Even before internet, people are/were repeating gossips, ideas reaching them through their peers.
“I’ve heard this from someone familiar with X”
The only thing, before internet these gossips were going through certain transformation by diverse individuals, now it is one single transformation by LLM
Was reading Simon Wardley's post on how "AGI date" is now "postponed", and remembered your posts from Twitter, decided to share :)
What is next? Robots by 2030, Quantum by 2035?
Get ready, build your understanding of principles, continuously improve you learning skills.
10/10
You may provide a counterargument saying some of us work on the same technology for decades. Parts of our ecosystem are different, and yes, in parts we still run COBOL. We deal with leveling up abstractions at the frontier.
9/
What's hot today? LangChain, Google ADK, CrewAI? Invest in underlying principles rather than in tooling alone. Principles are easier to combine during the next big reset.
8/
Your meta-skills better allow you to quickly learn and absorb some "crazy" ideas which sounded crazy just recently, but now are becoming the norm. You should never bet and invest in a tool alone, - instead, bet on principles.
7/
You better combine skills and practices into new skills and practices to move faster in the upward direction, the direction of new levels of abstraction, while the ecosystem does that around you.
6/
You better play the game of combination, while the ecosystem around us plays the game of combination. You better explode exponentially along with the ecosystem exploding exponentially.
5/
These skills help you navigate vertically in the exponential explosion caused by ever-rising levels of abstraction. This is what software development is. This is what we do.
4/
Of course, there are skills which are forever skills, they are fundamental. These forever skills are not tool-bound, language-bound, or ritual-bound. These skills are general principles and meta-skills.
3/
Whatever tools you master, languages, practices, or rituals (haha), whatever finally makes you an "expert", it will go away.
There is a Big Reset every five years or so. This big reset somewhat equalizes "experts" and novices.
2/
What Karpathy talks about here is something every software developer discovers again and again and again, every five years or so.
1/N
The entire history of software engineering is one of rising levels of abstraction.
This is as it was, is now, and always shall be.
x.com/karpathy/sta...
post a perfect album from the 90s that isn't nirvana, pearl jam, soundgarden, or alice in chains
Idea for new Black Mirror episode - everyone speaking what they actually think
EIP and DDD come to mind as brightest but there were many many more
Bjorn Bremb’s arguing that most systems in living organisms are neither completely deterministic nor completely stochastic, but in a middle domain: underdetermined
Having said that, I absolutely believe that every step made today gets us closer to ground-up spark, but we are still very far.
10/n (n=10)
So if spark is a product of ground-up development I get it, and I believe in it.
But imposed spark - attempt to make Frankenstein alive and then expect it so live as nothing happened, - I think this is Sci-Fi story.
9/n
Now this ecosystem is not necessarily a real world, - this can be digital. In digital ecosystem with rules, given the ecosystem persists in time and has ways to reward it's agents, there can be digital evolution.
8/n
So instead of being "Artificially" imposed, - such spark must come as a necessity, as a way to solve a problem of survival, and serve ecosystem by being good citizen of energy logistics.
7/n
No logistics to self-patch, self-build, self-heal, adopt to all kinds of environmental changes (energy sources etc.), by rewriting all protocols.
This spark will vanish and decay.
6/n
Now imagine this but about AI.
Imagine we gift the spark of consciousness to some computing device.
But it has no logistics to rewrite "spark" algorithm, or decrypt protocols of it.
5/n
So eventually that device will break, and it will serve as a nice item in a museum.
4/n
But, - that civilization has no capabilities of replicating this technologies - because there is missing knowledge there.
Decades of research behind each material. Algorithms and protocols running on the device having no source code.
Etc.
3/n
Their civilization will be amazed by capabilities of the technology, true.
They will take great care of the device, and will keep it safe, true.
2/n
I like to use this story in my argument about sentient AI:
Suppose we traveled to another part of galaxy and gifted undeveloped civilization marvel of technology - I don't know - plasma gun, or computer which can read thoughts, - or come up with your own example.
Then we leave them to be.
1/n
Again - the paper is nothing sensational, I don't see any groundbreaking work or methodology there
Most surprising in all this is that anyone in the world even considered that statistical model can "reason" the same way we do, and that this is seriously discussed.
2/2
To be honest, I don't get why this paper is so popular on social media.
A lot of famous people are sharing it second day with text similar to "Haha I told you".
I read the paper twice. They ask the model questions and observe collapsing behavior.
We - the users observe this daily.
1/2
It rains only a few times a year in SoCal, and when it does, it's the best sound to fall asleep to.
