HN user
Yes, sure. I will try to be more blunt, as I was not trying to complain about anything.
I think you have invested with agreeable results, but you missed fundamentals. Adjust course to "architecture first" and I expect a great product.
Group transform is out of order. It does not transform the group but the elements. This leads to the suspicion that position and rotation are not transformation chains on object trees but attributes. That would be the wrong architecture.
I am very sorry, but please explain. Why is this a nice looking Svelte / Three / CSG app, but the basics are wrong?
I also made this experience. As long as the public level of knowledge is high, LLMs are massively helpful. Otherwise not so much and still hallucinating. It does not matter if you think highly of this public knowledge. QFT, QED and Gravity are fine, AD emulation on SAMBA, or Atari Basic not so much.
If I would program Atari Basic, after finishing my Atari Emulator on my C64, I would learn the environment and test my assumptions. Single shot LLMs questions won't do it. A strong agent loop could probably.
I believe that LLMs are yanking the needle to 80%. This level is easy achievable for professionals of the trade and this level is beyond the ability of beginners. LLMs are really powerful tools here. But if you are trying for 90% LLMs are always trying to keep you down.
And if you are trying for 100%, new, fringe or exotic LLMs are a disaster because they do not learn and do not understand, even while being inside the token window.
We learn that knowledge, (power) and language proficiency are an indicator for crystalline but not fluid intelligence
The flaws and strengths of the 'woke mind virus' are the same. It prioritises ethical behavior annoyingly. 'I am more humble than you are!'
Turing a 180° leads to an capitalistic ethic free stance: I am right, because I am strong. This is ironically causally reversed. I am strong when I am right.
Maybe the age of aquarius is still far away. Hey society: That is fine. I can be quite strong to. So fuck off.
I think our billionaire are not grokking what they have done to themselves and it already shows.
Thanks, dang.
I think that using a LLM as the referred telepathy device to a wolfram-alpha/mathematica like general reasoning module is the way to AGI. The reasoning modules we have today are still much to narrow because of the very broad and deep search trees exploding in complexity. There is the need for a kind of pathfinder which could come from common knowledge already encoded in LLMs, like in o1. An system playing with real factual reasoning but exploring in directions coming from world knowledge.
What is still missing is the dialectic between possible and right, a physics engine, the motivation of analysed agents, the effects of emergent behavior and a lot of other -isms. But they may be encoded in the reasoning-explorer. And of course loops, more loops, refinement, working hypotheses and escaping cul-de-sacs.
There are people with great language skills and next to no reasoning skills. Some of them have general knowledge. If you ever talked to them, for a at least an hour freely meandering topics you will know. They seem intelligent for a couple of minutes but after a while you realise that they can refer fact, even interpret metaphors, but they will not find an elegant one, to navigate abstraction levels, even to differentiate root cause from effect or motivation and culture from cold logic. Some of them even ace IQ or can program but none did math so far. They hate, fear or despise rational results violation their learned rules. Sorry, chances are if you hate reading this, maybe you are one (or my English is annoyingly bad).
I love talking to people outside my bubble. They have an incredible broad diversity in abilities and experiences.
I wonder no more why my people, after Habeck announcing more than a uncomfortable winter, are united in supporting the people of Ukraine. He offered suffering for a cause. While the morale is quite different, the social mechanics repeat.