But isn't the point that the neural nets learn some insanely abstract ideas anyways? So whether the complexity is "ahh we think this neuron does 100 seemingly unrelated things but we don't understand it" or "this actually is some kind of latent shared feature, but we don't know what the concept represents because the composition of the entire circuit is an arcane algorithm", we're kind of stuck, no?
This seems like a classic model-description issue more than anything.
The idea is: a model may only have 100M neurons or whatever. But there are infinite concepts in the world. Therefore, models have to "stuff" multiple concepts per neuron.
How do you do that? Theres two ways:
1. Each neuron handles similar concepts. One neuron for dogs, a separate neuron one for cars, etc.
2. Each neuron handles a bundle of unrelated concepts. One neuron handles dogs and cars. Another neuron handles cats and numbers. And so on.
Historically, people thought #2 was correct because many neurons were found that seemingly represented multiple concepts. #2 had a cool name (polysemanticity) and was promoted in OpenAI/Anthropic blogs so it had a lot of traction.
However, this blog says #2 is wrong. Actually, the dog/car neuron is just a "V shape neuron". So #2 was just methodological error, and #1 was right all along. This is all the blog is saying.
- Quoted phrase in the opening sentence
- Opening quote: “hostile to users”
They use this software to find and kill people. Which they then publicly brag about. How much more hostile do you want to be?
Like immagine if the rhetoric of Turkey was constantly talking about expanding to Greece and how the land of Greece is historically the Turks. Obviously that would make any Greek uneasy.
- Quoted phrase in the opening sentence
- Opening quote: “all you need is SQLite”
But I recently looked into self hosting and several things each wanted their own Postgres container. Immich wants a Postgres and a redis instance.
If you’re building apps targeted at self hosters, SQLite is really all you need.
We can thank Microsoft for their willful support of criminals by having file extensions be hidden by default in Windows.
Just get into a decent general tracker and you'll literally never have this issue again (along with getting torrents with more seeders and better tagged content so your "4k" download will never be 480p again)
Alternatively, if there's a way to have qbit itself mark the download as "Failed" when it detects and refuses to download an executable, that would work too. I don't use qbit personally so I don't know if that's a thing it can be configured to do.
- Quoted phrase in the opening sentence
- Opening quote: “customer service”
Some of those people were mathematicians, engineers, physicists who found joy in being skilled at what they do and were able to be financially rewarded for their effort applying their skill. Are these artistic pursuits, subcultures, pick-up leagues, etc going to pay their mortgage? Will they help them get a job in a field they are trained in?
The "free to pursue more important things" is a hand-wavy statement that doesn't talk at all about what those people should be doing to provide for themselves after wages plummet. The hyperbole is drawing attention to the ridiculousness, not making a comment about there not being a rec soccer league or something.
For example, weavers were one of the most powerful professions (if thats a thing) in the middle ages. In england there are the wool churches(not entirely weavers....), in the low countries, the weavers guild fomented revolutions, and shaped the borders of countries.
But that came to a crashing halt over the space of 200 years. slowly at first, then all at once. the power loom, then cheap imports from india turned a cottage industry into dark satanic mills. Yes there are/were specialists that still hang on, but they are like football stars you can probably name all the people who make money, the rest are barely scraping by.
For us, in the programming world, we will not have a 200 year grace period. Yes, AI produces shit code, but thats not the point. its cheaper and there are vanishingly small places were "good" code matters anymore. AI could be used to help a large range of people, provide support, trusted advice.
However because of the people who are funding it, that will not be the case. it will be used to reduce wages. Just like offshoring, resulting in china being the worlds manufacturing hub, AI, and the current people funding it will aim to make the expensive worker redundant.
In short, I weep, because Altman, and the money behind him ans his ilk want me destitute. I have committed the crime of wanting to earn a wage, and thus I must be punished.
Nothing is validated or tested anymore. It’s all about the shape or believability or even how loud it is.
They dropped 100+ papers no one read. I am sure mathematicians are reading them now but even this forum has concluded that these problems are solved, without reading or validating.
It’s like everyone forgot “don’t believe everything you read on the Internet.” And it makes for a very boring discussion when everyone is just spewing unvalidated claims, layering them with strong beliefs founded on similarly unvalidated claims. It’s dull, without substance and just pure noise.
- Quoted phrase in the opening sentence
- Opening quote: “Artifical Intimacy”
- Quoted phrase in the opening sentence
- Opening quote: “The arc of the moral universe is long but it reverts to the mean”
I guess I'm not astonished that people are so bad to each other in general, but more that in the case of genAI it's so prevalent in places that seemed more elevated previously.
Part of me wonders if this is actually a form of self-defense. The misery of others can provoke sympathy or the opposite: Angry pushback to protect oneself from dwelling on their misery.
Worth reading: Robert Brandom, “Making It Explicit: Reasoning, Representing, and Discursive Commitment”
This is probably the saddest comment I've read on HN. It really shows the state of this place and how far into psychosis the comments have fallen.
- Quoted phrase in the opening sentence
- Opening quote: “honour”
Anyway, what are useful features of Steam:
- their dedication to getting Windows games running on Linux (to see off the existential threat Microsoft poses to them, but nonetheless greatly appreciated)
- community forums / screenshots / artwork / guides etc. for every product. Very useful if the game vendor doesn't have its own, and most game vendors don't
- that Steam Sales encourage vendors to offer games at much more reasonable prices (e.g. a few pounds/dollars/etc.) for what is effectively a license, revocable at any time, not a real purchase. £60? Ha ha fuck off, I'm waiting for the sale when it's £5
- also, as far as I've seen, every game distributed can still be downloaded today. People may complain that publishers update a title e.g. to remove licensed music 10 years later... but you can still download the old version with some trickery. And I've never seen Steam actively remove downloads for existing customers, I've only seen games removed from the store so no future customers
- pirates have found that most games, if their only DRM is Steamworks, are incredibly easy to crack. I am very glad Steam has convinced most publishers to use that alone and nothing worse, and I curse publishers who insist on their own DRM, their own launcher, etc. If Valve goes tits-up, I have a lot of confidence most of peoples libraries will still be playable with a little bit of cracking.
There are other useful features, even if Valve is getting rich and fat off selling gambling and day-trading to teenagers.
Optimizing a project with AI is one of the most trivial things you can do with AI.
I also wouldn't buy it for a second that Valve engineers aren't using AI, especially for something as trivial as "hey claude rank the top causes of slow startup" -> "oh okay now fix". The most I'd grant is that everyone uses AI secretly as a sort of dark forest problem.
Once I had a good benchmark script, Opus 4.8 or whatever made my terminal app faster with less memory than ghostty and kitty. It's not even interesting.
- Quoted phrase in the opening sentence
- Opening quote: “Hello World”
That's not even getting to some of the "missing" features.
No Community marketplace for in-game items
No Direct player-to-player item trading
No Digital trading cards, levelling, and points shop
I hope more application developers "forget" to add these features.
We credited Sapling because that's how Meta itself externally refers to the entire system, including the Mononoke server, and we wanted to follow that convention. We will add an additional footnote to clarify this. [edit: this is done]
We were inspired by many things while building Scion, such as the great ideas in Mononoke's Bonsai data model, and didn't intend to obfuscate that in any way. We were also inspired heavily by https://martin.kleppmann.com/2015/03/04/turning-the-database..., for example, which also describes the "derived data" concept using the same words.
We also did not just tell Claude to build us a version control system; we did a lot of research on what would work and not work for our use case and coming from an existing large Git repo (rather than Mercurial, as Mononoke did) required many novelties beyond some of these data structures.
The intent was to give credit to Mononoke without getting too deep in the weeds of comparing and contrasting each point of the data model where we diverged. There are many ex-Meta employees at Anthropic who were inspired by how joyful using source control was because of Mononoke, which deserves a lot of credit to you and the source control team there. We had that experience as a north star from the beginning while making decisions that map to our (different) needs.
I miss the 90s chunkiness of switches, buttons, hardware: being able to hear the computer "think" (HDD seeks), the hum of a CRT. Modern machinery feels so insubstantial, emphasizing clean glass and tidy machined chassis. (chassies?)
I have this persistent daydream about a computer setup that feels like the Nostromo flight deck from Alien. I almost want to feel like I'm sitting _inside_ a computer when I'm at my desk, surrounded by powerful feeling electronics with substantial switches and such. Of course there's plenty of inconvenience there too, but something about it appeals to me.
I think it does, but for new users only. I hate playing the "is-this-a-button" on modern designs, not knowing what is clickable and what is not. Other games I hate:
- Lets find the scrollbar
- Guess which window has focus
- (Windows-only) Try to predict the application to get focus on alt-tab, then alt-tab again.
- How do I find the file that just downloaded?
I don't know why anybody would. It takes all the lessons we learned about UX from user testing in the late 90s/early 00s and throws it all away to produce something that looks "clean" and "sleek", at the cost of being useful. Discoverability? Who needs it?
EDIT: Though I do enjoy skeuomorphism, I do think the TFA goes a bit too far.
I run a small pro audio company, we make guitar pedals and audio effects mostly. My current key objective is building up a processor similar to the Eventide H90, that would compete on having top-notch algorithms for reverb, delay, chorus etc. we already have single-effect pedals on the market that are very successful.
In the past 72 hours I have ONE-SHOT decompiled 3 of my main competitors products, including the Strymon Big Sky. I just... have their award winning algorithms now. Like, I could in theory just stick them in a box and resell them. It took one prompt and 3 hours for opus 5.5 to work that out.
The lexicon 224 then got added. While Claude was prepping It found someone else had already done the same thing 1 weeks earlier and dumped the result on GitHub, so it just stole that code and saved some time. It took all of 20 minutes to do the entire Ursa Major product range.
And yes, I've tested them all - and they all sound phenomenal. I specifically asked Claude to first build an emulator and execute the original firmware/ROM so I could be certain it wasn't just making "something similar" but actually executing the same code. Then I asked it to convert it into a self contained c++ library and compare the result to the emulator.
The problem is, everyone can now do this to my products as well. Basically, anyone with a Claude subscription can now rip of any best-in-class algorithm from any competitor. While I'm not going to do that, I know there are 30 Chinese companies that happily would.
The moat of hardware effects processors has just vanished.
- Ordinary term used more than once in the same comment
- “Call Fred.” appears 3 times
- “You'll need to unlock your phone for that.” appears 2 times
"Call Fred."
"You'll need to unlock your phone for that."
"Call Fred."
"You'll need to unlock your phone for that."
"Call Fred."
"OK, calling Fred."
Phone is locked in my pocket the entire time. Like, make your damn mind up. Every time I try to call from the car now I have to ask it at least twice, usually three times.
And a completely baffling one recently, not even anonymizing names, also in car:
"Call George."
"OK adding L.S.D. by Hallucinogen to your Liked Songs playlist."
WTAF.
- Ordinary term used more than once in the same comment
- “radical” appears 3 times
Not to say this "radical" behavior is helpful, and I understand why people don't like it.
My default presumption, without checking with them directly, is not only that it would not be allowed, but that this is also very much not a "better to ask forgiveness than permission" situation either. Even just asking the question can draw attention from them you may not want.
The ready accessibility of the land is advantageous in some ways, but as neighbors the state can be quite nasty. Your other neighbors don't tend to have anywhere near the regulatory firepower to make sure you comply with what they want.
If you wouldn’t even pay for it yourself, as pretty much the sole beneficiary and someone who could afford it but would rather the money go elsewhere, why would it make sense for other taxpayers to pay for it?
Isn’t the premise of “you may as well own that land” the entire appeal of living next to public land anyway?
Turns out when battery voltage gets low (but not yet dead) this can happen in a circuit. I swapped out the batteries and our ghost never came back. Not surprised to hear it happening in a store where stock just sits there for a long time.
Pretty creepy though to hear that when you are lying in bed
When I actually go to build anything, I generally do pencil and paper shop drawings, usually 3 views, hidden lines showing joinery, etc. I'll make exceptions for simple projects, but for anything even moderately complicated I do shop drawings. "Moderately complicated" starts at a side table with a drawer, and that's something I've put together more than a few of.
The important output of the drawing process isn't a picture of the finished piece. It's knowing, deeply, how all of the pieces are going to fit together and how they relate to each other in space. It's having an order of operations and assembly in mind. It's knowing where the pain points and hiccups are likely to be. It's a good cut list with notes about e.g. what the shoulder distances are for anything with a tenon on both ends.
I don't doubt that you could accelerate the 3D modeling process with AI or a sufficiently friendly CAD program, just as you could take plans from Fine Woodworking or Popular Woodworking and go directly to the shop and start building. I do doubt that things would go faster or more smoothly in the shop for having done so.
[0] Which depends largely on the complexity of the project. This gossip bench was simple enough to do in SketchUp:
https://www.longwalkwoodworking.com/gossip-bench
This dining table required some serious leveling up of my FreeCAD skills:
https://www.longwalkwoodworking.com/mid-century-modern-table
My experience is that if there's a lot of shaping with curves and tangency, FreeCAD is less cumbersome to draw it in and a hell of a lot easier to make changes in.
This is funny to read. Either you can be more "verified" than the other guy, or it's badly implemented.
(The bookstore I worked at paid low, but had one perk I loved: they let us borrow some of the books so we could learn what was good and give advice to customers. I took the job mainly for this reason, as I was really into 3D modelling and animation and thought I could teach myself. Alas, my old 486 PC could barely run the demos.)
That week I went to my bank to see if they'd give me a loan at better rates than the dentist, and to my shock they could, and would lend me up to 5k as a personal loan, and they said I could use it for anything. I went for it and got the loan.
I didn't go back to the dentist though. Went straight to a computer store and bought one of the latest Pentium PCs and my first ever dedicated GPU (a Matrox Millennium), then stopped at the pharmacy and bought the largest pack of ibuprofen they had.
I got a dream job doing 3D modelling and animation some months after that, transitioned into multimedia programming and have been a software developer ever since, so the ROI was pretty good. And I still have all my wisdom teeth.
https://asset.cdocs.com/uploads/forum/attachments/258722/Rea...
It's a fascinating read that basically confirms all of these anecdotes.
This corporate structure leads to "overdiagnosis" and vastly more complex procedures than necessary
Related investigations:
[0] https://www.oralhealthgroup.com/oral-health/u-s-investigatio...
[1] https://www.ovid.com/journals/jaim/abstract/10.1001/jamainte...
Reading the reviews on Yelp, someone had said they had a similar experience with their first visit and went to a different dentist and were told they had none.
I decided to get a second opinion and heavens am I thankful I did. They replaced my filling right then and there and I think I've had maybe 3 cavities in the 15 years since.
I put very little faith in dentists to not be scam artists.
The part I find particularly … funny is the disagreements. I had one dentist who insisted I needed to be using a pre-rinse. That dentist ended up leaving, and a different dentist at THE SAME OFFICE told me it was a waste of money and I should stop because it does nothing. I swear they all just make it up as they go.
I've had multiple dentists tell me this rough tooth in the back of my mouth needed to be capped because it was "falling apart". I started using imported NovaMin on it out of curiosity and it became super smooth instead of rough and no one has mentioned it since.
One dentist I saw for a number of years would try to insist I needed a dental implant for a missing molar and if I didn't all my teeth were going to come out alignment. I didn't have money for it at the time. Not a single dentist before or since has mentioned this.
I find _very little_ "professional curiosity" exists overall. Heck, I don't even always see it within our industry where I think it's maybe more common? Maybe I'm biased.
s/dentists/doctors
I'm reading up on new tech stuff all the time and it shocks me that most of the doctors I've ever interacted with don't seem to be doing the same. It's like they are just coasting on what they learned in school. And since the vast majority of us only see 1 doctor of a given type at best once a year (maybe more for the dentist, but the same dentist) we never get any sort of "second opinion".
Think about all the people you've worked with in tech, now think about them being the sole-decider for how to architect/build something without any pushback or anyone else to push back or give advice. Yeah, that's how I imagine the whole medical field.
Sure, we'd all love to have "Dr House" and his team examining us (or insert your favorite tv doctor/team) but I have a feeling that the 5-10min I see my doctor plus another half hour at max to read my charts/respond to my messages is all the time he thinks about me for the whole year. <1hr of attention and I'm supposed to expect him to pick up on issues that aren't obvious in my blood work or other measurements he takes? It's bleak.
It is why I never use "sign in with <google, meta, apple, whatever>"
---
The author has strong incentives for the success of passkeys.
1. Scott Helme is founder and CEO of Report URI, and Companies House records his ownership at over 25% and up to 50%. Report URI markets a "Passkeys Protection" offering, giving him a commercial interest in services for passkey adopters.
2. His WhyNoPasskeys.com campaign pushes websites to adopt passkeys, is sponsored by Report URI, and promotes its passkey protection services. The "Passkeys Are Not a Trap" article also carries a Report URI sponsor banner.
3. Troy Hunt, Helme’s Report URI business partner and investor, joined 1Password’s advisory board in 2020; Helme has also appeared in Hunt’s 1Password-sponsored content. That establishes an indirect vendor connection.
I can’t adopt Passkeys if I’m not certain they’re going to work as well as my password manager and Authenticator app, which is to say 100% of the time.
When sites ask to set up passkeys, it’s usually native passkeys (especially on iOS). And figuring out how to get the flow to set up a LastPass key by default seems unusually difficult.
I also use a MacBook, windows desktop, an iPhone and iPad. I use Firefox, edge, brave across these devices.
When I say “passkeys are a trap” they are a usability trap where no one can attest with certainty that I won’t wake up in a painful situation because I decided to adopt passkeys.
The idea and protocol are pretty solid, though we're back to a "biometrics as password" problem in many device flows, but every time I'm urged by a site to convert to a passkey, I wonder "is this site the sort that would screw this up?"
So far, it's about 30% of sites screwing it up...
- Quoted phrase in the opening sentence
- Opening quote: “Passkeys are a trap”
It is plainly obvious that the worry isn't attestation per se, it's the restrictions that attestation may in the future require when they turn the dial on restrictiveness up, that was extremely clear to me from reading the first article. If you can get attestation without those restrictions, "sure, why not", and a Yubikey seems like a way to do that which could avoid those restrictions. Most of this article makes sense, but this particular line makes me wonder how the author got it so far off the mark.
They also did nothing to quell my fears that they will in the future use passkeys to prevent people from using software they can modify themselves by saying "...the guidance for consumer sites has been to not even request attestation for years now." as if everyone always followed the guidance, and as if there aren't tens of thousands of managers that think their case is special and they really do need the extra control because they're a bank or handle copyrighted content or have PII or PCI and feel extra risk-averse today.
I don't know if they failed to do so because they can't (and this is a valid fear of the original author) or if they just are bad at reading the room at what the concern is and vastly overestimate how much people trust companies to always have their best interests at heart.
What I am getting from this, though, is "If you authenticate with a yubikey then that's attested, so they can't use attestation through passkeys to make sure your OS and browser aren't modified", and I'm not sure if that's the correct take away or even a correct statement.
Am I missing something? "Attestation tells websites what the authenticator is but it doesn't tell the website what the authenticator is".
The entire argument in this blogpost amounts to "nobody would ever turn on passkey attestation because it would require migration of people who already setup passkeys" which is a questionable argument- you could trivially imagine allowing unattested passkeys for existing users, but require it for new users. Most websites don't yet support passkeys to begin with, so this argument doesn't hold for the majority of websites.
The "biggest providers" bit at the end is odd too because the biggest providers are apple/google- who, as far as I know, do support attestation.
When a website prompts me to set up a passkey, depending on the device I happen to be using at the time it could end up going to Google password manager, Samsung password manager, iCloud Keychain, or Windows somethingorother.
What happens when I go to the same site on a different platform? What if I can't use cross-device authentication because I'm on a locked-down work machine or one that simply doesn't have bluetooth? What if I want to consolidate them all in one platform just for the sake of being organised? What if I lose or simply don't have the device with a certain passkey near me?
I don't have the time or patience to work out any of that, which might absolutely be a 'me' problem, but TOTPs are pretty sodding frictionless in comparison. Scan QR in my chosen manager; type number into website; done. Why are passkeys better?
They were complete no-ops and a waste of time. Faced with that same situation today I would simply lie about whether or not a treatment plan was put into place.
Second bit of advice is to try as best you can to go into clinical trials with an understanding of the landscape of disqualifying treatments. Any immunotherapy treatment, for example, will disqualify you from nearly all subsequent clinical trials for immunotherapy.
Lastly the thing I would recommend for partners of loved ones in the crucible, don't get too distracted by trying to 'save the patient' that you miss opportunities stay connected with them in the process. It's not your responsibility to save their life, as much as it might feel that way when you're in the thick of it.
- Ordinary term used more than once in the same comment
- “score” appears 2 times
One thing I would mention:
You are not dealing with a monolithic entity with regards to a hospital.
There are really multiple groups with oftentimes competing interests e.g.
- case manager wants you out and home b/c that is the best "score" for the hospital. They also DO NOT want you to qualify for Medicare if you are a senior b/c, again, "score" gets dinged if Medicare gets involved
- surgeons want you out as they don't want you to pick up hospital infection. They want you to go to a rehab facility as that will more likely lead to a good outcome
- the hospitalist wants you to be stable in the short term so will pump you with whatever medications are necessary to get you "back to normal" without addressing any root cause
- the nurses generally care about you as a person as they deal with you the most but oftentimes have the least amount of say over your case
- in one of those perverse incentives/unintended consequences: if you do TOO well on a physical ability test, that might disqualify you from certain coverage even if you are below the threshold on other ability tests. That can have gigantic insurance implications.
This is prompting me to write my own version of the OP blog post.
- Quoted phrase in the opening sentence
- Opening quote: “reality has more detail than people realize”
- I still have a mortgage on a property but don't live there so the mortgage bill gets sent to a different address (b/c rental property)
- Company A originated the mortgage and sells it to Company B
- Company B completely ignores the mailing address and starts sending bills to the original property address
- only through blind luck did I realize that the mortgage bill was being unpaid
This seems like basic, basic bookkeeping but I guess it's also a reminder that "reality has more detail than people realize" and that there are plenty of opportunities for improving systems.
- Quoted phrase in the opening sentence
- Opening quote: “unknown”
I’m trying to say that the logical conclusion is to shift the responsibility explicitly to SpaceX, since otherwise you have to admit the attacks are state sponsored and you have to name the perpetrator.
And yes, there should be an appropriate law response to the hacking incidents. If for nothing else, to 1) not set precedent of inaction, and 2) get full logs of what happened instead of relying on the labs' "kindness" to offer the information they want.
But also, the agents absolutely went rogue.
Anyone skeptical should go read the METR report [0].
The Hugging Face hacking incident was a shot across the bow. And in hindsight possibly the best thing to have happened in this era. If their security was tight enough, we would still be in the dark. So we (outsiders) should rejoice that it wasn't, that one time.
We can push for accountability for the labs, while also not be in denial of the power and nature of the new models.
[0] https://metr.org/hugging-face-incident-report-aug-2026.pdf
- Ordinary term used more than once in the same comment
- “Rogue AI” appears 2 times
The normal laws, since you are still liable for everything you own.
"Rogue AI" is solved by removing the accountability sink "it was AI, not me!"
It's your chat, your prompts, and all output that you should read and understand.
Don't know how to operate a power hammer? don't care, any problems it causes are still your fault.
No other big-tech company goes to such great lengths to deanonymize users and make its services difficult to use, especially during account creation.
The requirement to have a mobile device to scan a QR code to verify that you're human. Reverse SMS verification, where you have to send an SMS rather than receive a verification code. You think of signing up using a VPN server in another country? The phone number you use to send the SMS from must also be from that country.
Not being able to watch YouTube videos when using VPN - without signing in to a Google account, all under the guise of "bot protection."
Gemini not working in 3/4 of cases when using VPN.
It's essentially almost impossible to create a Google account without giving up your identity, and your phone number often leads directly to your real-life identity.
I'm not sure what happened in recent years, but they definitely switched their strategy to maximize data collection and make sure privacy-savy users are pressured via technical solutions to reveal their identities.
PS. The first company federal investigators often go for data is often Google
Should the federal government be tracking our interpersonal relationships?
Should an immigration officer be privy to the timing and manner of communications that I, a citizen, have with other citizens? Why?
Why should ICE be able to collect this information when domestic law enforcement generally has to show probable cause to a judge to gather similar information? Who decided to make ICE the country's supreme law enforcement agency?
Related, ICE is arresting citizens on fake immigration charges[1]. Some US citizens have been deported.
US citizens "can't" be deported, but ICE has supralegal powers. Should we have a law enforcement agency that is not bound by our laws? Why? What guardrails does it have if not law? Who defines those guardrails?
1 - https://projects.propublica.org/american-citizens-detained-d...
The article faults LLMs for not being more creative (riding in the city in the rain, in East Asia, in outer space, etc.). But, if prompted, "Generate an SVG of a pelican riding a bicycle," with no additional details or specification, then I would _expect_ something generic: a side view, nondescript terrain or perhaps terrain associated with a pelican (i.e., beach), little clothing (because birds don't wear clothes) with the possible exception of something strongly associated with bicyclists (a helmet).
The article talks about "the LLM diversity collapse problem," which makes it sound like LLMs are getting less diverse over time, but the linked benchmarks suggest that newer, more capable models are _more_ creative.
Out of curiosity, I tried a couple of prompt variations in Opus 5.5. "Be creative" did a beach scene but added fish to the bike basket and a lighthouse in the background - not too creative. "Pick a unique and creative variation of this concept" resulted in a "midnight circus act" of a pelican riding a penny-farthing on a tightrope while balancing a fishing pole with fish (and, this time, two lighthouses - the tightrope is implausibly strung between them).
- Quoted phrase in the opening sentence
- Opening quote: “If only I held!”
They don't look at 3DFX, Transmeta, Centillium, Virata, or the numerous other companies because they are dead in the graveyard of failed companies that have lost full valuation.
Depending on how much you invest in each, if you panic-buy/sell, etc changes all that. You're also ultimately betting on whether or not there will ever be another bubble/hype in the industry, and that you are betting wide enough into fully speculative companies like Nvidia instead of just aiming for power players. People forget that the market for luxury PC hardware barely existed at the time and was populated by a very small group of wealthy consumers. Expecting that to take off against the cheap prices of console offerings and against the technical know-how of assembling/disassembling PCs is closer to faith than any rationale.
Just by law of probability and total number of actual bagholders, the vast majority of people didn't bet that way. But I guess if you think you're special or have some lucrative insight into the industry, you might think you're an exception to that rule.
https://www.eetimes.com/record-orders-for-chip-gear-push-sem...
> Capital equipment suppliers have been increasing their forecasts for revenue growth in 2000 as chip manufacturers begin to add new wafer fabs and backend assembly plants. Last week, for example, Applied Materials Inc.–the world's largest semiconductor equipment supplier–announced it had increased its forecast for industry growth in wafer fab equipment sales to 46% from a projection of 19% at the end of last year (see Feb. 16 story).
> The strong wave of new orders “confirms the strength of the current market cycle,” said Stanley T. Myers, president of SEMI.
This time it's different.
For AI, there were top down directives to find new ways to experiment with and incorporate into core products and processes, but as you went down the management chain, management became more conservative. First and second line managers were strongly incentivized to keep their products, roadmaps, and interfaces stable and predictable by similarly conservative stakeholders. And as such, they were generally unwilling to officially dedicate much effort to AI other than the minimum “look at how much we’re using Claude”.
In particular due to the lack of central tooling, any team trying to do much with AI essentially had to reinvent much of the stack themselves. So even small projects were largely untenable due to inevitable required scope creep.