Too bad, so sad for the Mister Krabs secret recipe-pilled labs. Shy of something fundamental changing, it will always be possible to make a distillation that is 98% as good as a frontier model for ~1% of the cost of training the SOTA model. Some technology just wants to be free :)
It depends on what "premium"/"luxury" mean to you. For some, red leather that has been masterfully tanned and stitched lining the interior of their car is premium. For others, the ability to transport you hundreds of thousands of miles in any terrain and any conditions with equipment failures that are rare and easy to fix is "premium". Being swaddled in high cost materials while stuck on the side of the road in a snowstorm isn't exactly a "premium" experience.
Likewise, for some, there is nothing premium about a product that 1) becomes a paperweight when a single component fails or is no longer sufficient to satisfy the user's changing desires. 2) Hasn't had engineering time and BOM on high-cost materials devoted to making the device easy-to-repair, or has had engineering resources spent making the device hostile to repair.
Framework doesn't just give you permission to repair and modify their product, they have engineered and designed a product that is easy and intuitive to repair and modify, and made out of materials that are designed and selected to endure being touched and manipulated, one great example that probably comes to mind for many FW13 owners that have opened the device is the touchpad cable finger loop in the FW13.
As any technician or DIY enthusiast might tell you, the materials e.g. Apple uses that you interact with during disassembly aren't exactly robustly made, and there is no sign that care or good taste was used when designing the disassembly procedures. But again, it depends on what you want, for some fragility enhances their experience of an object as premium and they have no interesting in upgrading/repairing their own device so the quality of that experience is irrelevant.
You are a digital serf, dependent on the good will and love of a lord that gives you access in exchange for a tax.
I really wish free(libre) tools existed that allowed you to do your work. Hopefully they will in the future, I am sure someone has tried/is trying to build them.
Reversing demographic momentum isn’t so easy. You have a cascade effect which happens, and high costs of a massive elderly population can’t be borne by a sudden baby boom. Also, through the medical system, etc, you can force people to abort their children, but it’s harder to force them to make children.
Controversial take: Democracy and the US are awful at keeping secrets, and are incapable of winning by an information delta, if we followed your strategy we would surely be doomed. Our greatest advantages come when we work in the open, and share knowledge and empower ordinary people and the world with technology. As things stand, we are funneling our brightest minds into creating proprietary (secret) technologies... And it turns out the only people for whom the technology is uncopyable or secret are... American citizens. The "proprietary" technology is trivial to steal, and legal protections don't matter outside of our borders, the legal protections and subsidies afforded to those building proprietary (secret) technologies only deprives Americans of the ability to innovate, while in peer nations like China, individuals and startups are totally free to use and enjoy American technology without any restrictions.
But that only works if China reciprocates, which they show no sign of doing.
I’d imagine a Chinese citizen living, studying, or working in the US has access to a lot more advanced knowledge than a US citizen trying to do so in China.
Up to this point, the US has been the one with the advanced knowledge. We now face a world where the opposite might become true.
But using the previous example, I’d imagine a future hypothetical American going to China to study or work would face a lot more roadblocks to obtaining and extracting any advanced knowledge, especially anything with strategic importance.
It doesn't require reciprocation because it is a generalized version of the rebasing problem in software.
Over a big round table with cigar smoke in the air it's natural to come to the conclusion that the closed party can always outpace any set of open parties since it can take the public work and extend it with an advance that it keeps a secret.
In reality, we observe that open parties tend to win, or at minimum, if they lose, the closed party tends to have an entirely disconnected line of research that rarely incorporates ideas from the open party. In the rebasing metaphor, the reason for this is the free coordination an open party gets with other open parties. The closed party never gets to insert its advance into the shared state-of-the-art, so it loses all of the free maintenance of coordination, and it has to choose between paying the maintenance cost of integrating its secret advance with the public SOTA, dropping the secret advance and going back to parity with the public SOTA, or disconnecting from the public SOTA and going all hands in on its own ideas. The maintenance burden of integrating your ideas with the constantly moving SOTA may sound trivial but in practice it is usually prohibitively expensive if there are a lot of parties collaborating on the public SOTA and doesn't leave you with much time/budget to find new secret advances.
Right now in the US, we have all of the disadvantages of the open model: the closed parties of the world can cheaply take ideas they like from Meta, Google, OpenAI and mix them with private advances, and all of the disadvantages of the closed model: our domestic tech industry keeps all of its technology a secret from other domestic competitors, and gets none of the coordination benefits of open research / technology, independents and startups are not only unable to access information about the SOTA, but they are actively attacked by the existing monopoly players with any means available when they approach it independently, including using their access to massive capital to drain the talent pool, or being bought outright. And, as we are all too familiar with, the entrenched players don't even care that much about whether or not they can even use the talent efficiently, denying it to competitors is worth more.
> In reality, we observe that open parties tend to win, or at minimum, if they lose, the closed party tends to have an entirely disconnected line of research that rarely incorporates ideas from the open party
An obvious counter-example to this is the NSA/GCHQ and cryptography. They've repeatedly shown that they're a good 5-15 years ahead of everyone else.
Is this still true? I feel like I haven't heard of any crazy cryptography revelations for a while now. My assumption was that cryptography was a bit of a special case because it was only government/defense entities putting significant work into it, up until the Internet/digital telecommunications became prominent enough that there was great individual and private-sector demand for crypto. (Plus the whole mess with it being export-controlled, obviously)
Aside from doubts about whether or not this is actually the case the pertinent question that comes from my point is:
If cryptography researchers were keeping their results secret to within their institution / research circle, instead of sharing with academic community / public, would that advantage or disadvantage the NSA relative to the researchers? I think the answer is obvious, and it's a pretty excellent analogy for the US-China situation.
We haven't always been awful at keeping secrets, see the actual Manhattan Project. I like the optimism of your proposal, but how would those US companies continue the same level of R&D investment without those extra profits? If the government just directly invests, then you've just become the enemy.
The archeologists are already adding fake detail, just at a different level of abstraction. Did they constrain themselves to only painting in the places where they find remnants of pigment? No, otherwise there would be gaps, cracks, and random interruptions of other colors in the painted figures. And there's the guesswork involved in going from spectral analysis (+ other tools) of a pigment sample to an actual paint that could have been plausibly available to the artist.
Reconstruction, (similar to translation) is an art that combines carefuly study of evidence and craftfully filling in gaps and adding in detail where necessary (or leaving details unfilled and ambiguous to communicate the impossibility of total translation or reconstruction!) to present some communicable form of the original that gives the viewer some closer but imperfect access to it.
Whether intentionally or unintentionally, these researchers have cultivated a public perception that the classical statues we admire looked totally ridiculous and were actually hideous. It is difficult to interpret it as unintentional, when the more absurd your reconstruction, the likelier you are to get press attention and get invited to special events at international galleries.
To get closer to an answer to this you should still talk to the people doing the actual work?
I know that many scholars have an uncomfortable relationship to the PR work their research institutions are doing, but they themselves don’t strike me as unapproachable or closed to nuanced discussion. Seems weird to ignore that perspective and wildly speculate from the outside.
Who is inside and who is outside depends on your (subjective) spatial interpretation of the situation.
It could just as well be said that a bunch of scholars disconnected from the culture, history, and technique of fine arts (except as objects of scholarly interest) are wildly speculating from the outside about the nature of the objects, and people interested in these things are starting to ask "Why are these silly things being said about the topic I'm interested in? Are the people behind this pranksters?"
Anyways, if there is a misunderstanding here, which I don't doubt is the case for at least some of the people involved, why can't the discourse be had in public about it? The question has been asked as you suggest...publicly. Polychromic revivalists are free to respond in public, and we can all benefit from hearing the more nuanced perspectives get expressed.
How do you think public discourse spaces are created? By approaching and talking to people when you write about them! That doesn’t just magically happen …
I merely would have expected some humility when you characterize the work of other scholars from the outside without even talking to them. (Outside here is relative. Whenever you talk about scientific of scholarly work without talking to the people who do the work you are on the outside.)
If those scholars don’t want to talk to you, fair enough, probably no humility needed. If you don’t want to talk to them (which, fair enough, not everyone is cut out or wants to do journalistic work) you better be humble and maximally charitable, though.
> Whenever you talk about scientific of scholarly work without talking to the people who do the work you are on the outside
You are ignoring what I said and just reasserting your hegemonic view of scholarly institutions / scientific work. On the contrary, if you zoom out it becomes obvious that our academic research in these matters is ephemeral heat and noise that gets rolled into the dustbins of time.
Your central claim is that scholarly academic (mainstream) work ist disconnected from fine arts and as such outside of it, no fit to give meaningful answers.
That seems like a wild and weird take to me, contradiction everything I know about how the world works. But if that is your hypothesis then I don’t know how you can answer ist without actually engaging closely with those who you say are disconnected.
Gotta love only comparing the model to other openai models and just like yesterday's gemini thread, the vibes in this thread are so astroturfed. I guess it makes sense for the frontier labs to want to win the hearts and minds of silicon valley.
Please don't post insinuations about astroturfing, shilling, brigading, foreign agents, and the like. It degrades discussion and is usually mistaken. If you're worried about abuse, email hn@ycombinator.com and we'll look at the data.
reply