chaorace
Joined 7 years ago
Joined 7 years ago
It just so happens that weather is a phenomenon which affects all people and requires large, distributed, passive infrastructure to effectively manage. It's a classic case where the public option is bound to be more efficient in terms of absolute resource allocation.
On what basis do I assert that it's "bound" to be more efficient? Simple: weather affects the production, transportation, and logistics of virtually all goods. The costs of weather are therefore distributed equally across society regardless of government policy. Government is very good at delivering this specific type of centralized basic infrastructure in a cost-effective way (see also: roads), so if we're all paying for it together regardless this is a no-brainer policy.
The models get trained largely the same way as non-MoE models, except with specific parts of the model silo'd apart past a certain layer. The shared part of the model, prior to the splitting, is the "router". The router learns how to route as an AI would, so it's basically a black-box in terms of whatever internal structure emerges from this.
More or less, the art of making a successful video requires:
* An attention-grabbing thumbnail
* A curiosity-provoking title & premise
* A strong hook which convinces the viewer to put the screen down and let it run
* Editing which delivers the information at an engaging (yet monetizable) pace
* Packaging said information so that it is intelligibly balanced across the mediums (audio/text/video)
* ^^^ Doing this all in a style which still retains enough uniqueness to establish a repeat viewerbase
"The algorithm" is a system for efficiently delivering novel videos with these qualities to the audiences who will most eagerly consume them, which is an essential function for a platform with 2 billion monthly users. For every video on lowest-common-denominator celebrity junk, there's a dozen niche videos tailored to some ravenous subculture or other. Not all magazines are tabloids... but just about anyone can kill time with a tabloid, so that's what leads.
Unlike magazine stands, however, the platform will eventually learn to only show you the thumbnails for videos you'll want to finish watching. It's almost embarassing to share... but here's an example batch of 12 recommendations, almost all of which I'm likely to (eventually) click on and fully watch: https://i.imgur.com/dygfXXb.png
There's a lot of cool diagrams which I'd encourage skimming that link for, but here's the basic rundown: the goal of the described process is to achieve a creamy yolk like what would be produced via sous vide whilst eliminating the unpleasant jammy eggwhite texture characteristic of that process. The recipe involves 30 minutes of carefully transferring an egg back and forth between two vessels repeatedly: one boiling, one room-temperature. You do that 16 times in exact two-minute intervals in order to achieve the "perfect" egg -- very simple and convenient for the modern home-cook in a hurry!
Anyway... you can watch this guy on youtube make it so that you may eat some other, more sufferable meal more vicariously: https://www.youtube.com/watch?v=ahGGanfPDJw
This is a good analogy. Children are the people who they've been raised to become, so it stands to reason that people will give money and appreciation in the same ways that these things are originally given to them. These things are social constructs; we are inevitably taught how to use them by way of social dynamics. This is all to say that people love to support their darlings... but they've been socially conditioned to expect reciprocity in all transactions. That's how the sausage is made in content-based monetization -- you produce the actual product at a loss and then try to claw it back selling high-margin merchandise that nobody'd ever buy otherwise. The merch acts as social permission to finally do your part and pay the creators.
To risk stating the obvious: this is not a good thing and I think the majority of people would likewise agree. People should be fairly rewarded for their work and we should desire a culture which openly and freely encourages doing so. Culture, however, reflects society. The society we've created is transactional, so that's how people frame the spending of their money and efforts -- indeed, "spending" and "transaction" are practically interchangeable in our collective lexicon. Effort isn't strictly scarce in the same way eggs are, however, so we fail to value it.
> I don't know if this is the case around the world (probably is?) and I don't know what the solution is. It just sucks
It's not a total disaster... so we'll undoubtedly continue to ignore the cracks in the foundation. Martin still got paid good money for his efforts and it was good for him for a time. That podcast you like will sell enough t-shirts and get the rent paid on time, at least for a little while longer. This seems to be about as good as we've collectively agreed to make the world for the time being. A local maxima, so to speak: we've gotten stuck asking for more when less might do better. With a bit of luck and effort, however, we can still catch that pendulum when it eventually begins swinging in the other direction. That's my hope, anyway! For the time being I try to do the things I'd like to see become normal in a more decent world -- sharing generously, paying for the things I like, etc. -- because hopeless accelerationism is for chumps.
Conventional business models as currently implemented are fundamentally misaligned to the timescales associated with this product category. Products like these need a level of stability that can only be accomplished at the charter level of the corporation -- it needs to be fundamentally incapable of reneging on promises made.
Without that kind of reassurance, why should anyone trust this service with their links? The exchange is incredibly unequal. They receive full, permanent control of the content, access, and monetization of all things which I cite. I receive... a promise that my links will do what they already do, but maybe last longer.
The most realistic path out of this situation is that homeownership becomes increasingly inaccessible to the point where it hurts more people than it helps (i.e.: when the majority of voters cannot afford to own property). This will take a long time on the macro scale, considering the current 65% homeownership rate and the generational nature of property transfers. The micro scale will change faster wherever unattainable homeownership is already the reality, but it will ultimately be bottlenecked by state-level legislation. City governments lack the resources/authority to meaningfully incentivize developers or otherwise remove obsolete building codes.
Nevertheless, distributed power certainly is less dangerous than concentrated power, isn't it? Inequality is a metric of the overall concentration of such power in our free market society. More inequality means more power in the hands of fewer. They become more like kings each day, when does it stop?
If I tried to do that with man... I'd have to read (in alphabetical order) the documentation for 6 different command flags. That's how far I'd go to read about the flag -- to actually use it, I'd have to experiment with the command flag until I figured out the actual syntax using my imagination.
Let's say instead that I am in a rush... so I'm skipping that time-consuming process and instead scrolling furhter down for some practical examples... one full page later I find a bunch! ... Except there's only 4 and none of them demonstrate what I need (the syntax is completely different). The docs wasted my time exactly when I most needed it NOT to do that. If I'd instead just googled "xargs replacement example", I would have gotten something I can copy/paste in seconds and could have gotten on with my life!
The moral of the story is this: Don't tell without showing. Don't show without telling. Do both if your goal is to be understood.