|
|
|
|
|
|
by senttoschool
1099 days ago
|
|
|
I was thinking that maybe AGI won't be here for a long time. The reason is because humans take in so much data, process it, and use it without our conscience even knowing. For example, when driving on the road, we know inherently that if you see a heavily modified BMW with tinted mirrors, you need to be prepared to break because the chance of it cutting you off is higher than a Prius. To know why modified BMW drivers often do this, you have to understand that some people like to show off. To understand why someone would want to show off, you have to understand something else. And so on... There are probably even more subtle information that we take in without us knowing and we make decisions based on that data. There are just so many things that the human brain learns over a lifetime that any AI will have a hard time duplicating. |
|
|
A single intelligence is more powerful than any human intelligence in the planet, and more over, the closest new stuff coming will make them easily more intelligent: they can be assembled - several models, like 5 o 50.000 - to work together, to share information at a pace unavailable to humans right now. They can learn, the 50.000 LLMs simultaneously, 50.000 different things for they own, each one of them, and then just copy the resulting learning to the rest of the 49.999 LLMs. They all can do this and learn at a pace impossible for individual humans beings nor 50.000 "fully connected" humans (you take those 50.000 persons to a building, provide them communications means, radios, whatsapp, whatever: they won't be able to surpass the LLMs capability for instantaneous parallel learning).