We shouldn't worry about that. We should be grateful about the mountains of technical debt, because that will be the social security of the post boomers generations: working till death fixing the last hurrah of the boomers.
7,332 karma · joined January 26, 2021
I don't know why the taxpayer should subsidize fintechs.
In reality, USPS is starved of resources, and like with all other open source things, some billionaire would make millions out of it without contributing anything back.
We will can them scribes.
Of course they have been using our data all this time.
You don't want all you eval work to go to waste just because the discount inference provider you're using decided AcmeModel4.23 is as good as the Anthropics GilgameshEpic7.2 that you used to validate your solution.
We already have trouble enough with model retirement.
The distance between the usual consumer/b2b tech startup and the kind of company that sells software to the air force is even bigger than the distance between Google and the kind of company who hires Accenture.
This is something that requires an ethos completely different from how amazon works and the way jeff bezos sees the world.
By not limiting WIP they are actually making their employees less efficient.
It is not a demerit, actually I think that it is a very elegant outcome how many problems could be solved by generation and this probably show us how language is a far more important thing than we ever thought it was.
The "baby steps" are not there just for intelectual satisfaction, the baby steps give you the building blocks to understand things that become progressively more complex. The baby steps are tools, that once you get familiar with them, you'll find new and new uses everywhere.
A good practical illustration of this: It is pretty hard to figure out how differential equations can be useful if you don't understand what they are. And yet, I bet that on january of 2020, there was at least one math teacher deep down in the amazonian jungle playing with a compartimental model based on a system of ordinary differential equations, to figure out the diverse scenarios this new epidemic could play out if that scary new virus escaped China and those few cruise ships.
Another example: take the concept of backpropagation: the underlying idea behind it is the chain rule from calculus, used to work backward through a computation and calculate how each parameter affects the error. Now, it is way easier to understand it in a small CNN that you are training to recognize handwritten digits than trying to understand it in the context of a full transformer model. A digit classifier is something that is way easier and affordable for you to fiddle with stuff and develop and intuition on loss functions, gradients, backpropagation and overfitting.
In the second world war, Germany had impressive science and technology, but, as Stalin said, quantity is a quality in itself, and The Third Reich found out that it was an impossible task to get the metals and fuels for their advanced jet engines, rockets and U2 boats, when between then and those resources there were masses of tens of millions of under-equipped and malnourished soviet soldiers fighting what was an existencial war for them.
Who would know that optimizing for the next quarter would also not be a recipe for long term value creation in arts too?
Now, compare it with my own situation living in a medium mountain city in Brazil. Whenever I wanted to read a book published in the US, I had to go to a physical book shop in a major city at least some 100 miles from my city, and pay for each book, at least twice (probably more) that now I pay a month for o'Reilly.
Of course I would read each book cover to cover, and come back to it multiple times, until I had to start doing things like taping the cover back to the book.
The world should not be made more vulnerable just because someone in Idaho would be pissed because if they can't run Photoshop 2.0 anymore in a newer version of an operating system.
Then, after ruling your city, you could start exporting a few crows to neighboring cities, so they could recruit in your behalf.
After your first state, the whole country, then Canada, central america, western europe. Freight liners full of your crows' Army crossing the oceans. Enforcing non-fly zones, eliminating political enemies, spying, policing your domains, directing and supervising human workers.
Command a few AI data centers and a few research labs, put then to work 24/7 on neuralink interfaces for crows, so you could have AI-enabled, starlink connected squadrons of crows.
Then, an accelerated genetic improvement program, CRISPR, cloning, all the stuff. AI-Enabled-Starlink-Connected-Super-Crows. At your command!
Frankly, you should work more on your ambition levels.
First, it has a canon, every rationalist ends up referencing Yudowsky, arguments frequently refer back to earlier ones, and to make matters worse, by shorthand ("Molochs", "The Litany of Tarski", "conflict vs mistake theory")
Like the Talmud, rationalists like to use extreme hypotheticals to test principles. The contrivance of the situations is the point, as if stress-testing a rule by applying it to extreme situations to distil it to a single variable problem.
Both writing traditions keep opposing arguments in full. The Talmud records minority options and the full log of disputes that went for centuries. Rationalist's steelmanning is basically the same.
It is not surprising that Yudkowsky is Jewish and grew up Orthodox. A lot of other very important figures are also Jewish like Scott Alexander, so, it is plausible that they brought a lot of the disputation-as-study and commentary-on-commentary traditions with them and set the tone for the community.
Of course, I am not directly comparing the Talmud to Rationalist writing. The Talmud is ultimatelly a legal document, its hypotheticals feed a system that tells people how to act, compressing centuries of legal precedent, this is by nature dense, and this density makes it hard to read more than simply verbose. Rationalist blogs are often long due to a lack of professional editing.