A.I. Has Won. Hollywood Will Never Be the Same for Actors.
694The first 1,000 people to use the link will get a 1 month free trial of Skillshare: https://skl.sh/famefocus06221
Since way back, artificially intelligent machines were generally portrayed in movies as scary and evil... and bit slow and clunky, although that was probably unintended.
But either way, in general the habit has been to portray A.I. as the enemy and not to be trusted.
However, the irony of this is that in order to produce most modern-day movies VFX artists are actually using AI and Machine Learning tools to help them out.
And with robots getting a lot more advanced, it raises the question: Will A.I. replace actors?
Well, maybe not with robots, but in many modern movies, there are actually occasions where we've been watching A.I generated material and not even realized!
Like the music in this video? I made it!
Support me by getting it on any of these sites :P
Get it on iTunes: ► https://apple.co/2ENGfu9 ◄
Listen on Spotify: ► https://spoti.fi/3boTfCl ◄
Buy it on Amazon: ► https://amzn.to/2QVJZfk ◄
You see, little by little AI has been creeping into the visual effects workflow and it now forms a part of a variety of different tools:
Adobe After Effects Rotobrush 2 uses Adobe Sensei Machine Learning to predict the movements of the foreground to allow you to easily cut it out from the background, Nuke 13 also has a machine learning toolset and Maya has Neural Network capabilities.
The principal idea behind these types of tool is fairly simple.
For example, Imagine that you had never seen a flower before, you wouldn't be able to tell the difference between a real one from a fake one.
Now, if you had seen flowers before but didn't know their names,
you'd be able to tell whether a flower was real or not but you wouldn't know if it was a rose or a daisy. However, if you had seen a lot of different types of roses, from all different climates, not only would you be able to tell the difference between a Chinensis rose and a Gallacinae rose but also the country it originated from, and even perhaps how healthy the plant itself was.
Machine learning is basically this, compiling a library full of data that the tool can use as "Learning Data" so it can then study this data to identify patterns, similarities, or differences.
Deep Learning Tools
Framestore did a research project with software developers Weightshift to combine ragdoll simulation, keyframe animations and machine learning to take care of the on more menial and repetitive tasks, like for example, this man riding a horse. His body appears to be animated but it actually isn't. His body movement is being generated by a ragdoll simulation that is reacting to the horse's animation. This then allows animators to focus on the more artistic elements of the shot, like for example, having him turn around to see who's behind him.
This technology was used in things like the movement of this monkey's tail in tv series 'His Dark Materials'.
Weta Digital uses Deep learning in its facial pipeline and also in its FACETS system that can capture subtle details of the actor's performance and figure out which muscles have moved and how, then it maps those movements onto the digital characters.
Digital Domain also used their own machine learning software Charlatan with their proprietary tool Masquerade, in their real-time cloth simulation tool Elbor, and for facial capture with their real-time LiveDrive system.
As the use of A.I in tools and the tools themselves expand and evolve it seems that the only thing that limits their capabilities and effectivity is the amount of learning data they have access too.
When Digital Domain had to do a digital pick up shoot of Channing Tatum for Free Guy they were very limited with the amount of quality data they had access to for the machine learning, even so, their results were impressive, but with access to more learning data they could be even better.
Considering every VFX house has its own A.I software and its own Learning database we asked Digital Domain's VFX supervisor Niko Kalaitzidis whether a shared centralized machine learning database could be a future possibility.
He told us that after conversations with the head of the Charlatan team at Digital Domain, Matt Smith, that VFX houses could collaborate and share large datasets without exchanging them via Federated Learning.
This would allow each VFX house's tools to learn from another company's learning data without actually having access to or copying it.
Kind of like the difference between watching a film through your neighbors window instead of them burning you a copy.
What Does The Future Hold.
Well the possibilities of what could be accomplished with the help of Deep Learning and A.I are immense.
Read more in www.famefocus.com
This video is sponsored by Skillshare
Read more here: www.famefocus.com
Follow us on Twitter: https://twitter.com/focusfame