AGI will come from one or two people working by themselves, outside of academia, no more than 100k line of codes.
Where will artificial general intelligence come from?
191–200 of 252 posts
Re: Where will artificial general intelligence come from?
#192IMHO: If an AGI from the future came back to 2017, it could almost certainly create a new AGI from scratch on current hardware. What would it type into its terminal?
We, humans, are general intelligences, and we are incapable of creating AGIs on modern hardware. What makes you think artificial variants would be any more capable?
Re: Where will artificial general intelligence come from?
#193Earlier quoted context omitted.
Fortunately you don't have to write an AGI yourself. We have a very powerful tool called evolution that can do all the heavy lifting for us, we just have to set up the environment and the goals. I'm pretty sure we could create an AGI, given enough computational power and time, we're basically hardware-limited.
Nature had a planet and several billion years. Of course the goal wasn't AGI, it was survival, and horseshoe crabs have done a pretty good job of that. So have beatles, with their large number of species. How would you select for only AGI? It would be like selecting for only the greatest eyes (Mantis shrimp), but doing it over the entire tree of life. You still need a way to narrow down to the best eyes.
Survival is not a goal, it's just the score keeping system.
Re: Where will artificial general intelligence come from?
#194I gave this talk a while ago to a small group of attendees. It was not recorded (I saw some ask below). It's based on a document I wrote a while ago called "You suck at writing AI" (never published). The basic argument was that people are comically inadequate at writing complex code. You can't write the code to detect a cat in an image and the correct thing to do is to give up, write down an objective that measures t…
Fortunately you don't have to write an AGI yourself. We have a very powerful tool called evolution that can do all the heavy lifting for us, we just have to set up the environment and the goals. I'm pretty sure we could create an AGI, given enough computational power and time, we're basically hardware-limited.
I think that's what the person you are replying to is arguing for...
Re: Where will artificial general intelligence come from?
#195Can someone explain where the gif image on slide 69 comes from?
Re: Where will artificial general intelligence come from?
#196Earlier quoted context omitted.
It's called the self driving car - an AI that interacts with the world. It will be a launching pad to other AI agents.
At a more general level, you may find the book Society of the Mind interesting ( https://en.wikipedia.org/wiki/Society_of_Mind ). In this, Minsky proposes simple (mindless) agents that then are combined together to form a mind. That our mind is interacting with the world and launching agents to deal with things - that our own mind may be built this way. Yes, its a bit old (1986) and the current machine learning techn…
Minsky's 'society of mind' is actually not far from modern techniques, where deep neural nets have multiple distinct parts that solve different parts of the problem using each others' outputs. (Not to mention explicitly multi-part methods like actor-critic).
Re: Where will artificial general intelligence come from?
#197can a data center be shrunk down to the size of a consumer product within the next 50 years?
will we all own one and store massive amounts of information for purely selfish or inane reasons?
yes/yes - ai comes out of that.
no/no - we hit computing plateaus and ai becomes dm (decision maker), and we all own a pdm.
Re: Where will artificial general intelligence come from?
#198IMHO: If an AGI from the future came back to 2017, it could almost certainly create a new AGI from scratch on current hardware. What would it type into its terminal?
We, humans, are general intelligences, and we are incapable of creating AGIs on modern hardware. What makes you think artificial variants would be any more capable?
'Incapable' the way humans from the 1800s were 'incapable' of building heavier-than-air flying machines? Every human invention in history was 'impossible' until we pulled it off.
Re: Where will artificial general intelligence come from?
#199This slideshow is a mess.. it needs some kind of narrative.
Re: Where will artificial general intelligence come from?
#200I gave this talk a while ago to a small group of attendees. It was not recorded (I saw some ask below). It's based on a document I wrote a while ago called "You suck at writing AI" (never published). The basic argument was that people are comically inadequate at writing complex code. You can't write the code to detect a cat in an image and the correct thing to do is to give up, write down an objective that measures t…
There are at least two huge advantages of the decomposition by function approach compared to the black box solution space search approach:
- The former can be debugged. If your image analysis system embarrassingly tags black people as gorillas, you can go in and find the bug and fix it. That's not so easy if you're using a black box model.
- The black box approach is tremendously enervating. You code up a neural network architecture, launch a bunch of cloud GPU instances, and start training. If the results are bad... you try a new architecture. In the function decomposition world, you can actually use your knowledge and understanding of the system as an engineer to figure out what went wrong and why.