Incredible Athena
Creation of computer graphics for the film industry is a very complex process, requiring a lot of hardware resources, time and human strength. Intellectual, of course, as Comrade Lenin wrote. Today we will tell you about the company Bit Theory Inc. and its creator Allen Bolden, who together with the team does not just make beautiful pictures, but actually develops real artificial intelligence!
Creative artificial intelligence
The main trump card of BTI is the Athena software core, with which Bolden ... communicates. For example, he wants to see a scene with a green plane that flies above the clouds, and writes in a special field “a green jet flies above the clouds”. In response, a ready-made visualization of the scene immediately appears, which is fully suitable for further development by computer graphics artists. But the coolest thing about all this is that Athena learns in the process of solving each task, and then uses the accumulated "experience" in the following projects.
')
In many ways, Athena is similar to Watson, an artificial intelligence system developed by IBM. In 2011, she made a real sensation in the television quiz Jeopardy!, The analogue of which in Russia is “Own game”. Questions quiz Watson compared with a 15-terabyte knowledge base from a variety of areas. As a result, the supercomputer won the two champions of the show: they received 200 and 300 thousand dollars each, and Watson hit the jackpot - a whole million.
The creator of Athena believes that the algorithm of Watson's actions can be compared to how the rational left hemisphere of the brain works. The actions of Athena are more to the other, the creative hemisphere. The program understands what is written to it, but responds not with words, but with visualizations.
By the way, initially Bolden was not going to create artificial intelligence at all. His career began at Marvel Comics, where he came as an intern and suddenly realized that he really liked programming. To study this case, he temporarily quit his job and entered the University of Berkeley. After graduation, Allen received serious knowledge of artificial intelligence and network infrastructure and returned to Marvel. There he was already awaited by colleagues, who by this time were completely immersed in the movie business: "Hey, man, you cut in the computers, right?"
Bolden asked for help in solving the problem that arose during the implementation of one of the important projects. The deadline was tough - it was necessary to deal with everything in 5 days. It was then that the groundwork in the field of AI was useful. “I just thought that my knowledge would do some calculations faster, but ultimately something much more happened,” says Allen.
So, Athena transforms natural language into a quality blank for a future visual product. Bolden sees the main goal of his development in making the work of computer graphics artists more creative, and shifting all the routine parts like code and settings for creating animation onto the virtual shoulders of Athena.
By the way, when Bolden writes something to the program that she cannot understand, Athena puts out a series of question marks in this place. Then Allen has two options - make a more detailed description or give a link to the desired visual object that has already been created manually. After Athena completes its work, artists take its results as a basis for work in Maya and 3ds MAX. As long as they bring the graphics to perfection, Athena seems to be “watching” what specialists are doing, keeping all actions in their ever-growing database in order to use “knowledge” in the future.
The action algorithm of the graphs is as follows: first, the artists cut out fragments with actors, effects and other moving objects from each frame, then fill the frame with visualization of the environment and then completely recreate the entire scene in 3D.
Required performance
When Bolden created Athena, the system used the power of a cluster of 70 computers to work, and at the same time any other auxiliary computers that Allen managed to extract. Today, Athena works with a cluster of 30 Lenovo ThinkStation D20: they are mainly equipped with two quad-core Intel processors.
Two years ago, Bolden managed a team of 45 computer graphics artists. Today, the company employs 150 schedules in the US, plus external employees from Taiwan, Hong Kong and Seoul. In addition to Lenovo workstations where the entire Athena knowledge base is stored, the BTI team uses the ThinkStation D20, C20 and S20.
According to Bolden, BTI visualizes about 10,000 frames per week. Considering the fact that in one second there are 24 frames, then about 7 minutes of the finished video are obtained at the output. Of course, the team needs a lot of memory to work with such “heavy” material. Currently an array of 17 TB hard drives is used in a RAID1 configuration. Of course, this is not about the entire BTI database, but about the operational repository on which all work is directly performed.
Bolden compared the speed of different computers that he used. For example, a fragment from the movie “Transformers: The Dark Side of the Moon” with a length of 196 frames, a system with a six-core processor and 8 GB of RAM processed 96 minutes. A Lenovo ThinkStation D20 with two quad-core Xeon and 12 GB of RAM did the same task in just 25 minutes.
In another test, Bolden had to visualize 21 objects covered with hair drawn by computer graphics artists in Autodesk Maya. Individual hairs were more than 140,000! Lenovo Think Station S20 coped with each frame in about 5 minutes, and the other computers at the disposal of Bolden, simply did not pull this task.
And BTI’s leading specialist in 2D graphics conversion in 3D says that before his main program (The Foundry’s NUKE) had constantly “fallen” when solving complex problems, almost a couple of times a day. After Lenovo ThinkStation D20 appeared in the company, nothing like this happens at all.
BTI plans
Allen Bolden, of course, continues to improve Athena day after day. His goal is to continuously improve the efficiency of workflows, he wants to make the production of computer graphics faster and cheaper without losing quality. But the system is still at an early stage of development, so the 18-month R & D project was launched at BTI. The creator of Athena believes that productivity will greatly increase if you run the software core on 40 Lenovo workstations at the same time. In this lies the tangible commercial potential of using Athena and in external projects, and not just within BTI.
Creative artificial intelligence
The main trump card of BTI is the Athena software core, with which Bolden ... communicates. For example, he wants to see a scene with a green plane that flies above the clouds, and writes in a special field “a green jet flies above the clouds”. In response, a ready-made visualization of the scene immediately appears, which is fully suitable for further development by computer graphics artists. But the coolest thing about all this is that Athena learns in the process of solving each task, and then uses the accumulated "experience" in the following projects.
')
In many ways, Athena is similar to Watson, an artificial intelligence system developed by IBM. In 2011, she made a real sensation in the television quiz Jeopardy!, The analogue of which in Russia is “Own game”. Questions quiz Watson compared with a 15-terabyte knowledge base from a variety of areas. As a result, the supercomputer won the two champions of the show: they received 200 and 300 thousand dollars each, and Watson hit the jackpot - a whole million.
The creator of Athena believes that the algorithm of Watson's actions can be compared to how the rational left hemisphere of the brain works. The actions of Athena are more to the other, the creative hemisphere. The program understands what is written to it, but responds not with words, but with visualizations.
By the way, initially Bolden was not going to create artificial intelligence at all. His career began at Marvel Comics, where he came as an intern and suddenly realized that he really liked programming. To study this case, he temporarily quit his job and entered the University of Berkeley. After graduation, Allen received serious knowledge of artificial intelligence and network infrastructure and returned to Marvel. There he was already awaited by colleagues, who by this time were completely immersed in the movie business: "Hey, man, you cut in the computers, right?"
Bolden asked for help in solving the problem that arose during the implementation of one of the important projects. The deadline was tough - it was necessary to deal with everything in 5 days. It was then that the groundwork in the field of AI was useful. “I just thought that my knowledge would do some calculations faster, but ultimately something much more happened,” says Allen.
So, Athena transforms natural language into a quality blank for a future visual product. Bolden sees the main goal of his development in making the work of computer graphics artists more creative, and shifting all the routine parts like code and settings for creating animation onto the virtual shoulders of Athena.
By the way, when Bolden writes something to the program that she cannot understand, Athena puts out a series of question marks in this place. Then Allen has two options - make a more detailed description or give a link to the desired visual object that has already been created manually. After Athena completes its work, artists take its results as a basis for work in Maya and 3ds MAX. As long as they bring the graphics to perfection, Athena seems to be “watching” what specialists are doing, keeping all actions in their ever-growing database in order to use “knowledge” in the future.
The action algorithm of the graphs is as follows: first, the artists cut out fragments with actors, effects and other moving objects from each frame, then fill the frame with visualization of the environment and then completely recreate the entire scene in 3D.
Required performance
When Bolden created Athena, the system used the power of a cluster of 70 computers to work, and at the same time any other auxiliary computers that Allen managed to extract. Today, Athena works with a cluster of 30 Lenovo ThinkStation D20: they are mainly equipped with two quad-core Intel processors.
Two years ago, Bolden managed a team of 45 computer graphics artists. Today, the company employs 150 schedules in the US, plus external employees from Taiwan, Hong Kong and Seoul. In addition to Lenovo workstations where the entire Athena knowledge base is stored, the BTI team uses the ThinkStation D20, C20 and S20.
According to Bolden, BTI visualizes about 10,000 frames per week. Considering the fact that in one second there are 24 frames, then about 7 minutes of the finished video are obtained at the output. Of course, the team needs a lot of memory to work with such “heavy” material. Currently an array of 17 TB hard drives is used in a RAID1 configuration. Of course, this is not about the entire BTI database, but about the operational repository on which all work is directly performed.
Bolden compared the speed of different computers that he used. For example, a fragment from the movie “Transformers: The Dark Side of the Moon” with a length of 196 frames, a system with a six-core processor and 8 GB of RAM processed 96 minutes. A Lenovo ThinkStation D20 with two quad-core Xeon and 12 GB of RAM did the same task in just 25 minutes.
In another test, Bolden had to visualize 21 objects covered with hair drawn by computer graphics artists in Autodesk Maya. Individual hairs were more than 140,000! Lenovo Think Station S20 coped with each frame in about 5 minutes, and the other computers at the disposal of Bolden, simply did not pull this task.
And BTI’s leading specialist in 2D graphics conversion in 3D says that before his main program (The Foundry’s NUKE) had constantly “fallen” when solving complex problems, almost a couple of times a day. After Lenovo ThinkStation D20 appeared in the company, nothing like this happens at all.
BTI plans
Allen Bolden, of course, continues to improve Athena day after day. His goal is to continuously improve the efficiency of workflows, he wants to make the production of computer graphics faster and cheaper without losing quality. But the system is still at an early stage of development, so the 18-month R & D project was launched at BTI. The creator of Athena believes that productivity will greatly increase if you run the software core on 40 Lenovo workstations at the same time. In this lies the tangible commercial potential of using Athena and in external projects, and not just within BTI.
Source: https://habr.com/ru/post/244769/
All Articles