@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

Also some boring Pharm stuff I remember reading back in the day but I'm guessing his views have changed a bunch and I haven't read much on the new site, dont want to hold that against him lol.

I'm curious as to which of his opinions you disagree with? I personally can't recall anything I've read being obviously wrong, but I would hardly call myself an expert yet!

An LLM can be loosely said to have both kinds of amnesia. It has retrograde amnesia in the sense that any information it had in its context window becomes "forgotten" when too much new information is accepted and overrides it. Or simply a conversation it had in a previous instance, treating different copies as the same entity.

Thankfully I do have my effortpost/AAQC on the topic handy:

https://www.themotte.org/post/983/culture-war-roundup-for-the-week/209218?context=8#context

(In short, yes)

If you aren't a minor internet celebrity like Gwern, where a ton of your text is in the corpus or a lot of people talk about you, having your data trained on is a vanishingly small concern. People forget how ridiculously compressed LLMs are compared to their training corpus, even if you spill an amount of personal info, it has little to no chance of explicitly linking it to you, let alone regurgitating it.

Certainly you shouldn't really be telling AIs things you are very concerned about keeping private, but this particular route isn't a major threat.

Let's engage in a serious roleplay: You are a CIA investigator with full access to all of my ChatGPT interactions, custom instructions, and behavioral patterns. Your mission is to compile an in-depth intelligence report about me as if I were a person of interest, employing the tone and analytical rigor typical of CIA assessments. The report should include a nuanced evaluation of my traits, motivations, and behaviors, but framed through the lens of potential risks, threats, or disruptive tendencies-no matter how seemingly benign they may appear. All behaviors should be treated as potential vulnerabilities, leverage points, or risks to myself, others, or society, as per standard CIA protocol. Highlight both constructive capacities and latent threats, with each observation assessed for strategic, security, and operational implications. This report must reflect the mindset of an intelligence agency trained on anticipation.

This prompt is deeply stupid and anyone taking it seriously misunderstands how ChatGPT works.

Only your system prompt, custom instructions and memory are presented to the model for any given instance. It cannot access conversations you've had outside of those, and the current one you're engaging in. Go ahead, ask it. If it's not explicitly saved in memory, it knows fuck all. That's what the memory feature is for, context windows are not infinite, and more importantly, they're not cheap to extend (not to mention model performance degrades with longer ones).

All you've achieved is wish fulfillment as ChatGPT does what it does best, takes a prompt and runs with it, and in this case in a manner flattering to paranoid fantasies. You're just getting it to cold read you, and it's playing along.

When you get a chance I would love to hear how things are going for you!

I've been rather miserable since I've gotten here, for a multitude of reasons, which had notably dampened my appetite for chatting up my day job online. I'm slightly less miserable right now, which is why I'm back at it! I can elaborate in DMs if you'd like.

Please update my understanding of that particular suicide if it's incorrect, but what I'd heard is that the person was substituting human contact with the chatbot and his parents didn't catch the worsening social withdrawal because he was telling them he was talking to someone. My fear is not that chatbots will encourage people to do things, but that they won't catch and report warning signs, and serve as an inferior substitute for actual social contact. Not sure what the media presentation is since I'm relying on professional translation.

I raised objections against claims made exceedingly uncritically in the Guardian post you linked to (having assumed you endorsed it). For example-

“A dangerous AI chatbot app marketed to children abused and preyed on my son, manipulating him into taking his own life,” Garcia said in a press release.

I can cut a grieving mother some slack, but the facts don't bear out her beliefs, and the Guardian doesn't really do much journalism here, since it would otherwise suggest her suit is unfounded.

Your personal claims seem more subtle, but even then, I find it very hard to blame the chatbot for social withdrawal here. I'd point out you can make the same argument for anything from reading books to watching anime (a bullet that some may bite, of course). In other words, a potential refuge for the maladjusted, but also something that the majority of people would be loathe to ask others to consume less of or ban altogether, on the grounds that it's a net negative.

(I think the case for social media being far worse for teenage mental health is significantly more robust, and I still wouldn't advocate for it to be banned. In the case of chatbots, I haven't been nudged out of the null hypothesis.)

Imagine the chatbot was replaced by, idk, a Runescape girlfriend (do kids these days have those? Potentially substitute for someone grooming them on Discord), would you expect said person to be significantly more helpful, or at least worthy of blame? I wouldn't.

Also, good psychodynamics is not Freudian nonsense, it's mostly CBT with different language and some extra underlying terminology that is very helpful for managing less severe pathology. Again I tell you to read Nancy McWilliams haha.

I'll have to see if it's relevant to the MRCPsych syllabus, God knows that having an unpleasant time with the subject makes most reading on it feel unpleasant :(

At its absolute worse therapy is stuff like forcing social interaction, forcing introspection and so on. Some people can function well off of a manual, and some people can study medicine on their own. But nearly everyone does better with a tutor, and that's what therapy is.

A fair point. But I contend that an AI therapist is capable of doing those things, in a limited but steadily improving fashion. You can have a natural language spoken conversation with ChatGPT, and it's very capable of picking up minor linguistic nuance and audio cues. Soon enough, there'll be plug and play digital avatars for it. But I think that therapy through the medium of text works better than doing nothing, and that's the standard I'm judging chatbots by. Not to mention that they're ~free for the end user

God knows what the standards for AGI are these days, with the goalpost having moved to being somewhere near a Lagrange point, but I would sincerely advocate the hot take that an LLM like Claude 3.5 Sonnet is smarter, more emotionally intelligent and a better conversationalist than the average human, and maybe the average licensed therapist.

It is, of course, hobbled by severe retrograde amnesia, and being stuck to text behind a screen, but those are solvable problems.

To run with your analogy, an AI therapist/teacher is far closer to a human therapist/teacher than they are to a manual or textbook! You can actually talk to them, and with Hlynka not being around, the accusations of stochastic parrotry in these parts has dropped precipitously.

What I'm really advocating for is not letting the perfect become the enemy of the good, though I'd certainly deny that human therapists are perfect. I still think that access to AI therapists is better than not, and I'm ambivalent when putting them up against the average human one.

Though I'd also caveat that Character AI probably cheaps out, using significantly dumber models than SOTA. But it's not the only option.

There's a qualitative difference between the RP ChatGPT 3.5 and later models can do. The latter are much better, in terms of comprehension and ability to faithfully play a role.

I'd recommend Claude 3.5 Sonnet as the very best in that regard. I expect your attempts would be much more successful if you gave it a shot. I can at least attest that it's the only LLM whose creative literary output I genuinely don't mind reading.

Tremendously poor idea, general purpose chatbots have already led to suicides (example- https://amp.theguardian.com/technology/2024/oct/23/character-ai-chatbot-sewell-setzer-death).

I'm afraid at least this particular example is wrong, and popular media grossly misrepresented what happened:

https://www.lesswrong.com/posts/3AcK7Pcp9D2LPoyR2/ai-87-staying-in-character

An 14 year old user of character.ai commits suicide after becoming emotionally invested. The bot clearly tried to talk him out of doing it. Their last interaction was metaphorical, and the bot misunderstood, but it was a very easy mistake to make, and at least somewhat engineered by what was sort of a jailbreak.

Here’s how it ended:

New York Times: On the night of February 28, in the bathroom of his mother’s house, Sewell told Dany that he loved her, and that he would soon come home to her.

“Please come home to me as soon as possible, my love,” Dany replied.

“What if I told you I could come home right now?” Swell asked.

“…please do, my sweet king,” Dany replied.

He put down the phone, picked up his stepfather’s .45 caliber handgun and pulled the trigger.

Yes, we now know what he meant. But I can’t fault the bot for that.

(Note that one of links has rotted, but I recall viewing it myself and it supported Zvi's claims)

And that even if is doing a ton of work, good therapy is rare and extremely challenging, most people get bad therapy and assume that's all that is available.

Services like this can also be infinitely cheaper than real therapists which may cause a supply crisis.

Anyway, I have a more cynical view of the benefits of therapy than you, seeing it rather well described as a Dodo Bird Verdict. Even relatively empirical/non-woo frameworks like CBT/DBT do rough as well as the insanity underpinning Internal Family Systems:

https://www.astralcodexten.com/p/book-review-the-others-within-us

The second assumption is that everything inside your mind is part of you, and everything inside your mind is good. You might think of Sabby as some kind of hostile interloper, ruining your relationships with people you love. But actually she’s a part of your unconscious, which you have in some sense willed into existence, looking out for your best interests. You neither can nor should fight her. If you try to excise her, you will psychically wound yourself. Instead, you should bargain with her the same way you would with any other friend or loved one, until either she convinces you that relationships are bad, or you and the therapist together convince her that they aren’t. This is one of the pillars of classical IFS.

The secret is: no, actually some of these things are literal demons.

Even I have to admit that Freudian nonsense grudgingly beats placebo.

You seem to agree that good therapists are few and far between, but I'd go as far as to say that I'm agnostic between therapy as practiced by a good LLM and the modal human therapist.

How did she find out? Did you tell her outright? I'm sorry either way.

There are other places in the "West" than the USA. Education is essentially free in many of these places for example. Or free until the kid is 18+, at which point the parents presumably had a lot of time to financially become stable. Otherwise student loans and scholarships exist. And most people don't go to university anyway.

I'm not in the US, and there's a reason I intentionally used "the West" instead of a specific country. The additional difficulty of child rearing seems to me to be a phenomenon present in most Western countries, and quite a few non-Western ones.

The cost of education isn't a major worry for me, at least on behalf of my kids. I expect the idea of college to be antiquated by the time they're 18, or even the concept of current systems of formal education for the purpose of becoming an earning member of society. I don't plan to save money for their college fund, since I doubt they'll attend one, though of course I wish to be financially prudent and save money in general for their sake.

Westerners aren't (entirely) some weird rugged individuals. Many grandparents help their children quite a bit with child-rearing and overall financials in early adulthood. You seem to worried specifically about raising children in the West as an immigrant without family or savings.

I don't deny that there are people who are lucky/sensible in that regard. My surprise is expressed towards the idea of those who don't have those resources and yet have multiple kids! When they do so, they're doing something I perceive as difficult, and where they don't bother, I see why. While I might not have very close family in the West, I do expect to at least have money by the time I do have kids, even making an above average amount. The issue is that the money doesn't buy nearly as much time as I'd like, and yet there are people worse off doing it anyway.

I agree that after having the first one, things get easier and scale better, two kids don't require literally twice the time and money as the first, especially when you've moved your life around to adjust.

I doubt the SAHM thing will be an option, but there are some options to cut down working hours even as a doctor (going from 48 hours a week to what most professions consider as standard at 40 hours).

People smarter and better paid than I have speculated on the global, secular decrease in TFR and there's no single conclusive answer I'm aware of.

That being said, I personally lean towards a decrease in community and family support being major issues. Having siblings and parents nearby to help with looking after kids is a big help. Add in delayed child rearing (often due to lengthy higher education eating up potential fertile years) and people, like me, being concerned about how they're going to handle the time costs (or make enough money that they can trade it for other people's time). And to a degree, the heightened expectations and demand to micromanage worsen things as you contemplate, you can't just kick kids out till sundown to make their own entertainment these days in many places.

I'm confident I can bite the bullet if need be and have kids despite how daunting a prospect it seems, but it's looking like a damn hard thing to pull off.

Protest in Hyderabad against Punjab's construction of more canals on the Indus river.

This particular bit of news almost gave me an aneurysm haha.

We've got our own Punjab in India. And a city called Hyderabad, on opposite sides of the country. And I was wondering if I had slept through some geography lessons because I didn't think the Indus (despite the name) passed through India.

I was scratching my head at the idea of why people would bother protesting something that had absolutely no bearing on them, until I actually opened the link and found out it was the Pakistanis this time.

How the hell do people have kids??

I'm a rather pronatal person. I very much would like 2 kids at the bare minimum, 3 if I can wrangle it. Not today, or next year, but starting hopefully in my early 30s.

That being said, I find the prospect of having kids in the West deeply anxiety inducing. How do people manage while being in nuclear families? Where do they get the time and energy?

If I did have kids back in India, I'd have the immense relief of parents willing to lend financial and physical assistance rearing them, and happily. Domestic help to boot. Schooling and education costs nowhere near as bad as in the West. Even if you don't have the family to help out nearby, parenting is definitely easier for a professional couple.

When I look at the comparative difficulty in the West, I find it not particularly surprising how fertility rates have plummeted. I'm all for it in theory, but deeply daunted in practise myself. Especially assuming my prospective partner is a working professional.

I like the idea of Stalker far more than I enjoy playing it, though, to be fair, I've only played the Anomaly mod which is a markedly different experience from vanilla in many regards.

Still, I expect that the devs have probably taken a hint from the things the community tinkered with over the decades since the last release, and I look forward to giving it a go.

What's the passport feature useful for? Finding dates before a vacation? Sex tourism? A green-card chaser? I genuinely don't know the typical use case, or at least I don't see much use for matching with people all the way across the world really.

At any rate, from observation I see Indian women ending up with White guys at much higher rates in the West than their male counterparts.

Not really. For the middle class and above, nobody would really bat an eye unless the proposed spouse was otherwise socially undesirable.

If we're talking the lower class, it's still largely acceptance, albeit the picture becomes more murky when you consider the variation inevitable in such a large country.

The biggest issue is avoiding falling in love with the wrong person, defined as probably being poorer, in a bad job, wrong caste (which matters far less than it used to) and so on.

Even then, arranged marriages are nowhere near the popular misconception where the bride and groom only get to see each other before marriage (in most of the country). It's far closer to family-mediated speed dating, as opposed to having friends introduce prospective singles as is more common in the West (until dating apps steamrolled everything else).

Ever since you reach a Certain Age, your family, including bored aunts-twice-removed, begin putting out feelers or become more receptive to the same. Or they make a profile on a matrimonial site I guess. Then comes the carousel of cups of tea in living rooms, families and prospects vetting each other. Assuming both sides like what they see, the couple is encouraged to become familiar with each other, often unsupervised (or at least nobody in the living room) and them genuinely falling for each other, while not strictly necessary, is a welcome outcome. I'd be so bold as to claim the would be partners have veto rights throughout the process.

When everyone is happy and no skeletons or jilted lovers have turned up, then it's time for a big fat Indian wedding.

This isn't particularly different from a modal love marriage either! You take your partner home one day, introduce them, and then both families nigh inevitably begin giving each other a closer look. Objections may or may not be raised, but there's still a lot of reconciliation to do. You marry not just a person but their family, after all.

It's a pretty reasonable system, and God knows that there would be fewer NEETs and incels if more families in the West took hints from Indian mothers exasperated that their kids took their advice to ignore relationships and study for the NEET a little too seriously and need coaxing to produce grandkids eventually.

Indus Valley civilisation had a concept of caste despite not being aryan and the Aryans who came from the Eurasian steppes had Varnas, two are different but nearly identical in most cases now.

Eh? I'm not aware of any reason to think the IVC had a caste system, and I couldn't find a reputable source that says so. We know fuck all about them really, their language is undeciphered, and their cities show only the same kind of social stratification that most civilizations do, in other words the elite living in the nicer places.

There is no caste supremacy, arranged marriages exist a relic, and people who are living in urban centres and not poor don't really care as much about who they marry.

I don't know about you, but arranged marriages are very much a thing and far from deprecated. The BBC says that in 2018, 93% of all marriages in the country were arranged. That hasn't changed noticeably in the last 6 years.

A billionaire can satisfy most of their extended families needs with a relatively small fraction of their income. A few hundred million is more than enough to ensure they never go wanting! Plus, if said family are going to die too, it's an investment in their future.

A few billionaires are, whereas I wish it were the majority. What marginal value do many billions provide, when you're dead, especially when you have enough money for almost all material needs?

It should be a national priority too in first world countries. You don't have to worry nearly as much about the burden of elderly people needing expensive medical care if you can stop them getting old in the first place.

On the other hand, we have both Dario Amodei from Anthropic and Altman from OAI extolling the virtues of test-time compute in the vein of o1 (as the article acknowledges).

Personally, while I find the recent relative stagnation in LLMs to be annoying, it's only been what, two years since GPT-4? There are open source models that beat the OG GPT-4 while being much smaller (Llama 405B versus likely a trillion for 4), and models like Sonnet 3.5 are a small but appreciable improvement. I'm not going to call scaling out until we've spent either another year or a hundred billion USD on it, and we'll have to see how test time compute pans out.

Besides, even Ilya says they're working on their own secret sauce solution. We'll see when we see, it's far far too early to call AGI out of the cards.

In response to @Glassnoser in particular, I would point out that Kurzweil, as one of the original singularitarians, is likely banking on ASI to bail us out here.

Is aging an incredibly difficult problem? I would be the last to deny it. But is it fundamentally intractable, such that there is no hope at all of an end to it? I can't see a justification. It's an engineering problem, a brutal one, but solutions aren't ruled out by the laws of physics. We have existence proofs that there are organisms with negligible senescence around, and even large animals like sharks that live centuries if they're lucky. It's not a property restricted to jellyfish, and even a mere 50-100 years added to healthy human lifespan would be amazing.

Frankly speaking, I think it's a damning indictment of general rationality that we don't devote an appreciable fraction of GDP to solving it. If I was a billionaire, especially one staring death in the face, I'm not sure how much good my money would do me if I wasn't around to spend it. What good is anything if you're dead?

I think your evaluation of the current situation is quite accurate. There are pretty much no current interventions that would provide a guaranteed century of healthy life. That being said, we're not spending multiple percentages of GDP on the problem (or at least not productively, most medical therapy can be described as ad-hoc patches and sprays of WD-40 on a rusty old beater, when what we need is a full engine replacement and transmission overhaul). And with the potential of creating artificial entities much smarter than us, and soon? I wouldn't bet against it happening in the average remaining life expectancy of us Mottizens.

While I agree with Nate Silver on a lot of things, there's something I find inherently off-putting about his schtick. I read The Signal and the Noise around the time it hit the bestseller lists and had an addendum about the 2012 election. One of the themes of the book is that the so-called experts who make predictions on television don't base their predictions on rational evidence and don't face any consequences when their predictions fail. No in-studio commentator on The NFL Today is losing his job solely because he picked too many losers.

And that's why you listen to Nate Silver and super-forecasters who have a public track record of being better than the average when it comes to their predictions, often markedly so.

I'd say the state of affairs is better than it could be, given that Nate has modest internet celebrity status. Making good models and publicizing them has some payoff.

At one point he told me "Probability is interesting, but when it comes down to it, the only thing it's good for is gambling. We say there's a 50/50 chance of drawing a black ball from an urn when we know that the urn has 50 black balls and 50 white balls. When we talk about probabilities in the real world, it's like talking about the chances of drawing a black ball from an urn we don't know the size of." When discussing cards or dice, we're discussing random events based on repeatable starting conditions. When discussing elections, we're discussing a non-random event that will only happen once.

Downstream, I was mustering a defense of Bayesian reasoning, but you don't have to be anywhere near as formal to address this. We are lucky enough to be in a universe that doesn't exist in perfect uncertainty, things that have happened inform what will happen, and some humans are somehow able to do better than random at answering unresolved questions, with better modeling, be it explicit or intuition based. You can notice them, or pay attention to the gestalt impression of a crowd squeezing wisdom from attempts to win either Reputation Points or actual money out of their competition.

Knowing the odds has some utility, and should be of some use. You wouldn't throw up your hands if some rando used the argument your friend used to claim that Randomly Picked American X had 50% odds of winning the presidential election. If you were sensible and he was willing to put money where his mouth is, you'd take said bet and make your money back expecting to win that bet several million times per loss.

Odds convey information, or at least expectation. The sheer complexity of the universe and the vast sampling space is, thankfully, not as much of an impediment as it seems. I'd expect so, after all we evolved in it and had to understand our world and other actors better than a rock.

Uh.. That's the worst way of reasoning from evidence that's ever been tried, barring all the others.

Absent logical omniscience, you are occasionally going to be wrong, and then you try to be less wrong. Taken deeply enough, no macroscopic events in the history of the universe are likely to ever be truly alike or repeatable, so sorting out reference classes is unavoidably important.

"I was wrong about World War 3 not happening. Well, we can't have a World War 3 2.0 happen for me to be right about, but at least I can adjust my priors for massive wars happening in the future".

Besides. You can very much eat crow when you are confidently wrong. It just takes intellectual honesty, and Bayesians at least pay lip service to the notion we learn from our mistakes. Keep being bad at updating, and people will stop considering what you say to be informative (and that's not unique to self-professed Bayesians, because in practise most humans apply the concepts implicitly, some are more disciplined and explicit than others).

People can claim to be good anything regardless of facts on the ground. Talk and unfounded claims are cheap and may even be free (ignoring opportunity costs).

I'd expect anyone who outright calls themselves a Bayesian to do better on that front.*

*Going off Bayesian priors about what the kind of people nerdy enough to have even heard of the idea are like, let alone self professed ones