Hacker News Reader

Comment by erikpukinskis | Hacker News Reader

The thing people don't understand is that in order to simulate human intelligence, you have to be able to simulate TWO things:

1) A human brain

2) An entire human development up to the age of intelligence you are looking for

The first one is not the harder of the two.

Now, many AI researchers believe they can cut corners on the whole simulating an entire human lifetime thing, and that they can use a more impoverished simulation and make up for it on volume... say, just flashing a billion images in front of the AI hoping that's enough to form the specific subset of intelligences you are hoping for. Or letting the AI read the entire internet. But at this point it's an open question whether that could even theoretically lead to generalized intelligence.

View on HN

In order to simulate human-level intelligence, the machine doesn't necessarily need to be modeled on the human brain. It doesn't even need to use neural nets.

In order to simulate a human (which by definition only has human intelligence) then only your point 1 is true. Point 2 is not, because the standard way to know we've got 1 right is uploading/emulation: take existing humans, scan their brains at a sufficient level of detail (which will probably require destructive scanning), and use that as the base information of the simulation.

The neural nets still need to have experiences that are trained on a realistic simulation of the events which it needs to understand. Historical data does not train the networks the same way that interactive learning does.

There was an old study, in an earlier time of ethical strictures, where they took two kittens, paralyzed one, and strapped it onto the other one and let them run around and do kitten stuff. The paralyzed kitten saw everything the intact kitten did, felt the same breeze on its fur, but was completely blind to the universe. Without interaction, you cannot learn.

In order to provide that interactive environment you either need a robot body or a really rich virtual environment for your AI to grow up in.

At that point, they are developmentally limited to human timescales. No hyper-accelerated intelligence exponential.

> An entire human development up to the age of intelligence you are looking for

I've also found the lack of development time consideration a weakness in the AGI literature I've read. We can create intelligent humans now, but it takes not only reproduction, but immersion in our world and societies for something like 12-18 years for a fully intelligent agent to emerge. Maybe if some future DeepMind algorithms could inhabit a complex world like you find in Skyrim for a decade or two, you'd get something interesting.

You must be reading different AGI literature than me. More generally Embodied/Unembodied AGI has been a common discussion theme for a long time, just as much as the theme of what the right mix is for supervised/unsupervised learning / a priori knowledge. For instance here's a random Goertzel paper from 12 years ago: http://www.goertzel.org/papers/PostEmbodiedAI_June7.htm Perhaps the missing focus on timetables for embodied agents to reach human-level or beyond is annoying (such a timetable is not present in that paper) but to have that timetable problem to begin with would be great since we'd have already solved the much harder problem of creating something that can learn and generalize extremely well within an environment, even if it takes it 12-18 years to reach a sufficient level to be interesting. No matter how many years it takes, so long as we're certain it will get there in some number of years, the problem is greatly simplified to one of hardware and software optimization. So not that interesting by itself.

The brain is a mechanism that turns intelligence into action, language or memory. But who has ever said that intelligence is generated in the brain?

Pretty much everyone?

Reply on news.ycombinator.com