Compression is the trick. Its even philosophed about if compression = intelligence.
The LLM has to compress everyy question/prompt into its system. It does so by creating rules and ways of processing data (this can lead to AGI, world models or an architecture of sub architectures like an LLM + something else). So if it should respond in a way that only reasoning people can achieve, it might be able to learn a representation of what we call reasoning.
It read enough text in itself to even know about the concept of reasoning and how you would do that.
Even if this is only stochastic, it shouldn't be so devalued as your comment comes across.
A brazilian favela is not a representation for a global satelite network which consumes and produces a lot of energy and co2 to be send up and poisens our atmosphere when theybreak down.
There was no evaluation of the obvouse shitty problem in such areas.
And as i said, metal tubes do break radio signals.
That will not work for a long time for peak hours and remote areas with too little of customers.
Everyone has to travel into our city in the morning to work, everyone does it with their own car they are proud of and see as essential at the same time while standing in traffic for 20-30 minutes extra.
And then what do you get? You get YOUR car back, smelly, stinky, dirty for saving a little bit of money?
We already have a 'drive yourself' car rentel system here and the lowest price for an avg car is 8 Euros per Hour for the time and 40 cents for 1km.
I also don't think insurance will be that much cheaper for autonomous driving. The insurance for a normal car is not expensive already, we have very few car incidents.
It might even become more expensive because you are driving more km per year.
And i brought an argument. You said it makes space-x very valuable and i explain that the amount of payload we even send up is very limited which contracditcs 'very valuable'.
I then explain further what Elon Musks plan is to sell us his trillion dollar company or how you frame it 'very valuable' and explain why it doesn't work
Starlink has 10 MILLION customers. Thats just nothing.
All the investment in Fiber and mobile towers are long lasting investments.
Starlink NEEDS v3 to scale because they already have scaling issues. They need Starship, which doesn't work yet, to work to even send v3 up there.
And while Spacex has some first mover advantage, other companies start doing the same which will eat their margins. Makes it even more complicated to run all of it.
They have to do 300k orbit correction already last year, kessler syndrom can happen which will block access to space for all of us.
We don't even know yet how dangerous the poisoning of our atmosphere will be.
You still need a powerline to your house, sewer and water.
There are plenty of fibers and dark fibers on power pools.
Starlink doesn't 'just' pollute the night sky for EVERY SINGLE HUMAN (8 Billion people) it can also poisen our atmosphere when they re-enter and burn up.
Yeah but it doesn't become a Trillion dollar business if they don't solve the Starlink Satelite v3 issue.
They need Starship to be able to send v3 up, without v3 it doesn't scale well enough.
Starship still hasn't proven it can actually bring up the relevant payload high enough and they need it to be reusable otherwise costs will increase.
And they already exist and only have 10 Million customers. They need to get countries on their side like India but these countries are not stupid. Elon Musk showed them very clearly what he can do like his statements he did when Ukraine war started.
Right now we do not even have the antenna technology in current high end smartphones for 'easy to use, normal speed' mobile to satelite communication.
And funny enough, the more local mobile phones you would have, which want to send data to a satelite, the harder the problem gets due to interference.
With 5g we do already a lot of beam forming etc. Try beamforming into 500km space with uncoordinated random amount of mobile devices with very very little sending power and one satelite 'beamforming' its a few hundred square miles.
It will take years if not more to be technical capable to have modems so good that they can communicate with a starlink satelite in any reasonable 'day to day' way.
And Starlink already increased prices again.
And without Sparship and prooving that they actually can reuse it, they can't hold the price point.
Starlink satelites do not scale very well. They need v3 and even with v3 this doesn't scale efficently.
Starlink only has 10 Million customers, too expensive for most countries already.
Starlink brought internet to a lot of people who had it before already but made it easier for them.
Its still quite a interesting technology, given, but for the fact that he destroys potentially our atmosphere, has control over war critical tech, can do survailance and wants to send out A Lot MORE into our space, its a net negative for at least 7-8 BILLION people while 10 Millionen people benefit from it.
And they even increased the price just a few weeks back...
For less starving people? For less child abuse? For less climate change?
I look up at the night sky and i want to see stars and the endlessness of the universe and don't want to be reminded that Elon Musk will poisen our atmosphere.
The LLM has to compress everyy question/prompt into its system. It does so by creating rules and ways of processing data (this can lead to AGI, world models or an architecture of sub architectures like an LLM + something else). So if it should respond in a way that only reasoning people can achieve, it might be able to learn a representation of what we call reasoning.
It read enough text in itself to even know about the concept of reasoning and how you would do that.
Even if this is only stochastic, it shouldn't be so devalued as your comment comes across.
Who says that we are doing anything more magic?