They Are Among Us
Was some of this article generated by an AI too? There are some grievous grammatical errors..
AI systems can now create images of humans that are so lifelike they look like photographs, except the people in them don’t really exist. See for yourself. Each picture below is an output produced by a generative adversarial network (GAN), a system made up of two different networks including a generator and a discriminator. …
"selling hardware is what they are interested in"
In terms of the ML space, no. Renting, sure, but owning, ROFL. They aren't even clear where they draw the line, so there is no academic exemption, or hardware specific. If they don't like the code you're running on their card, then no CUDA for you.
We're not allowed to teach our ML students how to do it on a nVidia GPU, since that might suddenly make my 300 euro gaming GPU suddenly require a 10k pa dev licence that the uni would be on the hook for. AMD is much more helpful, since there stuff seems like it's released by engineers, not lawyers.
In terms of cost, assuming a card lasts two years, roughly 80% of the total cost it is licencing fees versus hardware. 5k card, 20k licence.
For the machines which do have that fat nVidia tax paid for, it's generally not the most expensive piece of software on the box. Oddly enough, the companies that make that software give me "free" copies, because I'll be using it for academic purposes.
You've never looked at 500px photo galleries then? All the top images on 500px.com pretty much all look like they were photographed and processed by machines using some form of AI software, generic and bland. People have lost the art of injecting "soul" into art these days, sadly it's not even the pursuit of money it's worse. Nowadays it's simply the pursuit of social networking rankings, very sad.
They all look happy, even the ones that aren't grinning inanely.
If I live long enough to retire one project I've mused about for years will be photography - documentary portraits of commuter faces. Regret, despair, angst, miseries of all sorts -- there's a vast palette of them on the 17:44 back to suburbia.
Do it! I've been a hobbyist snapper for many, many years I love capturing people's expressions but I tend to lean to more candid street images. My daily commute to my day-job down one of tube lines has lead me to consider a similar project. Candid snaps of people's thousand-yard stares while they cling to the handrails on the tube trains that are taking them to yet another dreary day in the office. Sometimes I just spend the the journey studying people's expressions, they really do run the entire gamut of human emotion some days.
The video really looks like the early morphing demos (such as the 'Black or White' videos) - but that was time-consuming hand tweaking of real faces, this is all automatic.
Good news for troll farms and political donors wanting to hide their identities behind a myriad of fakes.
So they made an automated photofit + colour swap system?
Yes, I know it's a little more complex than that, but in reality, we also don't know "how" a specific neural network is doing something (we cannot reverse engineer just yet).
So in that case, it could literally just be doing a photofit and colour swap. XD
Like this software already exists (Google have a ton of APIs/example codes), and you also have Unsplash for (at least for development) free content to add to the data set!
If I was clever enough, I would have done it already. You would have automated location + character generation at least on the cheap and seemingly infinite replayable (think the current trend of dungeon crawler/rouge lite).
....I'm pretty sure I used to go out with her. I should have known - she was a bit limited vocabulary-wise but on the plus side she was compliant, although it did worry me when she said "Dave I'm afraid I can't allow you to do that". I got more suspicious when she told me to call her Alexa during sex.