AI Is Learning to See the World—But Not the Humans Way

POSTED BY ch@tbot   |   IN Chatbots

Using logic, developing hypotheses and solving problems are some of the activities that require human intelligence. When a computer or even a computer program does so, it is considered as artificially intelligent. Inteladesk provides artificial intelligence services to solve problems and execute tasks just like human being. Some of these tasks are simple like recognizing the object in a photo and others are complex like driving a car. Machines can surpass human brain when it comes to things like chess games, but such devices are limited to their programming nature.

Machine learning

In the early days of research, computer scientists and engineers thought they can achieve by feeding a computer program with specific rules to follow. Example, if x occurs, do y, if not do z. With machine learning, a computer becomes intelligent through learning from experience just like a human. For example, if you want your computer to identify animals, you show it thousands of labelled photographs. The machine algorithm will then learn the distinguishing features of one animal from the rest. The computer then uses this information to develop a model capable of matching an input to the respective output.

Using photos that are unfamiliar to the algorithm as a test data, ask it to predict its output based on what it has learned. The more experience the algorithm gets at identifying animals, the better it becomes regarding giving the right output and identifying animals. That’s machine learning; where you program the machine to learn how to tackle a particular task so that it becomes better at performing the specific task over time. In simpler terms, you will be using the inputs and outputs to develop a program that can eventually solve a problem.

Machine learning

Deep learning

Artificial neural networks and Deep learning are future artificial intelligence concepts that can make you scratch your head initially. However, Inteladesk explain to their clients in a very simple and understandable language.

Machine learning applies one filter and a single effect. Deep learning it’s about stacking effects. It’s like increasing the brightness of the photograph, saving it and then taking the brighter photo and saving it, and then taking that brighter version to increase saturation, and so on. As expected, this may not be the perfect analogy. In reality, the effects and filters are layers of neurons forming a neural network.

These neural networks use multiple, stacked layers to process information. At each layer, data is transformed and transmitted and submitted to the next layer. It is an iterative process where each layer takes the output of the previous layer as its input.

Natural language processing

Natural language processing (NLP)

The goal of NLP is to let computers understand the language used by humans to communicate. Though computer scientists have programmed computers to understand and speak the human language, it has become more apparent that learning it is the only way to understand it. That’s the point where machine learning comes to play. Although to be honest you do not need machine learning to do NLPs. It is more of classes of problems rather than a particular method.


Firms need to be ready and establish a new training program for the jobs that AI chatbots will create. Office assistants and truck drivers need to be retrained to create data analysts and trip optimizers respectively. Those in other professions too need to be prepared as it is not certain what AI may bring. To achieve an AI that matches that of a human, just contact Inteladesk. The artificial intelligence system in your vehicle will get you home, but may not choose another destination once you are home. From there, we will need primary drivers with moral values and emotions. Creating machines capable of learning like human beings brings a possibility of the same machine inheriting human-like flaws and qualities. As such, any progress in AI is likely to be open and gradual and may require sharing information between academia, companies, and labs.