Artificial Intelligence

In 2019, we have started our visually communicatable artificial intelligence program. We wanted to communicate to our followers, how we can use Artificial Intelligence on semi-routine tasks that humans perform. Demonstrating this on visual platforms, and using the right tools were one of our key aims. Visually communicating our ability and capability on mastering AI models would make our followers / potential customers fully understand the possibilities of which processes of their businesses can be automated.

Neural Networks

We started modelling the problem and the Neural Networks to be able to play a Flappy Bird like game. We've created 2 releases v.1.0 and v.1.1 on it

Reinforcement Learning

As we follow trends and try to solve the AI problems the best way that adapt to the problem, we've focused on Reinforcement Learning. In a nutshell, Reinforcement Learning is a machine learning technique, that takes the future possibilities and statuses into account when deciding on an action. That is why it makes it a great technic to teach the computers how to play Atari Games, as humans always think on a few steps further before acting on an Atari Game. Of course, this can be applied to real world problems too. We know that the key problem and focus on Robotics and the AI that feeds them is the balance. So we've focused on the best machine learning method that ensures solving similar problems like balance.

DQN

In order to optimize the Reinforcement Learning, another UK Company called Deep Mind released an article in 2015, and named the new technic DQN in their research paper. The goal of this algorithm was to make AI play computer games in a more human-like manner. We used this algorithm to train our Acrobat model. The Acrobat, Epifanic AI v.3.0 actually tries to make the best movements in order to reach the height and make a round turn.



We are moving forward with our research on improving our knowhow so that we articulate semi-routine processes in business models in order to make them applied by Machine Learning / AI trained models.