The Singularity

[osmosisch]

Whoa, those are some eloquent words I said there. You sure that was me?

Blergh. Fixed.

[clyde]

Listening to Bruce Sterling talk about the utility fog I started to wonder: Is it physically possible for all the mass in the universe to simulatenously be energy instead?

[clyde]

This hour-long documentary is kinda boring, but it really nails it in that we are about to see massive job losses very soon.

[clyde]

I don't quite understand it, but everything linked in this blog post is kind of... it makes you think. I guess they are teaching AI to manipulate images based on what it manages to kind of see? It's weird.

http://317070.github.io/LSD/

 

It's pretty wild that google's image recognition software dreams of all the pet-photos people have been putting up for the last decade. 

[clyde]

I continue to think about this visualization of image-recognition machine-learning. This seems like such a big deal to me (because I can now kind of see how it works). I look at this Twitch stream and have all sorts of thoughts. This isn't just a trippy-demo. The demonstration of an algorithm that a program has developed for itself in order to identify images (with the help of humans tagging the relevant examples) is the first time I really have begun to comprehend how self-driving trash-trucks are going to pick up my trashcan, having never seen it before.

That this is a visual demonstration is also helpful for me to be able to grasp how more conceptual implementation of machine learning works/will work. I could totally see a doctor-robot recognizing the totality of my symptoms and physical diagnostics like weight and blood-pressure as a patient with a certain variety of illness using similar methods as this bot that adds pixels to match psuedo recognitions. 

But this visualization also helps me understand how my mind creates models of the world around me. Our ability of prediction has to compete with reality and we have more complex motives than a stream of suggestions from Twitch-users, but I see a lot of similarities between how I extrapolate from personal anecdotes and how this bot draws pagodas.

[clyde]

The visual hallucination thing applied to video:

https://twitter.com/kcimc/status/616662893120090113

...ah, apparently Google released the code so that people can start experimenting with it.

http://www.theguardian.com/technology/2015/jul/02/google-electric-sheep-image-processing-neural-network-algorithm

[clyde]

I've heard of using face-recognition technology for identifying emotions, but this is the first demonstration of quantifying the information that I have seen. 

 

[clyde]

Have any of you seen Sleep Dealers? The premise of that movie seems more and more likely to me as time goes on.

[Erkki]

Good article on singularity. https://aeon.co/essays/true-ai-is-both-logically-possible-and-utterly-implausible

[Cordeos]

Good article on singularity. https://aeon.co/essays/true-ai-is-both-logically-possible-and-utterly-implausible

"We have no idea how we might begin to engineer it, not least because we have very little understanding of how our own brains and intelligence work."

So much this.

[Erkki]

But actually, I think a thing that we could be worried about (besides AI taking the jobs thing, which I think will be good in the end) is that some kind of trainable dumb-AI based on neural networks or something is put into a situation where it is able to take a lot of decisions for a lot of people in some contexts, in a way that ends up increasing inequality and giving more and more power to those *humans* who control the AI, and also starts to move intelligence to where the money is.