Morpheus

David Burns Thinks Psy­chother­apy Is a Learn­able Skill. Git Gud.

Wob­bly Table The­o­rem in Practice

• Morpheus 5 Jan 2023 14:37 UTC

  16 points

  12

  on: Run­ning by Default

  Running upstairs is also a good example (the trick is getting to the point where you can run up the stairs fast enough that you only feel you are exhausted when you are already up). Or jumping downstairs. Generally, optimizing for fun is probably the most sustainable way of doing this. I don’t enjoy most sports that require endurance, but like doing things like running in short bursts, so it works pretty well for me. The biggest downside is probably strangers looking at you weird, but I mostly don’t notice and certainly don’t mind. Walking in a group is different, of course.

[Question] What is the ev­i­dence on the Church-Tur­ing Th­e­sis?

• Morpheus 20 Nov 2021 12:00 UTC

  11 points

  on: At­tempted Gears Anal­y­sis of AGI In­ter­ven­tion Dis­cus­sion With Eliezer

  Thanks, your numbered list was very helpful in encouraging to go through the claims. Just two things that stood out to me:

  39 Nothing we can do with a safe-by-default AI like GPT-3 would be powerful enough to save the world (to ‘commit a pivotal act’), although it might be fun. In order to use an AI to save the world it needs to be powerful enough that you need to trust its alignment, which doesn’t solve your problem.

  • What exactly makes people sure that something like GPT would be safe/​unsafe?

  • If what is needed is some form of insight/​break through: Some smarter version of GPT-3 seems really useful? The idea that GPT-3 produces better poetry than me while GPT-5 could help to come up with better alignment ideas, doesn’t strongly conflict with my current view of the world?

  Worth noting that the more precise #12 is substantially more optimistic than 12 as stated explicitly here.

  #12:

  “An aligned advanced AI created by a responsible project that is hurrying where it can, but still being careful enough to maintain a success probability greater than 25%, will take the lesser of (50% longer, 2 years longer) than would an unaligned unlimited superintelligence produced by cutting all possible corners.”

  This might come across as optimistic if this was your median alignment difficulty estimate, but instead Elizier is putting 95% on this, which on the flip side suggests a 5% chance that things turn out to be easier. This seems rather in line with “Carefully aligning an AGI would at best be slow and difficult, requiring years of work, even if we did know how.”

Time un­til grad­u­a­tion as a proxy for pick­ing be­tween (Ger­man) universities

• Morpheus 22 Oct 2022 15:12 UTC

  10 points

  3

  in reply to: gjm’s comment on: Wis­dom Can­not Be Unzipped

  What’s difficult mostly isn’t unzipping wisdom, it’s identifying actual wisdom and zipping it.

  Agreed.

  Maybe this is just because of it’s age, but “a stitch in time saves nine” strikes me more as an attempt of “pretending to be wise”. If I was trying to help, I’d try communicating a procedure one can actually follow:

  • “Set a 5 minute timer to figure out the biggest problems in your life.”

  • “Consider reading book X”.

• Morpheus 10 Sep 2021 21:14 UTC

  10 points

  on: The Best Soft­ware For Every Need

  If you already know an example of the tool you are looking for, but it is not available for your operating system (or some other reason it doesn’t fit your needs perfectly), then “NiceTool [best] alternative [YourOperatingSystem]” is often a good search. Alternativeto is also pretty great for that purpose. Lots of users list up- and downsides of tools which often makes it pretty easy to figure out if a tool would fit you or not. I’ve not yet tried another one, but alternativeto itself lists PrivacyTools and Product Hunt as popular alternatives.

  If you are just searching for tips for software tools you didn’t even know you wanted them, I’d highly recommend you look at the potpourri of this amazing course on tools for CS-Students/​Programmers. I can’t recommend this course highly enough. I’ve been going back to it again and again for helpful links on commandline tools or what to do if you messed something up with git.

[Question] What Is the Idea Be­hind (Un-)Su­per­vised Learn­ing and Re­in­force­ment Learn­ing?

• Morpheus 25 Oct 2022 21:27 UTC

  9 points

  6

  on: Beyond Kol­mogorov and Shannon

  Looking forward to the rest of the sequence!

  Nitpicks:

  Note that pure prediction is not understanding. As a simple example take the case of predicting the outcomes of 100 fair coin tosses. Predicting tails every flip will give you maximum expected predictive accuracy (50%), but it is not the correct generative model for the data. Over the course of this sequence, we will come to formally understand why this is the case.

  You are just not using a proper scoring rule. If you used log or brier score, then predicting 50% head, 50% tails would indeed give the highest score.

  for i in range(data_length/3):
  	data.append(0)
  	data.append(1)
  	data.append(np.random.choice([0,1])

  Here you are cheating by using an external library. Just calling random_bits[i] is also very compact. You might at least include your pseudo-random number generator.

  conceptual downside if we want Kolmogorov complexity to measure structure in natural systems: it assigns maximal ‘complexity’ to random strings.

  Maybe I’ve read too much Jaynes recently, but I think quoting ‘complexity’ in that sentence is misplaced. Random/​pseudo-random processes really are complex. Jaynes actually uses the coin tossing example and shows how to easily cheat at coin tossing in an undetectable way (page 317). Random sounds simple, but that’s just our brain being ‘lazy’ and giving up (beware mind projection fallacy blah...). If you don’t have infinite compute (as seems to be the case in reality), you must make compromises, which is why this looks pretty promising.

• Morpheus 21 Feb 2022 19:54 UTC

  9 points

  on: Seek­ing mod­els of LW’s aver­sion to religion

  One explanation why you would not find religion discussed much on Lesswrong: The sequences are definitely anti-religion in a lot of places, so this might filter out people who are good at epistemology all else being equal, but feel insulted by Eliziers writing. Someone wanting to discuss religion on LessWrong is therefore probably a strong signal for someone clueless who “doesn’t get it”. So discussing religion here is low status.

• Morpheus 6 Mar 2024 6:02 UTC

  8 points

  1

  on: Mor­pheus’s Shortform

  While there is currently a lot of attention on assessing language models, it puzzles me that no one seems to be independently assessing the quality of different search engines and recommender systems. Shouldn’t this be easy to do? The only thing I could find related to this is this Russian site (It might be propaganda from Yandex, as it is listed as the top quality site?). Taking their “overall search quality” rating at face value does seem to support the popular hypothesis that search quality of Google has slightly deteriorated over the last 10 years (although compared to 2009-2012, quality has been basically the same according to this measure).

  

  The gpt-4 translated version of their blog states that they gave up actively maintaining this project in 2014, because search engine quality had become reasonable according to them:

  For the first time in the history of the project, we have decided to shut down one of the analyzers: SEO pressing as a phenomenon has essentially become a thing of the past, and the results of the analyzer have ceased to be interesting.

  Despite the fact that search engine optimization as an industry continues to thrive, search engine developers have made significant progress in combating the clutter of search results with specially promoted commercial results. The progress of search engines is evident to the naked eye, including in the graph of our analyzer over the entire history of measurements:

  

  SEO Pressing Analyzer Graph

  The result of the analyzer is the share of commercial sites in the search results for queries that do not have a clearly commercial meaning; when there are too many such sites in the search results, it is called susceptibility to SEO pressing. It is easy to see that a few years ago, more than half (sometimes significantly more than half) of the search results from all leading search engines consisted of sites offering specific goods or services. This is, of course, a lot: a query can have different meanings, and the search results should cater to as many of them as possible. At the same time, a level of 2-3 such sites seems decent, since a user who queries “Thailand” might easily be interested in tours to that country, and one who queries “power station” might be interested in power stations for a country home.

  If we are worried that current recommender systems are already doing damage and expect things to get worse in the future, it might be good to actively monitor this to not get frog boiled.

• Morpheus 17 Oct 2022 15:29 UTC

  8 points

  1

  in reply to: Gordon Seidoh Worley’s comment on: My re­sent­ful story of be­com­ing a med­i­cal miracle

  Now the weirdest part: drinking coconut water will make the symptoms go away, and I can basically eat gluten if I drink coconut water at the same time. My best guess is that the high levels of magnesium in coconut water are doing something here, but that’s really just a guess. On the model, somehow the coconut water probably gets any gluten bound to the villi and causing inflammation to release and out of my small intestine and maybe also does something to help the inflammation to go down.

  How on earth did you privelege the hypothesis that coconut water helps? Is that common?

[Question] Does the ex­is­tence of shared hu­man val­ues im­ply al­ign­ment is “easy”?

• Morpheus 28 Feb 2024 1:03 UTC

  7 points

  0

  in reply to: Unreal’s comment on: CFAR Take­aways: An­drew Critch

  Rationality seems to be missing an entire curriculum on “Eros” or True Desire.

  I got this curriculum from other trainings, though. There are places where it’s hugely emphasized and well-taught.

  What are these places?

• Morpheus 30 Dec 2023 0:49 UTC

  7 points

  0

  on: 2023 Unoffi­cial LessWrong Cen­sus/​Survey

  I took the survey! Fun!

• Morpheus 22 Sep 2022 15:33 UTC

  7 points

  1

  in reply to: nem’s comment on: LW Petrov Day 2022 (Mon­day, 9/​26)

  You can look at the petrov day tag. 2019 and 2021 were successful.

Re­turns on cog­ni­tion of differ­ent board games

Cop­ing with Undecidability

• Morpheus 24 Nov 2023 19:17 UTC

  6 points

  1

  on: In­su­late your ideas

  With this in mind, consider the world through the eyes of an ancient lich or thousand-year-old vampire. It’s a worldview in which ephemeral gains are irrelevant. All that matters is permanent, generalizable knowledge—everything else will fade in time, and usually not even very much time. In this worldview, gears-level understanding is everything.