ff317

There's two issues here still, IMHO:

1) The LLM owners really can't guarantee that it won't directly plagiarize without attribution or licensing. Your code may contain a unique algorithm or method for solving something, and when someone asks the right question, your code may simply be the only answer it knows to give.

2) While the code being used as training input was open source and visible to the public to learn from, the models being built often aren't. It seems unethical to train from public data yet keep the resulting weights private and charge for access to use the trained weights.

Yeah this is the angle I look at the most, the Humans+Internet combo.

I don't believe LLMs will really get us much of anywhere, Singularity-wise. They're just ridiculously inefficient in terms of compute (and thus power) needs to even do the basic pattern-prediction they do today. They're neat tools for human augmentation in some cases, but that's about all they contribute.

I think, even prior to the recent explosion of LLM stuff, that the aggregate of Humans and the depth of their interconnections on the Internet is already starting to form at least the beginnings of a sort of Singularity, without any AI-related topics needing to be introduced. The way memes (real memes, not silly jokes) spread around the Internet and shape thoughts across all the users, the way the users bounce ideas off each other and refine them, the way viral advocacy and information sharing works, etc. Basically the Singularity is just going to be the emergent group consciousness and capabilities of the collective Internet-connected set of Humans.

Maybe rather than focusing so much on how to detect AI-generated content, we should instead focus on our general ability to validate the truthiness of content regardless of source. I don't really care if an AI wrote it, so long as the content is meaningful and informative. I do care if it's a load of junk, even if a human did write it.

Personally (as someone who was high-school aged in the early 90s):

1. FidoNet BBSes were one early pre-Internet form ( https://en.wikipedia.org/wiki/FidoNet ). Even once the early Internet was up and running for some, some of us lived in countries/places where there was no access, and so the loosely-interconnected FidoNet BBS system was the first form of online group communications I was exposed to. There were even some gateways passing email and usenet newsgroups between FidoNet and the Internet towards the end. I had several local friends, we all ran FidoNet nodes, and we did use it for interpersonal communications at times.

2. Usenet - I used this a little bit in the mid-90s for true communication, but mostly that was highly technical in nature, e.g. IETF discussions on protocol development issues, etc. Also lots of using it to download binaries of course, but that's not really chat. I mostly didn't "know" the people I was interacting with, other than by online reputation.

3. IRC - From around 1994 onwards for me (when I finally had real access to the real Internet), on the EFNet IRC network (and a couple others, to a lesser degree). I spent many hours of every day lurking on EFNet's "#hack" channel and several others. There was a real sense of an online-first community, and then there were smaller and often private channels of people you knew from your real life (the other local hackers in your immediate community, or people you met at conferences, etc). I won't defend the culture as necessarily stellar, especially by today's standards, but at the same time, there was a sort of sense of community where it didn't seem to matter much who or what you were in real life, a sort of "if you're here, you're one of us" vibe. I think at least around 1994-1997-ish, IRC was pretty central to my social life (not all of it of course, but still!). Some of it was more anonymous in nature, but a lot of it was people I had met in real life or were even local friends of mine.

I generally agree, but I tend to turn the perspective around to this: truly "Free Markets" are basically the natural, unevolved state of affairs when you don't have a better system in place. The whole reason we invented those better systems is to improve the "consequences for the majority".

I have no data, but I would hypothesize that the generational advantages do not disappear easily, even if a large chunk of the monetary value does.

Even just looking at from a purely monetary angle: maybe generation two manages to "lose" 70% of a $50 million inheritance. They've still got $15 million to start over with. Even if they lost 95% of the inheritance, they're still on a better financial footing than most with $2.5 million of capital laying around to rebuild themselves from. That's not enough to live like a king and never work, but if you're even remotely smart in managing that amount from somewhere in early to middle adulthood, that's still enough to at least never have a mortgage or worry about college for yourself or your kids. It's enough to start a successful small business or three as well.

But I think there's another angle on this. Even if the kids stupidly lose a lot of the cash, they're still riding on a ton of benefits from their enriched childhood: far better health thanks to a virtually limitless diet of high quality food and great medical care, the best schools and tutors, connection networks to all the other rich kids from their rich neighborhoods and schools, and learning (through osmosis if not explicitly) all about their parents and grandparents lessons of being in executive positions and how to invest wisely and preserve capital, what kinds of lawyers and insurances you should have, etc. Bottom line is they could literally set 100% of the inherited cash on fire and they'd still have a much higher chance of re-becoming wealthy than most.

I bought a new-construction place, and I was around during the process. After they were done with all the plumbing/wiring/etc, but before they filled in the insulation and sheetrock, I walked the whole house and took cellphone pictures of every bare wall to show where all the pipes and wires are run, and then labelled it all up by room. I plan to hand those off if I ever sell the place! I don't know why builders don't just deliver such a set of photos to everyone by default, it's been super handy!

We don't need M&Ms in the checkout aisle to feed and house 8B people on this planet. Don't conflate what I'm saying with a general anti-business sentiment.

There are many kinds of economic transactions that occur. The best kinds are the win-win transactions. I have an excess of X and a dearth of Y, you have the opposite, and we swap them to our mutual benefit and walk away happy in the long term thinking that we both made a good deal, and even a third party analysis by experts would agree it was a good deal. This is basically the kind of transaction that happens when I swap $2 for a bunch of fresh cilantro at the market, to cook a meal with. Or when I pay a skilled mechanic a reasonable fee to do maintenance on my car that I can't do myself.

Then there's the other kinds of transactions: the exploitative ones. One person substantially and noticeably wins and the other person equally loses. There is no upside for the loser in the bigger economic picture of our lives. Lots of basic examples of this are common, and some are borderline fraud. An example would be the mechanic who charges 10x what he really /needed/ to make a healthy margin, because he's the only mechanic in town and I'm immobilized by my failed car, and I'm poor, and I'm putting it on a credit card because I don't have the money but have to get the car working to keep my job.

I'm positing that, in a larger holistic sense, the transactions for the candy in the checkout aisle (all the advertising that goes into it as well!) are like that. They're not so much economically harmful: it may in fact be a "good deal" in a basic math sense to pay $2 for the candy's ingredients and manufacturing process. But at the end of the day, they're turning a profit and you're continuing an unhealthy sugar addiction and eventually dying of diabetes. It's a transaction that's explicitly designed to exploit you and harm you for their collective profit. This is not a win-win, at least not in a larger, holistic sense.

The whole sugar/candy industry is unethical in my mind. It may take some decades for society to catch up, but eventually eating sugar the way we do today will be seen like smoking and the tobacco industry.

I should make responsible decisions as an adult, but it's still fair to call that grocery store's actions unethical, and that maybe they should be regulated if they can't be ethical on their own.

Arguably, adults are worth protecting as well. A lot of marketing and even design intent for adult products is also extremely misleading and intentionally designed to prey on our collective fallibility. In that sense, it's at the very least unethical, and maybe should be illegal. We have some laws about truth-in-advertising, but they're very weak in a lot of scenarios.

I can give a real datapoint: Family of 4 in TX, suburban neighborhood, and we drive a ton of mileage (~25-30K/year) running kids around town and the occasional roadtrips. ~2 years ago, we financed solar panels on the roof, and around the same time, bought a brand new Tesla Model Y. We came from a slightly-larger gas SUV before that. I also switched electricity plans, and opted for one with free nights (9p-9a) and a high rate during the day, which doesn't buy back any power from my solar, because the math works out better that way. So basically the car recharges for free overnight, and the solar helps offset a decent chunk of the daytime usage. When I net /everything/ out (electric bill changes, solar monthly bill, car payment swap, fuel, and maintenance costs), all total on average I'm coming out ~$700/month ahead from making these changes.

I'll still write bash, but only if it's very trivial (~ <= 10-ish lines, no super-complex conditionals, etc). For anything else, it's worth stepping up to just about any more-robust language. If you don't like the hoops that something like Python makes you go through for basic shell-like operations, maybe try Perl? For middling-complexity scripts, perl can be a nice win over bash, without all the development overhead of Python. Perl5 is pretty much installed everywhere and universally compatible for anything but the very latest language features.

[Disclosure: I'm an engineer at Wikimedia involved with this project]

Wikimedia hasn't actually formally launched or announced this project. We've been working on it in the background for a long time, and we've just recently reached the point of wanting to do a slightly-broader round of both beta-testing the service and soliciting other meta-feedback on the plans with our Wikipedia editor and reader communities.

Unfortunately, the only way we can do this testing and feedback round is through a public interface like meta-wiki. And of course, once we've put out public information on how to use and access it, it's being picked up on various social media and news sites and publicized beyond what we desired at this stage. We knew that was a risk, but it's still a little jarring how fast it has shown up in places like this HN thread.

We absolutely plan to have a published privacy policy and other such things in place when the service is actually, officially launched. Launching some time in the future is not yet certain to even happen, as there is still internal evaluation/debate of the project's various costs and risk/benefit going on during this phase as well.

I think we'll never stop extracting oil in the foreseeable future, we'll just be doing a lot less of it. There will always be corner cases where diesel or other oil derivatives (or NG) are the /only/ reasonable option for fuel.

Think generators for an antarctic station that can only use solar in certain months and wind isn't enough (or reliable enough)). Even if efficient electric commercial aviation at scale ever happens, you can bet military jets will still need jet fuel. I'm sure there's many others. Methane for rocket launches?

On top of fuel, there's also the use of some fraction of oil in industrial chemical processes to make lubricants and plastics. There may be other chemical processes that can do it without oil, but they may be too costly (in terms of other extractive processes for ingredients, or in terms of yield, not just energy).

As they age, you're going to find a whole different level of arguing about how to raise kids on an every-other-week rotation with conflicting influences/decisions. Trust me, I've seen it. It can work, but you're gonna need to put in some work on the relationship with the other parent to get through it and find decent compromises.

I think at the heart of that debate, there lies a kernel of the essential progressive vs conservative debate on "progress" (and I mean these terms in the abstract, not as a reference to current politics). Even if you buy into the idea that the above (living forever as an AI / cyborg / whatever) is a good outcome, that doesn't mean it will work as planned.

Maybe society bets the farm on this approach and it all goes horribly wrong, and we all cease to exist meaningfully and a malevolent super-AI eats the solar system. Maybe it does kinda work, but it turns out that non-human humans end up losing a lot of the important qualities that made humans special. Maybe once we're cyborgs we stop valuing "life" and that changes everything about how we act as individuals and as a society, and we've lost something really important.

Progress is a good thing, but always be wary of progress that comes too quickly and broadly. Let smaller experiments play out on smaller scales. Don't risk our whole future on one supposedly-amazing idea. You can map the same thing to gene editing quandries. If there's a new gene edit available for babies that's all the rage (maybe it prevents all cancer, I donno), we really don't want every single baby for the next 20 years to get the edit universally. It could turn out that we didn't understand what it would do to all these kids when they reached age 30 and it dooms us. This is why I rail against the overuse of central planning and control in general (see also historical disasters like China's https://en.wikipedia.org/wiki/Great_Leap_Forward ).

No personal offense intended, but I really /hate/ this justification. "Open Source" should not be equated with "Hobby Projects". It should be possible to run large-scale important things on pure open source.

My employer (Wikimedia, which is a non-profit) has been running all our server infra on pure open source for many years, for very principled reasons. We even publish all of our internal infrastructure config in public git repos as well. We're an open-source-only shop that's doing very non-hobby work. We have thousands of servers, we run our own CDN on multiple continents, we have somewhere around a billion unique users per month, and our average global rate of inbound HTTP requests is around 135K/second.

However, it keeps getting harder to stick to our open source principles every year as more projects shift towards open-core models with this kind of "Anyone who needs feature X is surely a big corporation making billions who doesn't mind running closed source software" type of thinking. It's not just bad for us, it's bad for the whole larger open source and Internet ecosystems if all the important bits for running at scale are locked up and hidden away in closed-source software.

The salaries are much larger than what gets put in the "hosting" line, which merely covers things like leasing cage space in datacenters and paying for transit/transport/peering links.

I'm not gonna do another deep dive on the org structure, because I've done it before when these topics come up ( like this one half a year ago: https://news.ycombinator.com/item?id=33640817 ), but I'll provide my overall view as a 10 year engineering veteran at the WMF who has fought many battles in this org, which you can take or leave:

There are problems at the Foundation, like there are everywhere. Resources are not allocated as well as they should be, which is also common just about everywhere. However, by and large, the money is put to good use, and the employees of the Foundation are doing good and necessary work, and there's no way to run today's Wikipedia on today's Internet for anywhere close to $3M/year total organizational budget, so it's pretty disingenuous to use that as a comparison point.

Go look at budgets for any other org which runs a top-N important site on the Internet for comparison. Consider that we adhere to stronger values and principles around rights and privacy and open source, all of which /increase/ our operating costs.

WMF runs on a shoestring, relatively-speaking. WMF just had a ~5% layoff over budget reductions ( https://blog.legoktm.com/2023/04/05/wikimedia-foundation-lay... ) in response to a community outcry about our fundraising. We (the engineers) regularly run into budgetary and/or priority constraints pursuing new useful SRE-level projects that increase scalability, reliability, or directly improve user experience (e.g. latency for readers and editors, rendering speed, better APIs for various consuming parties, more edge caches, more hardware redundancy, etc). We also don't get paid industry equivalent rates, or get the stocks or bonuses that we'd get elsewhere.

You can make all the arguments you want about some past or present form of grift at some upper organizational levels, but I'll still contend that most of the money is going to good use, that it's very expensive to run Wikipedia responsibly and reliably, and that the foundation doesn't have the cash pile necessary to stop relying on donations for the long-term sustainability of its projects. Even if the budget we have was allocated perfectly, I'd still be arguing that we need more to do the job right.

The hash is keyed and rotates once a day, so it's not a long-term history. And in any case, nobody's trying to build a unique tracker out of it on the other end, which is the main thing (even in our internal analytics, we even throw away the PII we get directly from request attributes after 90 days).

I'm not saying it's easy or perfect, but Wikipedia does tracker-less analytics, and yet still produces https://stats.wikimedia.org/#/all-projects .

We do use cookies for part of the analysis, but they're not unique user-tracking cookies. Instead there's one that simply tracks a self-reported datestamp on when the user last visited, which looks something like:

WMF-Last-Access=2023-01-01,Expires=Wed, 01 Feb 2023 12:00:00 GMT

(we send a Set-Cookie for this with the current datestamp only to the 1-day accuracy, which expires ~32 days after it's set (but rounded to 12-hour accuracy), and is replaced constantly).

There's another more-recent one we use that's explicitly about differential privacy, which send back info on the hashes of the 10 most recent URLs you've visited on the site, IIRC. None of them are unique tracker hashes for a given user, though.

My retort would be:

If you have a singular employee doing a certain unique job, or a very small handful, then it's pretty trivial to instead manage them directly on a human level. You don't need Orwellian tech tools to have conversations with them and make a one-off evaluation as to whether they're giving it the effort and authenticity it should, and again: results matter. What were your expectations (in terms of accomplishments) in hiring this role, and are they being approximately met?

But if you have a lot of employees doing a similar job, then it just gets increasingly easier to be objective about it. Look at their actual outputs statistically, and find the outliers and focus effort on those cases. Does Joe's output sit two standard deviations below the median? Why? Maybe there's a good reason and you document it and move on, or maybe they're just being deceitful. I still don't think mouse/eyeball tracking or virtual timecards are going to help much.

I have yet to see any convincing argument for monitoring employee behavior/attentiveness/time, for any job role or level, really.

The whole world needs to hurry up and learn that what matters is what your employees really accomplished. You don't need to track hours, or clicks, or chat lines, or documents opened. You need to track that they're producing the accomplishments they were hired to accomplish.

Did they fix the software bugs you asked them to fix this week? Did they sign up 10 new customers today? Did they write that research report yesterday? Did they flip all the burger patties they were supposed to flip today?

As a manager, if you can't define productivity in real terms like that, and manage employee performance based on real work outputs, you're just doing it wrong. Someone else that does it right at a competing team or org is going to eat your lunch eventually.

Someone has to link this here, so it may as well be me :)

https://steve-yegge.blogspot.com/2006/03/execution-in-kingdo...

Not to knock the amazing contributions of Tim, but IMHO Gopher and HyperCard had already laid a lot of the important concepts down in various suboptimal ways. I think there were enough smart people using both in the 90s that it was virtually inevitable that something like Tim's work was bound to arise somewhere, somehow, as a merger of and improvement on those two sets of ideas.

Yes, but those are "real". Most peoples' life patterns are actually altered by week and weekend boundaries. Mar 1 this year fell mid-week on a Wednesday. It made no difference to most lives except that it made everyone's calendar ugly with duplicated / grey-ed out days at the end of a February calendar and the start of a March calendar.

I hear you, but: good software/services require continuous investment. It's kinda hard to set an up front one time price that covers that without it being ridiculously high. You're not buying one copy of a static piece of software: you're buying all the future bugfixes, security mitigations, server maintenance, future integration work (it has to work with the next upcoming version of your phone's OS, or the next version of your browser, etc), adhering to important future tech standards in the space. With a subscription model, those costs can be amortized over the lifetime that you're actually using the software/service, and can increase or decrease as necessary due to economic conditions as well.

I'll nitpick the "gold doesn't have intrinsic value" part: There's a limited supply of elemental gold on the planet, it takes some effort to extract it, and it has unique electro-chemical properties that make it productively-useful for industries. You could consider the jewelry use of it to be "merely aesthetic", but I would argue that even in that aspect, it has demonstrated intrinsic value. So many cultures over so many hundreds of years have valued its use in to enhance beauty via jewelry, which you can think of as an aid in attracting mates, etc, which definitely has intrinsic value to many humans.

Yeah, I figure if a bash script goes over ~10-20 lines, or involves really long lines, or quotes within quotes within quotes of different kinds, it's time to move on to Python or similar.

Seems like it would be simpler to fix this with regulation on housing "investment". Don't allow large corporations to buy up giant counts of single family homes (taking them off the market and jacking up real estate bubbles). Basically we need to stop the predatory act of spending billions to buy many thousands of houses out of the market just to turn around and rent them to those who can no longer afford to buy in the market they're effectively cornering. Some builders are now building whole neighborhoods full of single-family homes that are being sold directly to rental investment companies from the outset.

My off the cuff (and thus probably naive) thought would be to charge progressive federal taxes on home ownership which progresses based on the unit count owned by an entity. Whether you're an individual or a corporation: your first house is free (0% federal tax). If you own two houses, you pay 3% on the value (of both houses) every year for the right to own them. Maybe at three houses it's already at 10%, and by the time you get to ~10 houses owned, it caps out at 100% tax on the value every year, making large-scale home investing completely unprofitable. The funds from this new tax could be set aside specifically to fund affordable housing programs of some kind.

Under such a scheme, we'd still have rising prices from true single-owner-occupiers under the existing system, but it would slow down and be less-bubbly, at least in the national average, and builders will eventually build new homes to equilibrium, where regulations allow them to. Some moderately-wealthy people could probably still own a second vacation home, or 2-3 rental properties. Some extremely wealthy people might still own 10+ different homes for personal/family use at exorbitant tax cost, but at least the taxes from such scenarios go towards fixing things.