A thread for unboxing AI
The rapid progression of AI chatbots made me think that we need a thread devoted to a discussion of the impact that AI i
chez,
I've been thinking about this quite a bit lately, and I'm not convinced that a great ****ing off is in the cards unless or until we fully digitize consciousness in a way that allows humans (or what we used to call humans) to ditch our physical bodies.
The hurdles to maintaining physical meat suits in deep space are immense. Life has evolved on earth for millions of years in ways that are specific to earth's gravity, atmosphere, radiation shielding, etc. Terraforming is far, far, far more difficult and time-consuming than sci-fi movies would have us believe. And most rocky planets that might be remotely suitable for terraforming or immediate habitation are many light years away. If a great ****ing off ever occurs, I'm guessing that it will involve a digitized consciousness that has no need to be intimately connected to a particular planet.
It's possible that we could be forced to make the attempt in a serious way before we have transcended our physical bodies, but I suspect that such an attempt would have a super-high risk of failure and that any definition of success would involve the death of most humans.
Terraforming is pie in the sky. Gravity etc are a non issue.
If it wasn't pointless I'd bet a decent amount there are people alive today who will live to see people who will never visit earth.
Terraforming is pie in the sky. Gravity etc are a non issue.
If it wasn't pointless I'd bet a decent amount there are people alive today who will live to see people who will never visit earth.
Whether we can simulate gravity, shield ourselves from radiation, etc., is only part of the equation. We almost certainly can. But leaving the solar system is another issue entirely. Resources in deep space are meager, probably too meager to maintain hundreds of thousands (much less millions or billions) of human bodies for hundreds of years until you reach a more resource intensive area of the galaxy.
Low entropy energy will be available. With that self-sustaining ecosystems is errrr what we have now. There are significant but highly tractible problems. They are not close to show stoppers.
We dont even have to leave the solar system . But we will and it wont matter a damn how far away the next sun is. We will take our home with us. May take a while to get to hundreds of millions or billions - if that's the disagreement then it's arguement over.
Whether we can simulate gravity, shield ourselves from radiation, etc., is only part of the equation. We almost certainly can. But leaving the solar system is another issue entirely. Resources in deep space are meager, probably too meager to maintain hundreds of thousands (much less millions or billions) of human bodies for hundreds of years until you reach a more resource i
Hundreds of years? Isn't the closest solar system over 4 light-years away? How could we travel that far over hundreds of years? I don't know much about this stuff, but a quick search says the fastest a spaceship has ever traveled is 430,000 mph. 4 light-years is approximately 24 trillion miles, and traveling at 430,000 mph, it would take around 56 million years to get there.
Hundreds of years? Isn't the closest solar system over 4 light-years away? How could we travel that far over hundreds of years? I don't know much about this stuff, but a quick search says the fastest a spaceship has ever traveled is 430,000 mph. 4 light-years is approximately 24 trillion miles, and traveling at 430,000 mph, it would take around 56 million years to get there.
Time dilation but a 1 way trp.
Wouldn't that just speed things up from Earth's perspective? Those on the ship would experience the full 56 million years, but it would be a shorter period of time for the rest of us?
Hundreds of years? Isn't the closest solar system over 4 light-years away? How could we travel that far over hundreds of years? I don't know much about this stuff, but a quick search says the fastest a spaceship has ever traveled is 430,000 mph. 4 light-years is approximately 24 trillion miles, and traveling at 430,000 mph, it would take around 56 million years to get there.
I hadn't done the math, but in any case, I was imagining a world in which advancements were made that allowed us to travel orders of magnitude faster than any spaceship has travelled thus far.
Geoffrey Hinton continues to sound alarms about relying solely on market forces to guide AI development.
I think linguists play a pretty big role at least in llm development.
The British-Canadian computer scientist often touted as a “godfather” of artificial intelligence has shortened the odds of AI wiping out humanity over the next three decades, warning the pace of change in the technology is “much faster” than expected.
Prof Geoffrey Hinton, who this year was awarded the Nobel prize in physics for his work in AI, said there was a “10 to 20” per cent chance that AI would lead to human extinction within the next three decades.
Who are you going to listen to?
It actually didn't it's all unsupervised learning.
Linguistic (and semiotics) had a role in previous attempts to create AI through symbolic manipulation, but LLM training and development is gross rough number crunching with no external "insight" of any kind.
Just applied statistics basically
Geoffrey Hinton continues to sound alarms about relying solely on market forces to guide AI development.
Can't regulate it anyway worldwide so what's the proposal? To nuke countries that disagree with AI training regulatory frameworks?
The problem with regulating AI is quite simple: we might halt the progress of it; China won't. Needless to elaborate further.
The problem with regulating AI is quite simple: we might halt the progress of it; China won't. Needless to elaborate further.
That's true for most regulation which is required worldwide in order to "work" (like climate change).
If you can't guarantee full enforcement everywhere, in every relevant country in the world, at any cost, you are in complete bad faith proposing the west to regulate.
And btw any world where a regulation can be enforced worldwide with full force is a world where a world government exists, which is a model so risky and horrific we should always be against any proposal which requires a world government to be implemented
The problem with regulating AI is quite simple: we might halt the progress of it; China won't. Needless to elaborate further.
What if its regulated with adaptive intelligence?
It actually didn't it's all unsupervised learning.
Linguistic (and semiotics) had a role in previous attempts to create AI through symbolic manipulation, but LLM training and development is gross rough number crunching with no external "insight" of any kind.
Just applied statistics basically
Way out of my wheelhouse but my brother in law's brother in law has been a silicon valley guy for like 30 years. Just figured he was some programmer/engineer. Hung out with him during the summer and telling him how I was able to build an app with AI for some csv filtering/renaming I needed with zero coding knowledge. Well turns out he's actually a linguist specializing in 3 obscure African languages and most of his career the last 20 years has been developing AI for this kind of no code coding or whatever. Think Zer's post touches on this above. So I personally don't know exactly how it's implemented but linguists have at least been part of the process up to now for at least using everyday human language to be able to interact and instruct a machine.
Way out of my wheelhouse but my brother in law's brother in law has been a silicon valley guy for like 30 years. Just figured he was some programmer/engineer. Hung out with him during the summer and telling him how I was able to build an app with AI for some csv filtering/renaming I needed with zero coding knowledge. Well turns out he's actually a linguist specializing in 3 obs
Doesn't need to be a linguist either that's the point (the coding to run the LLM training are open source and amazon, google and others give them to you if you want to run your AI training, so that you buy their computing power).
Maybe as a linguist he knows which data to train the mini-LLM on sure, that saves time (and money).
But linguistic insights aren't used in LLM training. There is no linguist going around "teaching" the model anything or structuring the model activity attempting to replicate how humans think about language and so on
Hinton is a linguist?
Can't regulate it anyway worldwide so what's the proposal? To nuke countries that disagree with AI training regulatory frameworks?
We can regulate it's use in applications but it would be a very bad idea to try to regulate the development of AI. Falling behind economically will be an irrelvent problem compared to the military arms race that is well underway and exploding.
Another pervasive bad idea is denial about what is happening. Everything about politics/etc should be focused on the world we are entering and not the world we are fast leaving behind. Extinction is some possibility but the very fast rise of AI/robotics is an inevitability. This is an era defining stuff on a scale and speed never seen before.
