Software "neurons" can now recognise visual patterns and imagine images

[Toxic Tom]

 

The above images were all created by the AI neurons from noise after being shown images of various landscapes, buildings, animals etc.

 

Artificial Neural Networks have spurred remarkable recent progress in image classification and speech recognition. But even though these are very useful tools based on well-known mathematical methods, we actually understand surprisingly little of why certain models work and others don’t. So let’s take a look at some simple techniques for peeking inside these networks.

 

We train an artificial neural network by showing it millions of training examples and gradually adjusting the network parameters until it gives the classifications we want. The network typically consists of 10-30 stacked layers of artificial neurons. Each image is fed into the input layer, which then talks to the next layer, until eventually the “output” layer is reached. The network’s “answer” comes from this final output layer.

 

 

I think this is a really interesting bit of research. In the article they show how each neuron layer is responsible for a different level of image recognition, with the lower level neuron groups distinguishing corners and edges and the higher level neurons recognising certain patterns from previously seen images. I personally believe this to be closely related to the effects of psychedelic drugs such as LSD and psilocin as some of the neurons' outputs are very reminiscent of the psychedelic experience (which I think is an area of research that could shed a lot of light on the inner workings of the human brain).

 

Source:

http://googleresearch.blogspot.pt/2015/06/inceptionism-going-deeper-into-neural.html

[Kyuubixchidori]

so basically, if I take a bunch of drugs, ill think just like a computer? bad ass

[as96]

so basically, if I take a bunch of drugs, ill think just like a computer? bad ass

Just remember to liquid cool yourself.

[elfensky]

Cool.

[LukaP]

This is awesome. to think that merely a few years ago, we werent even dreaming of this kind of stuff, and now its here.

 

I wonder if theyre using that IBM neural network chip that can be connected to even more chips to increase complexity http://www.research.ibm.com/articles/brain-chip.shtml

[jos]

THe concept is nothing new... Google has done remakably good in creating deep neural network.. had been coding for neural nets for atleast 5 years now... The technique itself is really old it started in early 90's...My undergad project was deep neural network based image recognition for vehicles... and my first post gard was 3-d based recognition of finger prints from pictures irrespective of orientation 

[Toxic Tom]

so basically, if I take a bunch of drugs, ill think just like a computer? bad ass

 

I think the point is that it's not thinking like a computer, it's thinking "like a human".

[cesrai]

I saw IBM watson doing something like this and Nvidia had a program to run on there GPU to tell them what a CAT is.

[Kyuubixchidori]

I think the point is that it's not thinking like a computer, it's thinking "like a human".

so basically, if I take a bunch of drugs, a computer will think just like me? more bad ass

[Toxic Tom]

so basically, if I take a bunch of drugs, a computer will think just like me? more bad ass

Hmm I suppose so yeah

[Kyuubixchidori]

Hmm I suppose so yeah

keep in mind my post where 100% jokes.

 

this is pretty cool though. not sure how I feel about computers "learning". having computers control things is 100% okay in my mind, as long as they have predicatble end results. when computers start making their own choices, that's when it gets scary. 

[GoldSrc]

having computers control things is 100% okay in my mind, as long as they have predicatble end results. when computers start making their own choices, that's when it gets scary. 

 

Having computers control everything may be okay, but some advanced AI?, nope.