LLMs know nothing but are great at giving the illusion that they know stuff. (It's "mansplaining as a service"; it is easier to give confident answers every time, even if they are wrong, than to program actual knowledge.) Even your first case seems wildly optimistic. The second case is a lot of "maybes" and "we don't know how but we might figure it out" that seems like a lot to bet an entire farm on, much less an entire industry of farms.
We sure are looking at a shift in the job market, but I don't think it is a fork in the road so much as a Slow/Yield sign. Companies are signalling they are willing to take promises/hope to cut labor costs whether or not the results are real. I don't think anything about current AI can kill the software development industry, but I sure do think it can do a lot to make it a lot more miserable, lower wages, and artificially reduce job demand. I don't think this has anything to do with the real capabilities of today's AI and everything to do with the perception is enough of an excuse and companies were always looking for that excuse. (Just as ageism has always existed. AI is also just a fresh excuse for companies to carry on aging out experience from their staff, especially people with long enough memories/well schooled enough memories to remember previous AI booms and busts.)
But also, yeah if some magic breakthrough makes this a real "buggy whip manufacturer moment" and not just an illusion of one, I don't mind being the engineer on that side of it. There's nothing wrong about lamenting the coming death of an industry that employs a lot of good people and tries to make good products. This is HN, you celebrate the failures, learn from them, and then you pivot or you try something new. If evidence tells me to pivot then I will pivot, I'm already debating trying something entirely new, but learning from the failures can also mean respecting "what went right?" and acknowledging how many people did a lot of good, hard work despite the outcome.
Embeddings are still mostly just vectors into n-dimensional K-means clusters. It isn't "knowing" two things are related and here's the evidence, it is guessing two things are statistically likely to be related, based on trained patterns, and running with it without evidence.
It has no "semantic understanding" as we would define it. It's just increasingly good at winning cluster lotteries because we've increased the amount of training data to incredible heights.
Can you explain how you "know" two things are related? If I ask you the similarities between a cat and a dog, is your answer based solely on an understanding of their genetic phylogeny and how those genes express traits?
Grouping vectors in concept space is exactly how you create semantic understanding. The proof is in how good they are at creating semantically valid text. The fact that it took massive amounts of data is irrelevant. That just shows how much knowledge is encoded in all our language. It takes humans a ton of training to know things too.
I've been exploring related feelings in the related "For me, this is just a Tuesday" space. I had started to feel like a food tour group was stalking me personally. My usual Tuesday lunch. My usual Friday dinner. So forth. The first "gut feeling" was irritation. It's a lot of people showing up in a small space with random lectures distracting from whatever I was doing while eating (such as often reading a book).
It took a couple weeks, but I realized that I was the spoiled one and the other side of "For me this is a Tuesday" should be "I'm glad my local businesses are interesting to tours" and "It's easy to forget how impressed I might be with this lunch if I was visiting some other town, isn't it great it can be my 'usual Tuesday'?"
I started listening to some of the lectures. I could easily mock some of them and/or clarify/edit/fix mistakes in them, but also they can still be an interesting bit of perspective, including the way that tourists respond to them. It's fun to have that tourist perspective of the local area.
It's also a fun reminder to do and try more of the tourist things locally. When your "usual Tuesday" is someone else's exciting and desired vacation experience, what else are you seeing with the somewhat dulled eyes of being a local but would greatly enjoy if you thought like a visitor to your own city?
Space Cadet (Pinball) has the most direct answer: it was written largely in x86 assembler and didn't survive a 64-bit translation attempt. Raymond Chen says the ball would ghost off the table, fall down and end the game in seconds when trying to run in 64-bit math. Raymond even takes personal responsibility for the failure to keep Space Cadet alive and disappointment it didn't survive past Windows XP:
The larger answer to the rest of the games seems to be related: Windows trying to shrink its non-cross platform code "liabilities" and things it needed to translate between processor architectures. The games were never a priority for the Windows team. Most were either intern projects and/or contracted from "second party vendors". In Windows 8, Microsoft decided to completely contract all of the games to a second party, the strange and sometimes controversial Arkadium [1]. The Arkadium Solitaire and Minesweeper were installed by default for a while, but as Arkadium started injecting more ads and also quickly increasing the install sizes of the games, Microsoft did the natural thing and removed them as default installs so people would stop complaining about their size and/or ads and instead just adding shortcuts to install them from the Store.
AMC was one of the theater chains that did figure it out, but was also smart enough to realize that they didn't need the middle man and had a large enough chain to leverage. AMC A-List still exists. (Up to 4 movies per week at $23-35/month.)
Yeah, "MoviePass" was its own SV startup that tried to be the generic version of this pass and apply to "all" theaters. It burnt through a bunch of VC funds just to fail and it shouldn't have been a huge surprise that big chains like AMC decided to do their own similar passes on their own without a middleman.
I'd say it is indeed a huge surprise that a struggling company refused to do business with another entity which was trying to purchase tens of millions of dollars of its product.
AMC is the dumbest company (or more specifically, its CEO Adam Aron is the dumbest executive). MoviePass came in out of nowhere and became the largest purchaser of movie tickets... millions every week. And AMC actively fought against them, refused to even let them buy tickets at full price, and led the charge to drive them out of business. For what alternative? Mostly, nothing but empty seats.
AMC's stock price is $1.59 as I write this vs $50-70 while MoviePass was peaking around 2018. AMC had to do a 10-to-1 reverse stock split to avoid being delisted, they may need to do another one. They even got a brief "meme stock" spike over $250 and managed to do absolutely nothing productive (except pay the CEO more) with this new capital access.
Hah, yeah, my local AMC is a ghost town as people moved to newer and/or better maintained theaters. (Not entirely AMC's fault, they bought a decades old theater that was sort of on its last legs after four other companies controlled it in as many decades.) Most of the screenings I attend are empty or nearly so. I almost need an app to find the full AMC screenings. (I know it is at least as much a factor of which nights of the week I attend. I could attend busier nights. But even then this theater's busiest nights now are not what they were way back when I was in school.) I do sometimes miss seeing something with a large audience.
There was a Harkin's location a few blocks from me kind of like that... it was old, but still really busy... the only difference is the land became so valuable they wound up selling anyways. The next nearest Harkins and AMC have much worse parking, even if newer/larger theaters, because they're sharing in strip-malls.
Yeah, the dynamics and shifts in trends of where the newer theaters get built and which theaters become the busiest theaters is fascinating to me, and especially how much nostalgia influences which theater I want to spend the most time at.
I lost my favorite theater in the early 2020s and then its sibling, and last locally-owned and operated theater, to a landlord screwing with rents to try to attract an inner-city Publix. (A deal which still may not actually happen.)
That leaves the AMC as the last regularly attended theater of my high school and college days still standing.
A newer, more popular Cinemark is in the city's biggest mall. I still remember when mall theaters were the worst/cheapest/smallest places. This has flipped now that most of what's left of the malls are the new theater/Dave & Busters at the one mall or the Top Golf/Puttshack at the other mall with the rest of the malls seeming now just weird appendixes to the new attractions. Meanwhile, I don't want to deal with Mall Traffic, which is still a thing in these flipped malls, I don't entirely know why.
Most of the rest of the most popular theaters all wound up in the Exurbs, two beltways away from the city's downtown, presumably due to cheaper land, and I don't want to commute that far to regularly watch movies.
Haven't even considered the theater from when I was a kid... I was in a smaller town, and until I was about 13yo, there was a single theater with two screens. I remember waiting about 3 hours in line across two showings to see Return of The Jedi opening weekend. Aside, I also remember when you could get the movie, a small popcorn and a small drink for $5, and not being able to convince my dad to fork over the extra when the theater raised its prices... that's when my friends and I would stop at Walgreens on the way and buy candy there instead of at the theater.
Yeah, mall theaters are just kind of hellish... mostly because of the parking/organization... I know why they're laid out how they are, just really wish they'd switch it up to make the theater easier to access if that's all you want.
The new mall one is facing the parking lot like a classic "anchor" department store and doesn't even have a door connecting to the mall, you have to leave the theater to visit the mall. It leaves for me a lot of questions about what the point of having built it attached to the mall was. (Same thing with the Top Golf at the other mall. At least the Putt Shack has a window wall inside that mall. I don't remember if it has a door, though.)
> I would argue that any sufficiently large system reaches a point where more code is in fact the opposite of what it needs.
I have absolutely worked on code bases I would describe as "marbleized bricks" where the best thing I can do is carve out the statue they already contain. There's a great satisfaction in making PRs that mostly delete things, but the later result is a program that works faster, has fewer bugs/edge cases, is easier for the next person to debug.
The LLMs certainly can add more layers of marble. Companies don't often know how much more they need an artist with sculpting tools more than a bricklayer.
Part of it is why carry a physical wallet anymore when your phone (and Apple Watch) can store all of your cards? There are even US States now that let you store your Driver's License in Apple Wallet. (And a version of Digital ID based on a US Passport that works for TSA and sometimes but not always US Customs.) There's an increased ability to leave the house just carrying your phone and not a physical wallet.
In a gym context specifically: a lot of gym wear doesn't have pockets. Being able to leave your phone and physical wallet at home or in a locker and use your watch (which you also use for workout tracking) for every membership card swipe, vending machine electrolyte/protein drink purchase, and gym class ticket can be very convenient.
If you have an Apple Card they already assume that you have an iPhone to tap to pay with. Why pay for hardware in the card that duplicates hardware your phone already does better?
(Better as in Phone tap to pay has an extra layer of security that card tap to pay does not. But also yes, cynically, better for Apple because Apple gets a small cut in Phone tap to pay to help pay for that extra layer of security.)
Using an Apple Watch for tap to pay is really nice, for what it is worth.
Handing your phone to the guy working the gate at the parking lot is awkward. Will he need to hand it back and forth for face ID? Handing a credit card like everyone else does is better, but why is this heavy titanium card one of the few that doesn’t work on his tap reader as expected?
I wear a garmin when I work out but otherwise want a mechanical watch with no tracking or distractions.
In the modern world, the guy working the gate at the parking lot should be handing you the reader and not care what thing you are swiping, inserting, or waving at it. I think Europe gets it. America is slowly catching up.
But if you have to pass a card because some business still feels that they have to touch your cards in this age, the EMV chip on the titanium card is just fine. They should just insert the card after a second or two of wondering why it doesn't tap. If their reader only takes tap to pay, then their reader needs to placed in a more convenient place for you to tap things like your phone or able to be passed to you for you to do what you need to do to pay.
No one should be physically handling my payment method anymore but me. It's not a great service to walk 20 feet away from me with my credit card just to make a payment at some hard to reach terminal. In fact it's a security risk for skimming and card stealing (and always has been).
LLMs know nothing but are great at giving the illusion that they know stuff. (It's "mansplaining as a service"; it is easier to give confident answers every time, even if they are wrong, than to program actual knowledge.) Even your first case seems wildly optimistic. The second case is a lot of "maybes" and "we don't know how but we might figure it out" that seems like a lot to bet an entire farm on, much less an entire industry of farms.
We sure are looking at a shift in the job market, but I don't think it is a fork in the road so much as a Slow/Yield sign. Companies are signalling they are willing to take promises/hope to cut labor costs whether or not the results are real. I don't think anything about current AI can kill the software development industry, but I sure do think it can do a lot to make it a lot more miserable, lower wages, and artificially reduce job demand. I don't think this has anything to do with the real capabilities of today's AI and everything to do with the perception is enough of an excuse and companies were always looking for that excuse. (Just as ageism has always existed. AI is also just a fresh excuse for companies to carry on aging out experience from their staff, especially people with long enough memories/well schooled enough memories to remember previous AI booms and busts.)
But also, yeah if some magic breakthrough makes this a real "buggy whip manufacturer moment" and not just an illusion of one, I don't mind being the engineer on that side of it. There's nothing wrong about lamenting the coming death of an industry that employs a lot of good people and tries to make good products. This is HN, you celebrate the failures, learn from them, and then you pivot or you try something new. If evidence tells me to pivot then I will pivot, I'm already debating trying something entirely new, but learning from the failures can also mean respecting "what went right?" and acknowledging how many people did a lot of good, hard work despite the outcome.
reply