Artificial intelligence? Machine learning? Deep learning?
Artificial intelligence is the attempt to reproduce or simulate intelligent behaviour with machine learning and deep learning forming sub-categories.
Programmes using machine learning can employ algorithms to predict people's actions in order to, for example, detect credit card fraud. Deep learning goes a step further and uses hierarchical layers in order to carry out the process of machine learning. That means that artificial neural networks are developed that are akin to the human brain. Complex concepts can be learned by building them from simple, hidden layers. Data is processed in the first layer before being moved up to the next where it is processed further, before being moved up to the next, and so on. This model can have endless numbers of layers and go to incredible depths, and is the reason behind the name “deep learning”.
Deep learning technology is being utilised, for example, in driverless cars to differentiate between road signs, other cars and pedestrians. It also takes the form of intelligent voice control in computers and smartphones. The application possibilities are practically infinite and much more precise than machine learning. The only disadvantage is that deep learning needs a lot of computing power.
How many levels of artificial intelligence are there?
Three to five levels of artificial intelligence are currently recognised. At the first level, AI doesn't exert any influence and people are still in control of everything. The next levels are defined by their position on a scale of autonomy—how much influence AI has and how independently it can operate. This starts with assistant-style roles and reaches all the way to self-thinking and self-operating. Currently, however, AI is still operating at the lower levels, and many “machines” are not yet capable of autonomous working. But this could all change in the next few years. Self-driving cars are just the beginning.
Artificial intelligence requires a lot more computing power than any other current applications, in particular deep learning algorithms that rely on high performance capacities to analyse and interpret information piece by piece. This cannot be achieved without acceleration using Graphics Processing Units (GPUs). Powerful GPUs currently on the market include the NVIDIA® Tesla® graphics processor This enables companies to accelerate their most demanding high-performance computing (HPC) applications and hyperscale workloads in the data centre. This enables companies to accelerate their most demanding high-performance computing (HPC) and hyperscale workloads in the data centre. The advantage being that petabytes of data can be handled much faster than when using standard CPUs. Whether deep learning, energy research, virtual desktops or driverless cars, Tesla GPUs deliver enormous performance to enable extensive calculations and simulations in in next to no time.