1. Efficient recursive transform of kv embeddings into polar coordinates
2. Quantize resulting angles without the need for explicit normalization. This saves memory via key insight: angles follow a distribution and have analytical form.
Could add a scheduled GitHub action that compacts long history into a vector db and then have the agents also check that vector db in addition to md and git history.
The world is full of Juliuses. And if one works with enough people one can suddenly realize that they too are a Julius relative to someone smarter and more introverted. Worth considering this before dismissing someone as yet another Julius. Oh and everything doesn't suck.
> And if one works with enough people one can suddenly realize that they too are a Julius relative to someone smarter and more introverted.
No, Julius is not a spectrum. There is a line between being one or not being one. It’s not just a slider between “socially outgoing” and “technically competent”, it describes a particular type of individual.
> Oh and everything doesn't suck.
I think it was pretty clear I didn’t mean literally. Obviously the Sun doesn’t suck, nor does water, nor do an infinity of things which humans could not have as hand it.
Where does the machine begin and end? Even a fountain pen is a highly advanced mechanism which we owe to countless generations of preceding, inventive toolmakers.
Fountain pen is still more or less the same tool as the lowly stick left partially in the campfire. It is just packaged more cleanly perhaps. It is not drawing for you or writing for you.
My point was that humans are very connected to and identify deeply with their tools. Probabilistic autocomplete we are so excited about these days is just another slab on a deep stack of abstractions humans use to interact with the world.
A stick and the campfire are also tools that do not pre-exist. Just try to make a campfire without a matches or try to make a stick without a cutting tool. Also try to write the next great novel using a stick and a campfire instead or a fountain pen. Tools that are available become the defining factor of the great works a generation can produce. Nothing is different this time.
I feel for the author. Until recently it used to be that writing was a way for humans to project their thought into time and space for anyone to witness, or even to have a conversation. Oh how I miss that dead art of having a good one.
It used to be that you knew where you stand with colleagues just from how they write and how they speak.
Had this Slack memo been written by someone who just learned enough English to get their first job? Or had it been crafted with the skill and precision of your Creative Writing college professor's wet nightmare muse?
But now that's all been strangely devalued and put into question.
LLMs are having conversations with each other thanks to the effort of countless human beings in between.
God created men, Sam Colt (and Altman) made them equal.
I have a vision of some future advertisement going more-or-less like so:
Exec A: Computer, write an email to Exec B, to let them know that we will meet our projections this month. Also mention that the two of us should get together for lunch soon.
AI: Okay, here is an email that...[120 words]
[later]
Exec B: Computer, summarize my emails
AI: Exec A says that they will meet their projections this month. He also wants to get together for lunch soon.
In my vision, they are presenting this unironically as a good thing. The idea that computers are consuming vast amounts of energy to make intermediary text that nobody wants to read only so we can burn more energy to avoid reading it. All while voice dictation of text messages has existed since the 2010s.
It gets to the basic question... what is the real point of communication?
Exec B is too busy gorging their brain on the word salad I am feeding it through her new neural link. But I now have just upgraded my body to the latest Tesla Pear. Want to meet up? Subscribe for a low annual fee of..
The consistency is that most left-wingers are pro environment and anti-corporation. So it makes perfect sense for them to oppose generative AI, which serves to enrich corporations and harm the environment.
American tech bros come up with democratizing tech every five years. And then oops, now the tech has enslaved us or just made the tech bros rich at the expense of everyone else. Oops.
The thing with “left wing” positions is that it depends on the conditions you live under. It does not depend on, like tech people get so incredibly tunnel-visioned about, the tech in isolation. You embrace and use the mills if you collectively own them; you smash them if they are being used against you.
I won’t claim that you are on the side of the billionaire tech bros. I don’t know if it is intentional.
I think this marathon attitude of not trying to win but to hit personal targets could be applied in other areas of life and in other sports even with clear "winners". None of this is about the destination. We all arrive at the same one.
Python's stdlib is not really something to brag about. Very inconsistent, multiple ways of doing the same thing, a lot stuff in there people recommend against actually using...so much cruft acquired over decades.
It's why projects use so many packages in the first place.
Indeed. The standard library was primarily designed in, and for, an era where you couldn't just download multiple megabytes of third-party code on a whim.
Why do you think what you describe being excited about does not warrant the current level of AI hype? I agree with your assessment and sometimes I think there is too much cynicism and not enough excitement.
the current level of AI hype amongst a lot of people, but especially investors and bosses, is that you can already give an AI a simple prompt and get it to spit out a fully functional, user-ready application for you. and we're so incredibly far off that.
the things that AI is able to do are incredible, but hype levels are just totally detached from reality.
> is that you can already give an AI a simple prompt and get it to spit out a fully functional, user-ready application for you.
But it can already do that. Isn't that the whole "one-shotting" thing?
The problem is, of course, that it won't be optimized, maintainable or have anyone responsible you can point to if something with it goes wrong. It almost certainly (unless you carefully prompted it to) won't have a test suite, which means any changes (even fixes) to it are risky.
So it's basically a working mockup generator.
I am so, so tired of "semi-technical" youtubers showing off new models with one-shots. The vast majority of actual devs who use this stuff need it to work over long-term context windows and over multiple iterations.
The thing is, we've already had "working mockup generators" — a.k.a. prototyping tools — for decades now.
If you come at the problem from the direction of "I draw a user interface; you guess what it's supposed to do and wire it up for me", then all you need to solve that problem (to a first-order approximation) is some plain-old 1970s "AI" heuristics.
The buzz around current AI coding prompting seems to be solely generated by the fact that while prototyping tools require you to at least have some training as a designer (i.e. understanding the problem you're solving on the level of inputs and outputs), these tools allow people with no experience in programming or design to get results. (Mainly by doing for UIs what genAI image/video tools do for art: interpolating the average of many ingested examples of how a designer would respond to a client request for X, with no regard for the designer's personal style†.)
† Unless prompted to have such regard... but if you know enough to tell the AI how to design everything, then you may as well just design everything. Just as, if you know art well enough to prompt an AI into developing a unique art style, then you likely know art well enough to just make that same art yourself with less effort than it takes to prompt and re-prompt and patch-erase-infill-prompt the AI into drawing what you want.
from what i can tell, the one-shot thing only works on youtube.
you might produce something that looks usable at first, but the actual application functionality will be significantly broken in most ways. it maybe works enough to do a demo for your video, but it won't work enough to actually distribute to end-users. and of course, as you say, it's not testable or maintainable in any way, so fixing what's broken is a bigger project than just writing it properly in the first place.
Because to justify the current hype and spending, these companies have to have a product that will generate trillions of dollars and create mass unemployment. Which they don't have.
reply