
No IMO. (I'm also skeptical of competitiveness with expert jailbreakers, but this isn't a crux. Edit: It does seem to be competitive with Scale's jailbreakers, but I think some human jailbreakers are much better and I expect the gap is more stark for harder to jailbreak models.)
2024-11-04T21:48:42.534Z
4
RJgHw7xfBAE8WXDNQ
vYWDEtJ2b8tkbG7Rv
Behavioral red-teaming is unlikely to produce clear, strong evidence that models aren't scheming
behavioral-red-teaming-is-unlikely-to-produce-clear-strong
https://www.lesswrong.com/posts/vYWDEtJ2b8tkbG7Rv/behavioral-red-teaming-is-unlikely-to-produce-clear-strong
Buck
2024-10-10T13:36:53.810Z
fN5dMfrhfXpKASsPw
> You can use the first X% of scenarios at deployment time and audit them with humans. Only if you don't have privacy guarantees? Also, for the honeypots, you can potentially generate vast numbers of honeypots, use a trusted model to review outputs, and then only have humans look at the cases where the trusted model...
2024-11-06T19:41:01.232Z
3
hwKAAYkExpGoZjudk
RveeCTcoApkAtd7oA
Anthropic: Three Sketches of ASL-4 Safety Case Components
anthropic-three-sketches-of-asl-4-safety-case-components
https://alignment.anthropic.com/2024/safety-cases/
Zach Stein-Perlman
2024-11-06T16:00:06.940Z
YNeJLoSnQe4JtebAb
You can still inspect the underlying CoT to know if you have big issues. It is worth noting that the baseline proposal was already incentivizing the Face to be deceptive to the extent to which this proposal is important (the whole hope was to allow for hiding deception). So, I agree that my proposed change trades off ...
2024-11-15T23:56:56.587Z
2
noHkAy2euqAnWjaGw
TecsCZ7w8s4e2umm4
5 ways to improve CoT faithfulness
5-ways-to-improve-cot-faithfulness
https://www.lesswrong.com/posts/TecsCZ7w8s4e2umm4/5-ways-to-improve-cot-faithfulness
Caleb Biddulph
2024-10-05T20:17:12.637Z
BnF5rxYX6wTEsccq8
> On the other hand, your version of Daniel's protocol loses this property. I'm talking about training only the Face, not training the policy (shoggoth) at all with the proposal I'm imagining. And, these should clearly be separate models such that the training of one doesn't generalize to the other. So, making the F...
2024-11-16T03:59:27.950Z
3
WedYLsDdvuXcBdRov
TecsCZ7w8s4e2umm4
5 ways to improve CoT faithfulness
5-ways-to-improve-cot-faithfulness
https://www.lesswrong.com/posts/TecsCZ7w8s4e2umm4/5-ways-to-improve-cot-faithfulness
Caleb Biddulph
2024-10-05T20:17:12.637Z
Xm5u6RGkeCgye4j2n
To the extent that complexity makes succesfully red-teaming our approaches too difficult, then we could try to mostly depend on simple approaches and only use more complex approaches only if we are confident that at least they won't can't make the situation worse. My general view is that complex methods are only produ...
2024-11-16T06:21:40.180Z
17
GdXtDhRv6grfxFSoj
TBLztZivgMbpYCPvt
Using Dangerous AI, But Safely?
using-dangerous-ai-but-safely
https://www.lesswrong.com/posts/TBLztZivgMbpYCPvt/using-dangerous-ai-but-safely
habryka
2024-11-16T04:29:20.914Z
Qv9GyCZprvto8Rzqr
I agree there are various ways the Face could cause problems via influencing the training of the actual policy. (Both more wild influence that would require clever manipulation and a more mundane reduction of disincentive for problematic reasoning (though I'd note this mundane issue is the core goal of the original pro...
2024-11-16T17:02:24.897Z
2
XtAuNPTn7LCcSG3Lf
TecsCZ7w8s4e2umm4
5 ways to improve CoT faithfulness
5-ways-to-improve-cot-faithfulness
https://www.lesswrong.com/posts/TecsCZ7w8s4e2umm4/5-ways-to-improve-cot-faithfulness
Caleb Biddulph
2024-10-05T20:17:12.637Z
p6piai4E8rcBEfPBM
Or get some other work out of these systems such that you greatly reduce risk going forward.
2024-11-16T18:02:06.655Z
5
J4eAiiBD7jgmkaZSH
nnvn6kaBLajiicH8e
Sabotage Evaluations for Frontier Models
sabotage-evaluations-for-frontier-models
https://assets.anthropic.com/m/377027d5b36ac1eb/original/Sabotage-Evaluations-for-Frontier-Models.pdf
David Duvenaud
2024-10-18T22:33:14.320Z
WRE4WTa2m2EDSNu4q
Sure, I mean that I expect that for the critical regime for TAI, we can literally have the Face be 3.5 sonnet.
2024-11-18T23:09:01.255Z
2
WEqxky9gvdmgGYbba
TecsCZ7w8s4e2umm4
5 ways to improve CoT faithfulness
5-ways-to-improve-cot-faithfulness
https://www.lesswrong.com/posts/TecsCZ7w8s4e2umm4/5-ways-to-improve-cot-faithfulness
Caleb Biddulph
2024-10-05T20:17:12.637Z
LA9WegmwnjTytwkEm
> I guess one way of framing it is that I find the shoggoth/face idea great as a science experiment; it gives us useful evidence! However, it doesn't make very much sense to me as a safety method intended for deployment. Sadly, gathering evidence of misalignment in deployment seems likely to me to be one of the most e...
2024-11-18T23:10:43.788Z
2
WEqxky9gvdmgGYbba
TecsCZ7w8s4e2umm4
5 ways to improve CoT faithfulness
5-ways-to-improve-cot-faithfulness
https://www.lesswrong.com/posts/TecsCZ7w8s4e2umm4/5-ways-to-improve-cot-faithfulness
Caleb Biddulph
2024-10-05T20:17:12.637Z
jKCD3NLizgzN4BhRX
I was asking about HSB not because I think it is similar to the process about AIs but because if the answer differs, then it implies your making some narrower assumption about the inductive biases of AI training. > On generalizing to extremely unlikely conditionals, I think TD-agents are in much the same position as o...
2024-11-19T22:49:48.027Z
2
GJGjSFhFwFubw9aCF
YbEbwYWkf8mv9jnmi
The Shutdown Problem: Incomplete Preferences as a Solution
the-shutdown-problem-incomplete-preferences-as-a-solution
https://www.lesswrong.com/posts/YbEbwYWkf8mv9jnmi/the-shutdown-problem-incomplete-preferences-as-a-solution
Elliott Thornley (EJT)
2024-02-23T16:01:16.378Z
7BkGjJcuGrKsBtY5b
> Q: Why bother splitting shoggoth and face? Why not simply blind the reinforcement process to the internal CoT, like OpenAI claims to be doing with o1? You coulld still have a paraphraser too. > > A: Because if you use the same model weights to do both the internal reasoning and the external token outputs, there's mor...
2024-11-20T00:50:06.347Z
14
null
Tzdwetw55JNqFTkzK
Why Don't We Just... Shoggoth+Face+Paraphraser?
why-don-t-we-just-shoggoth-face-paraphraser
https://www.lesswrong.com/posts/Tzdwetw55JNqFTkzK/why-don-t-we-just-shoggoth-face-paraphraser
Daniel Kokotajlo
2024-11-19T20:53:52.084Z
ZfBiftLAnyT34ap9o
Testing sounds great though I'd note that the way I'd approach testing is to first construct a general test bed that produces problematic behavior through training incentives for o1-like models. (Possibly building on [deepseek's version of o1](https://api-docs.deepseek.com/news/news1120).) Then I'd move to trying a bu...
2024-11-20T19:41:57.441Z
10
TeHRdof83qz9gaKHp
Tzdwetw55JNqFTkzK
Why Don't We Just... Shoggoth+Face+Paraphraser?
why-don-t-we-just-shoggoth-face-paraphraser
https://www.lesswrong.com/posts/Tzdwetw55JNqFTkzK/why-don-t-we-just-shoggoth-face-paraphraser
Daniel Kokotajlo
2024-11-19T20:53:52.084Z
XrENSNCXqWMwfek3f
One aspect of untrusted monitoring which this post neglects, but which seems very important to me, is that we'll be uncertain about which model should be untrusted and the extent to which these models are untrusted. This suggests that ensembling over untrusted monitors could be very promising (and much of the benefits...
2024-12-01T03:27:18.248Z
27
null
GCqoks9eZDfpL8L3Q
How to prevent collusion when using untrusted models to monitor each other
how-to-prevent-collusion-when-using-untrusted-models-to
https://www.lesswrong.com/posts/GCqoks9eZDfpL8L3Q/how-to-prevent-collusion-when-using-untrusted-models-to
Buck
2024-09-25T18:58:20.693Z
fskAJy9gdMMcywknb
Zvi noted that this is an error in [this comment](https://thezvi.substack.com/p/repeal-the-jones-act-of-1920/comment/79099406). Zvi claims this is fixed in the substack, idk about the lesswrong post.
2024-12-01T06:29:58.330Z
2
swEeEC2GZDAFMaQcG
dnH2hauqRbu3GspA2
Repeal the Jones Act of 1920
repeal-the-jones-act-of-1920
https://www.lesswrong.com/posts/dnH2hauqRbu3GspA2/repeal-the-jones-act-of-1920
Zvi
2024-11-27T15:00:06.801Z
yhsYFYnTFdqK6EqtM
I donated $3,000. I've gained and will continue to gain a huge amount of value from LW and other activities of Lightcone Infrastructure, so it seemed like a cooperative and virtuous move to donate.[^cost] I tried to donate at a level such that if all people using LW followed a similar policy to me, Lightcone would be ...
2024-12-02T00:36:34.679Z
57
null
5n2ZQcbc7r4R8mvqc
(The) Lightcone is nothing without its people: LW + Lighthaven's big fundraiser
the-lightcone-is-nothing-without-its-people
https://www.lesswrong.com/posts/5n2ZQcbc7r4R8mvqc/the-lightcone-is-nothing-without-its-people
habryka
2024-11-30T02:55:16.077Z
NyNYZCHCHPPcm3AWg
("well approximates AIXI" seems unlikely to be an accurate description to me. Maybe "closer to what AIXI is doing than to base GPT-4 without a sampling loop" or "dominates human competence" are more accurate.)
2024-12-11T19:32:08.572Z
11
3hEMaorfJyQc4WQuT
byNYzsfFmb2TpYFPW
o1: A Technical Primer
o1-a-technical-primer
https://www.youtube.com/live/6fJjojpwv1I?si=5qYt5BZq3dWQoq0i&t=265
Jesse Hoogland
2024-12-09T19:09:12.413Z
SgmntncSunhHwra8D
> All results are based on runs where the models are given a goal that they later realize is misaligned
2024-12-15T20:45:46.083Z
2
iAAFECnHG7F3h58kH
YTZAmJKydD5hdRSeG
Would catching your AIs trying to escape convince AI developers to slow down or undeploy?
would-catching-your-ais-trying-to-escape-convince-ai
https://www.lesswrong.com/posts/YTZAmJKydD5hdRSeG/would-catching-your-ais-trying-to-escape-convince-ai
Buck
2024-08-26T16:46:18.872Z
hmKctjAvm4ppyni4Y
I often refer to the ideas in this post and think the fundamental point is quite important: structural advantages in quantity, cost, and speed might make AI systems quite useful and thus impactful prior to being broadly superhuman. (The exact estimates in the post do pretty strongly assume the current rough architect...
2024-12-15T21:37:44.106Z
12
null
5sWNnbHRkExfLaS49
Before smart AI, there will be many mediocre or specialized AIs
before-smart-ai-there-will-be-many-mediocre-or-specialized
https://www.lesswrong.com/posts/5sWNnbHRkExfLaS49/before-smart-ai-there-will-be-many-mediocre-or-specialized
Lukas Finnveden
2023-05-26T01:38:41.562Z
JLb3yx6fA7k6zuwSA
ASI alignment could be trivial (happens by default), easy, or hard. If it is trivial, then "if we build it, we all die" is false. Separately, I don't buy that misaligned ASI with totally alien goals and that full takes over will certainly kill everyone due to[^base] [trade arguments like this one](https://www.lesswron...
2024-12-15T22:41:46.965Z
16
uKSLixKuydqZJaYta
cqF9dDTmWAxcAEfgf
Communications in Hard Mode (My new job at MIRI)
communications-in-hard-mode-my-new-job-at-miri
https://www.lesswrong.com/posts/cqF9dDTmWAxcAEfgf/communications-in-hard-mode-my-new-job-at-miri
tanagrabeast
2024-12-13T20:13:44.825Z
FbccrNvFWcSkjS3oQ
You said: > How does it not show that it would try to escape if it could? > > Gemini still tried to escape 1% of the time, even without any behavioral nudging. However, it isn't that they found Gemini would try to escape 1% of the time in a natural internal usage context, they gave it a conflicting goal (and informe...
2024-12-16T18:10:47.408Z
7
tcwF3QGr2iDQHRFpD
YTZAmJKydD5hdRSeG
Would catching your AIs trying to escape convince AI developers to slow down or undeploy?
would-catching-your-ais-trying-to-escape-convince-ai
https://www.lesswrong.com/posts/YTZAmJKydD5hdRSeG/would-catching-your-ais-trying-to-escape-convince-ai
Buck
2024-08-26T16:46:18.872Z
pGyvYuRCAYELHajCB
As in, for the literal task of "solve this code forces problem in 30 minutes" (or whatever the competition allows), o3 is ~ top 200 among people who do codeforces (supposing o3 didn't cheat on wall clock time). However, if you gave humans 8 serial hours and o3 8 serial hours, much more than 200 humans would be better. ...
2024-12-21T21:08:48.648Z
10
fYELkmvB5ebJHgxek
Ao4enANjWNsYiSFqc
o3
o3
https://www.lesswrong.com/posts/Ao4enANjWNsYiSFqc/o3
Zach Stein-Perlman
2024-12-20T18:30:29.448Z
YTESYYcC87yncj8uX
> scaring laws lol
2024-12-21T21:09:54.801Z
8
d8pGJMEadpd7kR93Z
5BbcaPPeAoe9nJJwC
Anthropic leadership conversation
anthropic-leadership-conversation
https://www.youtube.com/watch?v=om2lIWXLLN4
Zach Stein-Perlman
2024-12-20T22:00:45.229Z
7zyzD4wkHFAmfasHS
This plan seems to underemphasize security. I expect that for 10x AI R&D[^prod], you strongly want state proof security (SL5) against weight exfiltration and then quickly after that you want this level of security for algorithmic secrets and unauthorized internal usage[^usage][^evenhigher]. Things don't seem to be on ...
2024-12-22T18:09:17.312Z
34
null
SvvYCH6JrLDT8iauA
When AI 10x's AI R&D, What Do We Do?
when-ai-10x-s-ai-r-and-d-what-do-we-do
https://www.lesswrong.com/posts/SvvYCH6JrLDT8iauA/when-ai-10x-s-ai-r-and-d-what-do-we-do
Logan Riggs
2024-12-21T23:56:11.069Z
o5oPjKzppkfKfBLje
Donating to the LTFF seems good.
2024-12-22T23:30:33.076Z
22
RqQ9hZMERCavvaKis
jb4bBdeEEeypNkqzj
Orienting to 3 year AGI timelines
orienting-to-3-year-agi-timelines
https://www.lesswrong.com/posts/jb4bBdeEEeypNkqzj/orienting-to-3-year-agi-timelines
Nikola Jurkovic
2024-12-22T01:15:11.401Z
yj7hySp2JdsfxrBQg
I think the best bull case is something like: They did this pretty quickly and were able to greatly improve performance on a moderately diverse range of pretty checkable tasks. This implies OpenAI likely has an RL pipeline which can be scaled up to substantially better performance by putting in easily checkable tasks ...
2024-12-23T17:52:27.248Z
24
d9oFg4HosqpKqyhet
Ao4enANjWNsYiSFqc
o3
o3
https://www.lesswrong.com/posts/Ao4enANjWNsYiSFqc/o3
Zach Stein-Perlman
2024-12-20T18:30:29.448Z
hQedm7pvNvHPy4Q6o
> progress from o1 to o3 was only three months Can't we just count from announcement to announcement? Like sure, they were working on stuff before o1 prior to having o1 work, but they are always going to be working on the next thing. Do you think that o1 wasn't the best model (of this type) that OpenAI had internally...
2024-12-23T17:56:02.942Z
5
ocPQPKtc3wX3rDkDw
Ao4enANjWNsYiSFqc
o3
o3
https://www.lesswrong.com/posts/Ao4enANjWNsYiSFqc/o3
Zach Stein-Perlman
2024-12-20T18:30:29.448Z
oCRtLKRv4xydReJkq
Yeah, sorry this is an important caveat. But, I think very superhuman performance in most/all checkable domains is pretty spooky and this is even putting aside how it generalizes.
2024-12-23T18:10:18.990Z
8
Hgr8daXytktc4tFWR
Ao4enANjWNsYiSFqc
o3
o3
https://www.lesswrong.com/posts/Ao4enANjWNsYiSFqc/o3
Zach Stein-Perlman
2024-12-20T18:30:29.448Z