Garrett Baker

• Garrett BakerMay 9, 2025, 5:28 PM

  17 points

  0

  in reply to: johnswentworth’s comment on: Ori­ent­ing Toward Wizard Power

  FYI (cc @Gram_Stone) the 2023 course website has (poor quality edit:nevermind I was accessing them wrong) video lectures.

  Edit 2: For future (or present) folks, I’ve also downloaded local mp4s of the slideshow versions of the videos here, and can share privately with those who dm, in case you want them too or the site goes down.

• Garrett BakerMay 9, 2025, 4:24 PM

  2 points

  0

  in reply to: amitlevy49’s comment on: Ori­ent­ing Toward Wizard Power

  Not really, the hypothesis is that John has depression, and of all the antidepressants, bupropion is the best (if it works for you).

• Garrett BakerMay 8, 2025, 8:13 PM

  17 points

  4

  on: Ori­ent­ing Toward Wizard Power

  A few points:

  1. All of these things sound obviously awesome and fun to do

  2. Cooking is imo another class of wizardry

  3. Have you tried the antidepressant bupropion? Ask David for the pitch

  Because of 1, I think the difficulty you’ll find building (or finding) this community is not whether or not what you’re saying “resonates” with people, but whether they have the expertise, energy, or agency to put in their share of the work.

• Garrett BakerMay 3, 2025, 5:03 PM

  7 points

  5

  on: What’s up with AI’s vision

  The boring hypothesis here is the model was actually trained on the id-location-from-picture task, and wasn’t trained on the id-object-location-in-pixel-art task, and pixel art is surprisingly nontrivial for models to wrap their heads around when they’re still trying to understand real world pictures.

• Garrett BakerMay 3, 2025, 4:57 PM

  3 points

  0

  in reply to: johnswentworth’s comment on: What’s up with AI’s vision

  My understanding is its not approximately all, it is literally all the images in geoguessr.

Lighthaven Se­quences Read­ing Group #33 (Tues­day 5/​6)

• Garrett BakerApr 29, 2025, 6:38 PM

  4 points

  0

  in reply to: johnswentworth’s comment on: john­swent­worth’s Shortform

  Apparently there already exists a CUDA-alternative for non-Nvidia hardware. The open source project ZLUDA. As far as I can tell its less performant than CUDA, and it has the same challenges as firefox does when competing with chromium based browsers, which will only get worse as it gets more popular. But its something to track at least.

• Garrett BakerApr 29, 2025, 5:53 PM

  5 points

  0

  in reply to: sam’s comment on: sam’s Shortform

  I recommend you read at least the first chapter of Getting Things Done, and do the corresponding exercises. In particular, this one, which he uses to provide evidence his model of productivity is correct

  I suggest that you write down the project or situation that is most on your mind at this moment. What most bugs you, distracts you, or interests you, or in some other way consumes a large part of your conscious attention? It may be a project or problem that is really “in your face,” something you are being pressed to handle, or a situation you feel you must deal with sooner rather than later.

  Maybe you have a holiday trip coming up that you need to make some major last-minute decisions about. You just read an e-mail about a new and pressing issue in your department. Or perhaps you just inherited six million dollars and you don’t know what to do with the cash. Whatever.

  Got it? Good. Now, describe, in a single written sentence, your intended successful outcome for this problem or situation. In other words, what would need to happen for you to check this project off as “done”? It could be as simple as “Take the Hawaii vacation,” “Handle situation with customer X,” “Resolve college situation with Susan,” “Clarify new divisional management structure,” “Implement new investment strategy,” or “Research options for dealing with Manuel’s reading issue.” All clear? Great.

  Now write down the very next physical action required to move the situation forward. If you had nothing else to do in your life but get closure on this, what visible action would you take right now? Would you call or text someone? Write an e-mail? Take pen and paper and brainstorm about it? Surf the Web for data? Buy nails at the hardware store? Talk about it face-to-face with your partner, your assistant, your attorney, or your boss? What?

  Got the answer to that? Good.

  Was there any value for you in those two minutes of thinking? If you’re like the vast majority of people who complete that drill in our seminars, you’ll be experiencing at least a tiny bit of enhanced control, relaxation, and focus. You’ll also be feeling more motivated to actually do something about that situation you’ve merely been thinking about till now. Imagine that motivation magnified a thousandfold, as a way to live and work.

  If anything at all positive happened for you in this little exercise, think about this: What changed? What happened to create that improved condition within your own experience? The situation itself is no further along, at least in the physical world. It’s certainly not finished yet. What probably happened is that you acquired a clearer definition of the outcome desired and the next action required. What did change is the most important element for clarity, focus, and peace of mind: how you are engaged with your world.

  But what created that? Not “getting organized” or “setting priorities.” The answer is, thinking. Not a lot; just enough to solidify your commitment about a discrete pressure or opportunity and the resources required dealing with it. People think a lot, but most of that thinking is of a problem, project, or situation—not about it. If you actually did this suggested exercise, you were required to structure your thinking toward an outcome and an action, and that does not usually happen without a consciously focused effort. Reacting is automatic, but thinking is not.

• Garrett BakerApr 28, 2025, 6:04 PM

  4 points

  2

  in reply to: Wei Dai’s comment on: Wei Dai’s Shortform

  This argument seems only convincing if you don’t have those destructive values. One man’s destructive values is another’s low-hanging fruit, and those who see low hanging fruit everywhere won’t give up on the fruit just because others may pick it.

  Since bad people won’t heed your warning it doesn’t seem in good people’s interests to heed it either.

  An analogy is one can make the same argument wrt rationality itself. Its dual use! Someone with bad values can use rationality to do a lot of harm! Does that mean good people shouldn’t use rationality? No!

• Garrett BakerApr 28, 2025, 5:17 AM

  4 points

  0

  on: Our Real­ity: A Si­mu­la­tion Run by a Paper­clip Maximizer

  Yet the universe runs on strikingly simple math (relativity, quantum mechanics); such elegance is exactly what an efficient simulation would use. Physics is unreasonably effective, reducing the computational cost of the simulation. This cuts against the last point.

  This does not seem so consistent, and if the primary piece of evidence for me against such simulation arguments. I would imagine simulations targeting, eg, a particular purpose would have their physics tailored to that purpose much more than ours seems to (for any purpose, given the vast computational complexity of our physics, and the vast number of objects such a physics engine needs to keep track of). For example, I’d expect most simulations physics to look more like Greg Egan’s Crystal Nights (incidentally this story is what first convinced me the simulation hypothesis was false).

  One may argue its all there just to convince us we’re not in a simulation. Perhaps, but two points:

  1. Given the discourse on the simulation hypothesis, most seem to take our physics as evidence in favor of it, as you do here. So I don’t think most think clearly enough about this for our civilizational decisions to be so dependent on this.

  2. The simulators will have trade-offs and resource constraints too. Perhaps they simulate few highly detailed simulations, and many highly simplified simulations. If this is exponential, in the sense that as the detail decreases the number of simulations exponentially increases, we should expect to be in the least detailed world consistent with the existence of sentiences and for which its not blatantly obvious we’re in a simulation.

  Of course this argument would break given sufficiently different physics from ours, enabling perhaps our world to be simulated in as much depth as it is very cheaply. But then that seems intuitively at least very unlikely & complex a hypothesis.

• Garrett BakerApr 27, 2025, 11:01 PM

  9 points

  0

  in reply to: JustisMills’s comment on: AI Self Por­traits Aren’t Accurate

  Yeah I think I agree with all of this, so I do think most of this was miscommunication/​interpretation.

  the combativeness did also make me a little sad.

  Sorry about that, I think my comments often come across as more negative than I intend, I try to remember to take a step back afterwards and rewrite things to be nicer, but I often forget or don’t realize in the moment its necessary.

• Garrett BakerApr 27, 2025, 5:57 PM

  4 points

  6

  on: AI Self Por­traits Aren’t Accurate

  If LLMs can be sad, that sadness would probably be realized through the firing of “sadness” features: identifiable patterns in its inference that preferentially fire when sad stuff is under discussion. In fact, it’s hard to say what else would count as an LLM experiencing sadness, since the only cognition that LLMs perform is through huge numbers of matrix operations, and certain outcomes within those operations reliably adjust the emotional content of the response.

  Best I can tell, your argument here is “either there’s a direction in activation space representing sadness and this is what ‘sadness’ is or something else is going on, and I can’t think of anything else, so the first thing must be true, if any sadness is going on at all”.

  Suffice it to say, Reality has never played very well with those trying to make arguments from their lack of imagination, and I think you need to do much, much more work if you want this argument to have any sway.

• Garrett BakerApr 26, 2025, 8:58 PM

  5 points

  2

  in reply to: Ben Goldhaber’s comment on: bgold’s Shortform

  Your second option seems likely. Eg did you know community notes is open source? Given that information, are you going to even read the associated whitepaper or the issues page?

  Even if you do, I think we can still confidently infer very few others reading this will (I know I’m not).

• Garrett BakerApr 26, 2025, 3:19 PM

  2 points

  0

  in reply to: Linch’s comment on: Why Have Sen­tence Lengths De­creased?

  Smol r'\. (.*)\.'! y? Clear! Had big r'\. (.*)\.' b4. & abstract. ppl no get. Now: Smol. Clear.

  Good. Think clearer. y? Smol ⇒ deep.

• Garrett BakerApr 26, 2025, 3:04 PM

  3 points

  5

  in reply to: faul_sname’s comment on: faul_sname’s Shortform

  A possible longer term issue with this is when future generations of models are pre-trained, this style of code will be a significant fraction of their training data which will only grow over time, so just as its been hard to get models out of the “chatgpt-ese” due to simulators reasons, it may also be hard to get models out of this messy code basin, even before you do any code RL, once they realize their chat-models and they’re “supposed to” talk like this.

  I say issue, because it does seem worse to have a trend in the direction of AI code un-readability by humans have some momentum behind it, rather than just being a result of easily changeable RL fine-tuning.

[LAPTOP REQUIRED] Lighthaven Se­quences Read­ing Group #32 (Tues­day 04/​29)

• Garrett Baker 24 Apr 2025 22:50 UTC

  5 points

  3

  in reply to: Neel Nanda’s comment on: MichaelDick­ens’s Shortform

  In that case I think your response is a non sequitur, since clearly “really care” in this context means “determiners of what they end up doing in practice re influencing x-risk”.

• Garrett Baker 23 Apr 2025 2:19 UTC

  2 points

  2

  in reply to: johnswentworth’s comment on: john­swent­worth’s Shortform

  Conjecture seems unusually good at sticking to reality across multiple domains.

  I do not get this impression, why do you say this?

• Garrett Baker 22 Apr 2025 21:44 UTC

  3 points

  1

  in reply to: Tenoke’s comment on: Eri­oirE’s Shortform

  In this case prediction markets will be predictably over-optimistic, and expert consensus is very split.

• Garrett Baker 21 Apr 2025 17:28 UTC

  3 points

  2

  on: Re­search Notes: Run­ning Claude 3.7, Gem­ini 2.5 Pro, and o3 on Poké­mon Red

  There is a question of interest here though: why does pixel art work so well on humans despite literally nothing in real life being pixel art?

  I’m reminded of Gwern’s comments on the difficulty of getting GANs to generate novel pixel art interpolations

  Pixel-art anything is derivative of a photorealistic world. If you look at 8-bit art and standard block sizes like Mario in NES Mario, if you were not already intimately familiar with the distribution of human faces, and had to learn starting with a completely blank slate like a GAN would, how would you ever understand what a human face was such that you could imagine correct variations like Princess Peach or Luigi? Or if you wanted to generate Pokemon, which are all based heavily on real animals, how would the model know anything about horses or zebras or hamsters or butterflies and be able to generate a sprite of Butterfree independently? If you look at the Pokemon GAN failure cases closer and compare them to ‘real’ Pokemon, you start to realize to what an extent each Pokemon is derivative of several real-world animals or objects—Pokemon in some ways do not exist in their own right, they are only shorthand or mnemonics of other things. Pikachu is the “electric mouse”: but if you had never seen any electricity iconography like ‘lightning bolts’ or any rodents like hamsters or chinchilla or jerboa, how could you ever understand an image of a ‘Pikachu’ or generate a plausible rodent variation of it? If you could, you’d need a lot more Pikachu training data, that’s for sure. (One is reminded of the joke about the mathematicians telling jokes. They knew all each other’s favorites, you see, so they only needed to call out the number. “#9.” Sensible chuckles. “#81.” Laughter. “#17!” Chortling all around. The new grad student ventures his own joke: “…#112?” Outright hysteria! You see, they had never heard that one before…)