@self_made_human's banner p

self_made_human

amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi

16 followers   follows 0 users  
joined 2022 September 05 05:31:00 UTC

I'm a transhumanist doctor. In a better world, I wouldn't need to add that as a qualifier to plain old "doctor". It would be taken as granted for someone in the profession of saving lives.

At any rate, I intend to live forever or die trying. See you at Heat Death!

Friends:

A friend to everyone is a friend to no one.


				

User ID: 454

self_made_human

amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi

16 followers   follows 0 users   joined 2022 September 05 05:31:00 UTC

					

I'm a transhumanist doctor. In a better world, I wouldn't need to add that as a qualifier to plain old "doctor". It would be taken as granted for someone in the profession of saving lives.

At any rate, I intend to live forever or die trying. See you at Heat Death!

Friends:

A friend to everyone is a friend to no one.


					

User ID: 454

Funnily enough, I have an AAQC on the Dodo Bird model.

It seems like the most parsimonious explanation, but I would say that it doesn't disqualify therapy as a valid therapeutic intervention. A lot of the people being sent to therapy do not have access to a discreet, thoughtful friend who will keep secrets. That might well be a service worth paying for. What isn't in dispute is that therapy works in the first place, even the models that use bonkers frameworks.

I see no reason LLMs can't make for okay therapists, and they are definitely better than the quality of some I have personally met.

At the end of the day, I'm just glad that therapy isn't the only tool in my arsenal, and I can dish out the fun drugs. Psychologists are so painfully restricted in what they can do.

The paper seems to have been published on April 2025.

Gemini 2.0 Pro and 3.7 Sonnet came out in February 2025. Claude 3.5 Sonnet came out in June 2024 and was better than the version of 4o out then.

At the very least, the authors should have made a note that they weren't using the SOTA, or that the SOTA would have moved significantly by the time of publication. To do less is mild dishonesty. This isn't 2022, the pace of progress is evident.

4o is also what powers chatgpt.com so it's the model that most casual users will get the output from.

True, but that's OAI being cheap, and not an indictment of the utility of LLMs for translation. It's akin to claiming TVs suck, and then only using a cheap and cheerful $300 model from Walmart as the standard.

My criticisms stand, namely that LLMs only get better, they're "good enough", and that this is a net improvement over the status quo. It remains to be seen how much better the SOTA is over 4o or DeepL.

Huh. I was confident that I had a better writeup about why "stochastic parrots" are a laughable idea, at least as a description for LLMs. But no, after getting a minor headache figuring out the search operators here, it turns out that's all I've written on the topic.

I guess I never bothered because it's a Gary Marcus-tier critique, and anyone using it loses about 20 IQ points in my estimation.

But I guess now is as good a time as any? In short, it is a pithy, evocative critique that makes no sense.

LLMs are not inherently stochastic. They have a (not usually exposed to end-user except via API) setting called temperature. Without going into how that works, it suffices it to say that by setting the value to zero, their output becomes deterministic. The exact same prompt gives the exact same output.

The reason why temperature isn't just set to zero all the time is because the ability to choose something other than the next most likely token has benefits when it comes to creativity. At the very least it saves you from getting stuck with the same subpar result.

Alas, this means that LLMs aren't stochastic parrots. Minus the stochasticity, are they just "parrots"? Anyone thinking this is on crack, since Polly won't debug your Python no matter how many crackers you feed her.

If LLMs were merely interpolating between memorized n-grams or "stitching together" text, their performance would be bounded by the literal contents of their training data. They would excel at retrieving facts and mimicking styles present in the corpus, but would fail catastrophically at any task requiring genuine abstraction or generalization to novel domains. This is not what we observe.

Let’s get specific. The “parrot” model implies the following:

  1. LLMs can only repeat (paraphrase, interpolate, or permute) what they have seen.

  2. They lack generalization, abstraction, or true reasoning.

  3. They are, in essence, Markov chains with steroids.

To disprove any of those claims, just gestures angrily look at the things they can do. If winning gold in the latest IMO is something a "stochastic parrot" can pull off, then well, the only valid takeaway is that the damn parrot is smarter than we thought. Definitely smarter than the people who use the phrase unironically.

The inventors of the phrase, Bender & Koller gave two toy “gotchas” that they claimed no pure language model could ever solve: (1) a short vignette about a bear chasing a hiker, and (2) the spelled-out arithmetic prompt “Three plus five equals”. GPT-3 solved both within a year. The response? Crickets, followed by goal-post shifting: “Well, it must have memorized those exact patterns.” But the bear prompt isn’t in any training set at scale, and GPT-3 could generalize the schema to new animals, new hazards, and new resolutions. Memorization is a finite resource but generalization is not.

(I hope everyone here recalls that GPT-3 is ancient now)

On point 2: Consider the IMO example. Or better yet, come up with a rigorous definition of reasoning by which we can differentiate a human from an LLM. It's all word games, or word salad.

On 3: Just a few weeks back, I was trying to better understand the actual difference between a Markov Chain and an LLM, and I had asked o3 if it wasn't possible to approximate the latter with the former. After all, I wondered, if MCs only consider the previous unit (usually words, or a few words/n-gram), then couldn't we just train the MC to output the next word conditioned on every word that came before? The answer was yes, but that this was completely computationally intractable. The fact that we can run LLMs on something smaller than a Matrioshka brain is because of their autoregressive nature, and the brilliance of the transformer architecture/attention mechanism.

Overall, even the steelman interpretation of the parrot analogy is only as helpful as this meme, which I have helpfully appended below. It is a bankrupt notion, a thought-terminating cliché at best, and I wouldn't cry if anyone using it meets a tiger outside the confines of a cage.

/images/17544215520465958.webp

Does he? Wouldn't surprise me, but I think we need weebs subject matter experts to disambiguate on our behalf.

I'm reading the paper, but initial issues that caught my eye:

  1. They're not evaluating GPT-4. They're using 4o. The exact implementation details of 4o are still opaque, it might be a distill of 4, but they're not the same model. As far as I can tell, that's a point of confusion on your part, not the authors.

  2. 4o, even at the time of publication, was not the best model available. Very far from it. It is a decent generalist model, but not SOTA. It isn't the best model, the second best model, or even the third... best model, on almost any metric one opts for.

I have, as far as I'm aware, never claimed that LLMs match or outperform professional human translators. My core contention was that even basic bitch LLMs are good enough, and an improvement over previous SOTA, including DeepL, which this paper supports.

This would hold true even if the authors had chosen to use something like o3, Opus 4, Gemini 2.5 Pro etc. It is unclear to me if they were aware that better options were unavailable, there's little reason to use 4o if one wants to know what the best possible output is.

And even if it is true, it doesn't matter. The models are consistently getting better. We have a new SOTA every few months these days.

  1. This is a very bad way to showcase ChatGPT's writing chops. When I use it, by virtue of personally having native fluency, I am far better equipped to critique its output. The original user, a Japanese person somewhat conversational in the language, is simply not in a position to be able to tell where it could have made better stylistic choices. Note that my objection above is to do with the fact that the availability of ChatGPT represents a massive improvement over the previous status-quo. If he had simply asked it to translate his original Japanese, it would have done a perfectly fine job.

  2. LLMs are not a monolithic entity. When most people talk about ChatGPT, they're ignorant of the fact that there are options beyond the slop-tier 4o. You may or may not know the difference, but the average normie complaining doesn't. My own workflow involves using multiple different models, multiple attempts, prompts that I know qualitatively improve results and my own ability to discern between outputs.

For example, when I ask for a translation that preserves the writers' rather unique character:

Kimi K2:

Today's stream was absolutely perfect! The second I dropped the comment "Please step on me, Step-on-me!" my oshi, Haachama herself, shot back with "Gross!"—and I was already soaring. Then she hit me with "You're such a perv!" and I just about ascended straight to heaven. I'm floating in pure bliss right now; this is the most peaceful moment of my entire life. And the thing that's got me buzzing more than anything? Haachama's birthday live is going down on Sunday, August 10 at 9:00 PM! I'm ready to cheer my lungs out. Just picturing that day makes me feel like I'm literally sipping her bathwater—yeah, that level of hype!

Gemini 2.5 Pro:

Tonight's stream was PERFECT! I commented, "Funde kuda-STEP ON ME!"* and my oshi, Haachama, actually responded with "Kimoi!" ("Gross!"). And then she even added, "You're such a perv!" THIS IS INSANE!! I feel like I'm in heaven right now. This is the most peaceful moment of my entire life. And to top it all off, the thing I'm looking forward to most is Haachama's birthday live on Sunday, August 10th at 21:00!! I'm gonna support her with everything I've got. Just imagining that day's stream makes me feel like I'm drinking her bathwater.

What a sterling example of making the dream of perfection the sworn enemy of the merely better. As others have pointed out before, the most likely alternative, in the absence of ChatGPT, would have been this poor fellow resorting to Google Translate or other, far simpler ML solutions. There isn't an abundance of fluent English and Japanese speakers willing to proof read random YouTube comments.

I don't speak Japanese, but I see nothing particularly objectionable in the translation. It might not capture all nuance, but it gets the gist of it across. Learning language takes time, probably years, and by the time this gentleman gets good enough that he needs or appreciates the nuance, LLMs will be even better at the job.

When you told me you're fluent in Japanese the other day, this was really not how I expected it to become relevant haha.

"Today's stream was perfect! When I commented "Please step on me!" my oshi Haachama reacted by saying "Gross!" and then she even said "You're way too much of a pervert!" - it was incredible!! I'm feeling like I'm in heaven right now. This is the most blissful moment of my life. And what I'm most excited about is that Haachama's birthday live stream is on August 10th (Sunday) at 9:00 PM!! I want to support her with everything I've got. Just imagining that day and the live stream makes me feel like I'm drinking her bathwater."

Claude knocks it out of the park as far as I'm concerned.

Even 4o:

Today's stream was absolutely perfect! When I commented, "Step on me, please!" my oshi, Haachama, actually reacted and said, "Gross!" Then she went on to say, "You're way too much of a pervert!" It was insane!! Right now, I feel like I'm in a sexy kind of heaven. This is the most peaceful moment of my life.

And on top of all that, the thing I'm most looking forward to is Haachama’s birthday live on Sunday, August 10th at 9:00 PM!! I'm planning to cheer her on with everything I've got. Just imagining that live makes me feel like I'm drinking her bathwater.

Fun stuff, but that's really not post-singularity. It's not even post-scarcity!

My guy, you asked me to go easy on Turok last time. Look how he repaid your charity.

If there's anything useful to be said on class resentment, you won't find Turok saying it.

Uh.. It's fine? I'm genuinely okay with "hard" scifi having speculative elements. My original objection was solely that Avatar represents harder scifi than Vinge's work. Nothing in Avatar outright breaks the laws of physics as we know them. This isn't a particularly big deal, since speculating on future advances in physics and engineering is part of the appeal of science fiction in general.

This is the opposite of an effort post, but I'm fond of La Vieille Ferme Rosé. It blew up on TikTok as "Chicken Wine" (a fact relayed to me by an ex), and I think it tastes great for something that costs £9 at the local supermarket.

(I have no desire to develop expensive tastes)

Strictly speaking, I'd be happy with either. If I have to choose, the one who's actually my age. More longterm potential there.

Wait, do I know who is who?

I mostly agree, but I would assume that evolution incentivized attraction to total fertility. If you shack up with a woman in her late teens, you can, at least in theory, get many more kids out of her than you can in the late 20s. We live in a very unusual period of time, when we can take negligible infant or childhood mortality for granted.

The state of the art of AI in Star Wars isn't much better than today. Funnily enough, I recall that there was minor plot point, where a malevolent bounty hunter droid of minor infamy ended up seizing control of the Death Star's mainframe, and began growing into a superintelligence.

Unfortunately it had very bad timing, as mere minutes later the place got blown up in a spectacular fashion.

So much of the Star Wars franchise doesn't plain make sense. It's soft science fantasy at best, and even I'm not shameless or contrarian enough to earnestly make that claim.

I understand how that makes sense in-universe, but my objection is that makes no sense in the real world. I see no viable mechanism by which a real ASI could pull that off IRL, without simply forcing everyone into a simulation it controls. As I've said in a reply to @TitaniumButterfly, not even God can make 2+2=5.

It is a good conceit for a story, but it doesn't apply to reality.

My objection is that I strongly doubt that even a superintelligence can enact such a sweeping change to the laws of physics such that it could meaningfully constrain the ability of different levels of something as vaguely defined as "intelligence" within the galaxy. The only remotely feasible way I can envision to do this would be to create a universe from scratch, or at least run a simulation where you have utter authority.

Superintelligence != omnipotence, even if they can be ridiculously powerful.

Huh. I mean, at that point, you might as well say that a superintelligence can make 2+2=5!

The only way I can see this working is if it took control of the lightcone, and then forced everyone else into a simulation where it had perfect control.

The kind of "exclusive" ephebophilia you mention seems vanishingly rare to me. Often the kind of men opting to go for much younger women are doing so both because they like them, and because they would have a harder time with the older ones who aren't so tolerant of them being weird/lazy/unsuccessful. Think aging musician or bartender with no other credentials, they can convince a college sophomore that they're super cool, but don't appeal to women in their age group looking for something serious. That is rational from their perspective, they're getting laid at the end of the day. There are rare exceptions like Leonardo Di Caprio, who can afford to make a break for it when they cross 25, as he can easily afford to be picky.

At the end of the day, men tend to prefer youth and correlates for fertility. They might even have preferences that their partners be dependent on them, which describes so many goddamn people that I wouldn't call it pathological. Men were the breadwinners for most of recorded history.

I am being performatively lazy here. Turok genuinely isn't worth my time, and I'm confident that almost all of our regulars are well aware of his bad behavior. That being said, I appreciate you sharing the link to previous warnings.

For what it's worth, I don't see anything wrong in continuing a thread in a new CWR. Most users would prefer more engagement or at least eyeballs on their posts, and once the thread becomes obsolete, it's very unlikely that a significant number of people will even read anything you have to say.

Was it? I don't seem to recall that being the case. And even if that's the nominal explanation, there's no way in hell it would work IRL.

I think even the label of "ephebophile" is an artificial one. It probably encompasses almost all men. The average man cannot tell, at least by looking, if someone is 16 years old or 18 and a day. If they've got tits and a nice figure, they'll make just about everyone sexually attracted regardless of age. The only difference worth noting is that, at least in the West, there has been so much social conditioning that people are loathe to accept this fact, and those who do take the risk knowingly likely have other issues.

Back when OkCupid's blog wasn't a joke, they shared a chart showing response rates for women by age. The curve peaked at 18, and declined after. I am convinced that if the platform allowed younger women to sign up, then the actual modal value would be ~16 or 17. We have successfully pathologized very natural behavior, but oh well. I'm lucky enough to have a thing for MILFs, albeit they're just women my age these days.

Thank you for the tip! Dover took a lot out of me, but I should be down in England for a while yet and might find the time to see the Sisters (if they're single). If not, then I'll bookmark it for another date.