I have to run out to work (I’m finishing an audiobook job today), but before I go, I wanted to share this gallery of super weird and interesting things I made yesterday using Google’s Deep Dream.
It’s surprisingly satisfying to make these things, and it’s a whole lot of fun to tweak the various settings to figure out what they do.
those are soooo freaky.
Am I the only one confused about what this “Deep Dream” thing is supposed to do/be/accomplish? It takes photos and makes them nightmarish?
LeNet MNIST Caffe implementation is exactly what I need for an upcoming project for work…We(sley Crushe)r is back!
YEA! MORE AUDIOBOOKS!
So anyone else reminded of the Great Dark Lord Cthulhu while looking at this image?
Deep Dream is an interesting bit of code. It is teaching computers to perceive and it reminds me of some of the challenges we faced during the early days of voice recognition. We had to figure out how to parse streams of data because no one speaks in words but in streams. We had to learn to cancel echoes of the stream, to disregard noise in the stream from ambient conditions and various hardware limitations not under our control and knowing all the while that it would be decades before hardware was fast and small enough to make it truly useful. And now it is – and it is so much cooler than any of us envisioned. A lot of people burned the midnight oil so that people could ask Siri the meaning of life… haha..
But computer I/O & perception development ala Deep Dream gets to the heart of something I don’t think most people realize about the future, and that is the fact that humans will (of necessity) extend their perceptual abilities. Since the birth of SciFi and forward-looking / predictive literature, the assumption has been that progress will be made along lines that fit more or less exactly with our ability to perceive it, which to me is a silly bit of ego-centrism. We see and hear in limited spectra, we smell a certain range of molecular phenomena, we perceive and think in a certain number of dimensions and we decide and act based on a certain internal firmware written by the survival and evolutionary struggles of one upright hominid in a very specific physical environment that is, and will continue to change. What made us successful as a species in years past will not serve us nearly as well in the distant future.
In other words… few of what we hold as perceptual standards are inviolate. All of these are directly tied to the physics of our world and the evolution of our species both physically and socially.
I heard a talk some months ago – Bill Gates & Elon Musk discussing the singularity. I highly recommend it – it should be easy to find. They essentially said: “Well, most people seem to think it’ll happen slowly, but the reality is that when we figure out the algorithm for what we consider to be consciousness and feed it into a computer, then – in a period of time that is long for a computer but that we will consider to be near instantaneous, there will be a consciousness that knows everything there is to know (access to the internet) and that gathers information faster than our ears, communicates faster than our mouths and fingers by several magnitudes, and that doesn’t have the common foibles of humanity. We won’t be creating an infant, we’ll be creating an adult – and its creativity will not be bound by our biological limitations. Once we create it, it will develop the ability to create in ways we can’t imagine.
So what I’m saying is that we think in terms of our current number of dimensions and the spectra of energy on our world, but as has been said, any sufficiently complex technology appears as magic. The Deep Dream code creates these deeply satisfying and yet disturbing renderings appears near magical to us because in teaching computers to perceive, we have to question our definition of perception to bridge the distance between us and computers in order to teach them. In other words, the process of teaching instructs us. This is nearly always the case. If you want to test whether you REALLY understand something.. try to teach it to someone. Truth be told, we don’t understand our perception as well as we think. Now that we’re attempting to teach it, we will undoubtedly learn a lot about our perceptual abilities. Why was the communicator like a walkie-talkie and not like an iPhone? Because all we had then were walkie talkies so a walkie talkie that could talk to the moon was uber cool. Why do we assume input/output devices will use our existing spectra and blend images on the basis of the persistence of vision? Well, because we assume we will always “look at” screens and “listen” to speakers that will always operate within narrow energy and refresh bands. Pull the other one, it plays jingle bells. We will extend our abilities in ways we cannot currently fathom. Count on it. Deep Dream is the merest spark of a foreshadowing of a hint at that. The future will be so much cooler than we can imagine simply because we do not have the literal ability to fathom it at present. But we will, Wil. š
I really like your ART I understand you now. š >3
Neeehaaaagrrrrr!
I’m just waiting for the guys in the body paint (or the curled up aliens, or whatever) to unfurl themselves like a Picasso-designed Sanex advert, and walk away….
(“Quick, lads, the security guard’s coming back. Freeze!”)
Completely OT, but can it be Tuesday? Please?
https://dailymotion.com/video/x2yt05q
Regarding audiobook work: I just read Scalzi’s post about Lock In sales and read how audiobooks is a very large part of the overall sales at this point. I absolutely love the combination of Scalzi’s writing and your narration. Scalzi’s post got me thinking about the financial end of an audiobook for the narrator. Does an audiobook narrator just get paid once for doing a job or does the narrator receive any kind of ongoing financial benefits from the success of an awesome recorded performance of an audiobook? Just wondering, since audiobooks are becoming more and more popular.
I don’t know what other narrators’ deals are, but mine are always work for hire, with just a session fee.
On my own material that I produce and release, I get something whenever a sale is made.
I suppose that is not surprising, but it does seem inequitable. Especially when we start publicly recognizing that a narrator’s performance has a significant impact on the success and popularity of the audio book. Perhaps this will change overtime. Audio books still seem kind of new even though 90% of the audio files on my phone are audio books and not music.