I don't think I'm completely oblivious to social status, but I've never understood at a deep level the way most people seem to process concepts like "fame" and "celebrity". I have never had the experience of being awestruck by a person, or elevating them above personhood, though I admire plenty of people. With the few brushes with, maybe we can say "microcelebrity" I've had myself, the opportunities and status benefits, while nice, seemed not worth the bizarre distorting effect on social reality it has, like the thing where someone's heard of me and talks to me like they know me when we've never met is uncomfortable at best, and most people considered various degrees of famous I've met who I came to respect seemed to be similarly jaded with this awestruck or even worshipful reaction some people seem to have when they idolize someone. I really think this whole cluster of behaviors is unhealthy and weird, and the fact that mass-communication technologies and the massive societal resources bent toward persuasion (both commercial and political) have drastically amplified it is probably one of the major causal factors in the polycrisis of the modern world
There's a pretty stark difference between medical science, medical practitioners, and medical institutions.
Medical science has done a lot to earn credibility over the last couple hundred years. It fully eliminated a disease that was a leading cause of death in multiple human civilizations. It reduced rates of infant mortality by staggering numbers. It created vaccines and surgeries that have each saved countless lives and effectively cured life-altering disabilities and injuries, and demoted tons of diseases from death sentences to mere manageable disabilities or even mild temporary inconveniences
Medical practitioners are a mixed bag that ranges from total altruists to total mercenaries. Most people can learn a skill and incentivizing any skill via things like money or social status creates some goal-misalignment, but also dramatically increases the prevalence of the skill and also the community surrounding it, including research, pedagogy, and infrastructure. I think there's considerable value in not automatically trusting every medical practitioner, but on the whole many of them have earned credibility
Medical institutions have mostly been responsible for the existence of trustable practitioners and research, though in the rich world and especially the US specifically there are significant misalignments that come from capture from power-hungry parties in both government and the private sector. The American medical establishment in particular seems at this point near unsalvageable, having broken so badly that it often deliberately tanks quality of care to enforce state power or to turn a profit. We should probably rethink a lot of how institutions fund research, administer hospitals, prioritize care, act as financial buffers, what legal and enforcement authority we vest in medical practitioners sanctioned by the institutions, etc. So I agree with you there, especially in the United States specifically, though much of this was apparent well before 2025 (practically before I was born)
I don't know if the balance of evidence supports significant changes in sexuality and eroticism or not, but I think the way you've made the case that it doesn't here is unconvincing. Consumption of erotic content on the internet and actually engaging in sexuality as a participant are drastically different matters and both the sign and magnitude of any correlation between them is hard to pin down. From my own anecdotal perspective, there's a weak but significant anticorrelation between how much porn/erotica people consume on the internet and how much they engage in sex or kink or even relationships with other people. Maybe the sample of people I've met isn't indicative, but I would say neither is anything you're using as supporting evidence here
There is no sustainable future without repair and maintenance, and there are significant barriers to repair and maintenance imposed by governments in the name of "Intellectual property".
I don't think (from a sci-fi reader's perspective) this is a particularly new idea, and with all due respect to Hinton as a machine learning pioneer, my disinterest in this topic is not going to be alleviated by saying some names that can claim expertise. Stories about ASI are essentially the same thing as stories about advanced alien civilizations or gods. They essentially act as a repository for hopes, fears, and generally expectations one has around the concept of being dominated and ruled by something more powerful than we can imagine defeating. Telling stories of these kinds can do a lot to examine the relationships we have to power, how we've come to expect it to behave, what it's done to us, and for us, but they're not newsworthy meaningful predictions of the future and never contain good advice for what one should actually do, so it's weird to keep treating them as such
You can get pretty far by mostly just claiming to use it "when it makes sense" but you do meet people who are very pushy about it. Hoping that calms down as knowledge of the downsides becomes more widespread
With all due respect, "wow this is cool, I wonder what it'll turn into" is basically the mandatory baseline stance to take. I'm lucky that's where I'm still basically at, because anyone in a technical position who shows even mild reticence beyond that is likely to be unable to hold a job in the face of their bosses' frothing enthusiastic optimism about these technologies
For all the advancement in machine learning that's happened in just the decade I've been doing it, this whole AGI debate's been remarkably stagnant, with the same factions making essentially the same handwavey arguments. "Superintelligence is inherently impossible to predict and control and might act like a corporation and therefore kill us all". "No, intelligence could correlate with value systems we find familiar and palatable and therefore it'll turn out great"
Meanwhile people keep predicting this thing they clearly haven't had a meaningfully novel thought about since the early 2000s and that's generous given how much of those ideas are essentially distillations of 20th century sci-fi. What I've learned is that everyone thinking about this idea sucks at predicting the future and that I'm bored of hearing the pseudointellectual exercise that is debating sci-fi outcomes instead of doing the actual work of building useful tools or ethical policy. I'm sure many of the people involved do some of those things, but what gets aired out in public sounds like an incredibly repetitive argument about fanfiction
I mean, even if that's true, being based in the EU might matter a lot given how keen that bloc is on becoming more technologically sovereign from America and China right now
We have an impossibly pervasive network of sensor blisters littered throughout our lives, to the point where I don't feel comfortable discussing certain sensitive topics in most other people's homes, but every step of the way most normal people have given the same refrain: "oh, the tech companies probably already have all my data anyway"
Now that those tech companies are working closely with an American regime that seems increasingly willing to disregard the rule of law and public perception to round up people they deem undesirable in large numbers and put them in concentration camps, and we have natural language processing tech that can pretty effectively filter through large amounts of text for some semantic analysis, I hear some of the more attentive people coming to the barest hint of a realization that this situation is unacceptably dire
People heavily invested in AI, obviously. We have an economy that's heavily driven by investor perception, and while "AI is taking jobs" sounds bad for workers or humanity as a whole, it can still be a sales pitch for investing in AI
The 2022 drop for SWE is easy for me to explain, and it's not on these analysts' list of factors (though I'm not an economic quant, I don't know how you could really control for it): In 2017, a tax bill was passed that cut a particular tax incentive in 2022 in an effort to be counted as "revenue neutral" despite being otherwise a massive tax cut overall. The incentive in question was a writeoff for "Research and development". This means that in 2022, it got effectively much more expensive to hire anyone who falls under that category, including developers not directly necessary for the day-to-day function of a business (hell, one might argue they would have counted anyway) and scientists of most kinds. That this hit big firms, which have a higher relative amount of R&D efforts going at a given time, first makes a lot of sense.
For customer service, my explanation is that companies literally do not care about customer service. Automated phone trees, outsourced call centers whose reps have no real power to help a customer, and poorly-made websites have been frustrating people for decades, but businesses never seem to try to compete on doing better at it. It's a cheap win with investors who want to hear about AI initiatives to lay off yet even more of this department, because it doesn't matter if the quality of service declines, there are no market or regulatory forces that are punishing this well enough to ever expect firms to stop breaking it, let alone fix it
This places a lot of faith in the following assumptions:
1. Efficiency measures as written to benchmark this coupling with economic productivity overall
2. Monetary assessments of value in the context of businesses spending money corresponding with social value
3. The gains of economic productivity being distributed across society to any degree, or the effect of this disparity itself being negligible
4. The negative externalities of these processes scaling less quickly than whatever we're measuring in our productivity metric
5. Aforementioned externalities being able to scale to even a lesser degree in lockstep with productivity without crashing all of the above, if not causing even more catastrophic outcomes
I have very little faith in any of these assumptions
It's crazy to me that people will reference Keynes' prediction of leisure without acknowledging that we chose not to do that. The dystopian way in which work has become more competitive, intensive, and ill-compensated even as economies have supposedly continued to become more productive is the result of policy choices, not some inevitable fact of the universe
This is honestly why I love C++ ranges right now. The "pipe" syntax is a "left-to-right" of writing very powerful map/filter operations in contexts where I'd want a list comprehension but in a much more sensical (and for that matter customizable) order
This view of social interaction seems symptomatic of internet socialization, specifically social media. Since engagement is often driven better by controversy, the overall incentives in that context make an experience like yours more likely. Similarly, LLM providers who offer a public-facing chat service tend to be incentivized to tune their models to foment addiction. While people can indeed treat each other awfully sometimes, I think you could benefit a lot from seeking out specifically in-person interactions, perhaps in a context with a shared activity that can take some of the edge off of direct interpersonal engagement
Yea. It's kind of the same error, in a way, as people who assume that there were no old people in the middle ages. The overwhelming majority of the increase in expected lifespan between then and now comes from drastic decreases in the infant and child mortality rates. While current medicine is only really making slow, incremental progress on letting the oldest people live longer, even if this was the bulk of the advancement you wouldn't see the kind of movement on overall life expectancy you'd get out of reducing those, and that's just on the pure statistical basis of how the metric is constructed. But on top of that, I think it's nearly impossible to understand just how many infants used to be stillborn, and how many diseases we essentially eliminated. The death of a child from an illness used to be a fairly common tragedy, now it is a rare one.
It's just a little internet toy that probably cashes out to be a slightly more impactful version of "memento mori", but you could add a little backend complexity without collecting any more demographic information and get a more accurate life expectancy given only one's current age from extant actuarial tables. If you wanted to be extra cheeky, you could have it adjust on a regional basis based on IP address too
Mostly the principle that direct, tangible, and irreparable harm is being done by the speech. I called these "hairy" exactly because they are less clear-cut than most cases. The line between satire or speculation and slander, for example, can be kind of notoriously hard to draw, many "doxxing" incidents mostly involve finding already-public information, and incitement to violence is pretty difficult to demonstrate in a satisfying way
The reason I bring up these examples is that they are what I consider the edge cases of freedom of expression. There are genuinely difficult tradeoffs between values, but allowing stuff that bothers some people to be published is not one of them
It's crazy to me that people can simultaneously understand that censorship is bad and still think we can allow for exceptions for things like "obscenity/pornography" or "hate speech"
Free expression means that objectionable things will be said, even published. There are certainly hairy exceptions, like doxxing, slander, or incitement to violence that can put people in immediate danger, but stuff like this clearly doesn't fall in that category, and giving finance the ability to censor in this way is not signficantly better than governments doing so
I really like this usage of recurrent modules to augment attention-based models, and I think this is a really cool result and a fruitful avenue for future work
This exact kind of sloppy equivocation does seem to be one of the major PR strategies that tries to justify the massive investment in and sloppy rollout of transformer-based language models when large swaths of the public have turned against this (probably even more than is actually warranted)
The effects of wealth on outcomes meaningful to most humans look mostly like a step functions, with some continuous-lookin' bits here and there. Anecdotally my assumption is that partying x wealth is bimodal, with peaks near the top and the middle-bottom of the wealth distributions. But homeless people and people scraping by on two minimum-wage jobs, both of which comprise a larger portion of america than when I was a child, are unlikely to be able to host or even often attend parties, like I think there's a sharp step to basically zero after a certain poverty line unless you're in a really integrated community (also less common over that time period, to be fair)
I like that this delves into the relationship between "helicopter parenting" and this trend, and maybe I missed it, but I find that it conspicuously lacks economic precarity and the decline of real wages over this time period as an explanation. Hosting social events does cost free time and money and most people have way less of both in real terms than the period it's comparing to
Every industry in America, and especially tech players, work to lock in their customers, paid or not. People who are dependent on their phones don't make choices like that, and anticompetitive behaviors are becoming less illegal and easier
At this point "vote with your wallet" is basically a delusion in contexts like this
The bitter lesson is a good demonstration of how people have really short memories and distributed work loses information
Every AI "breakthrough" comes at a lag because the people who invent a new architecture or method aren't the ones who see its full potential. Because of the financial dynamics at play, the org or team that sees the crazy-looking result often didn't invent the pieces they used. Even if they did, it's been years and in a fast-moving field that stuff has already started to feel "standard" and "generic". The real change they saw was something like more compute or more data
Basically, even the smartest people in the world are pretty dumb, in the sense of generalizing observations poorly
Some people got a career path like this, and in fact I either personally have or know someone who has experienced all of the things you're listing here as perks. I'd even say you likely get all that if you're in the 95th percentile of seniority and job stability among devs, or if we're talkin' about ten or fifteen years ago. Sounds like you're in one of those categories? Congrats. I'm not doing terribly myself either, albeit not quite that well. This always seems to be the argument. "How very privileged we are! Devs have so much going for us, we ought to be grateful! Clearly there is no reason we should collectively bargain." I dunno. I'm not into doing oppression olympics about my working conditions? There are still things that suck about how labor arrangements are in this profession, and this has become especially apparent during the recent multiple-year enormous layoff cycles in tech, and this is also indicative that we are, as a profession, perhaps losing some of the very scarcity-driven leverage that creates these conditions. Also, this kind of view really seems almost explicitly designed to buck solidarity with people who, in the same career path, don't have as many options as us, possibly because they are more junior or have nontraditional backgrounds or whatever. In fact, working conditions having high variance between devs for various reasons is one of the problems unions try to solve