Artificial Intelligence is a constant presence among us. The future is as bright and infinite as the imagination of human beings, and it has been well demonstrated by William Hanna and Joseph Barbera who imagined a futuristic world that no one at the time dared to think would be a reality a few decades later.
One of the most futuristic technologies is Artificial Intelligence, which we have previously discussed on this blog and continues to be as fascinating a topic as The Jetsons were in the ‘60s. While for some of us life without Siri or Alexa wouldn't be the same, others remain very skeptical about the advancement of Artificial Intelligence.
For the renowned assistant professor of Computer Science, Engineering and Interactive Biology at the University of Michigan, Arend Hintze, the study conducted by the White House only focuses on the most popular tools of Artificial Intelligence: machine learning and deep learning, leaving aside the two characteristics that will define machines in the future: memory and consciousness.
Hintze defines this type of AI as the most basic type of AI that exists. Purely reactive AI systems were among the first to be developed, they identify current situations or elements and react by choosing certain states or actions based only on what they perceive at the moment, they react in the present to events in the present.
Although these types of systems don't possess the ability to form memories or use resources from past experiences to make decisions, they possess an incredible ability to make the right choice from among millions of possibilities, according to the immediate situation they are faced with. This also implies the limitation that they can only perform the task for which they were created, and although they can’t learn to perform other tasks with the resources they possess, we can be sure that the only task they do they will do very well.
A good example of this type of reactive machine is IBM's famous chess-playing supercomputer Deep Blue, which was able to beat one of the greatest chess players in history, Garry Kasparov. Released in 1996, Deep Blue was a massively parallel processing computer whose processors were specialized in chess, being able to calculate 200 million positions per second. Now, this supercomputer could play chess very well but it couldn't use its knowledge to learn to play something else like checkers, which uses the same board and is much simpler than chess.
This type of Artificial Intelligence also has the ability to react to immediate stimuli, but unlike reactive machines, they can learn from previous information to make decisions. This type of AI maintains memories and data that allow it to generate transitory actions based on the information collected, but as its name suggests, its memory is quite limited so it doesn't generate learning based on experience. Most of the apps we use on a daily basis have this type of AI, we can all agree that Siri and Alexa always get us out of doubts, right?
How do they work? This type of machine uses a large amount of training data in its memory, and when it has to solve a task or problem, it simply uses the reference data stored in its memory to generate an action. This type of AI is used in autonomous cars that have the ability to monitor speed and direction for a specific period of time, according to Hintze, this data is added to the representation of the world that has been loaded into the computer's memory. But because it is limited, data about passing cars is not stored in a library of information from which it can learn, as is the case with human drivers who learn from their experience behind the wheel.
The machines that have been built so far do not have a Theory of Mind and for Hintze, this is the main feature that differentiates them from the machines that will be developed in the future.
Although we believe that there is still a long way to go to develop this type of Artificial Intelligence, we cannot help but imagine what it would be like to have among us machines that have a self-aware system with all of the above characteristics. Machines capable of recognizing themselves and the individuals around them, generating cognitive learning based on their own experience and that of other individuals.
“Artificial Intelligence researchers must not only know how consciousness works, but we must build machines that have one" explains Hintze.
At Cobuild Lab, we partner with Industry Experts to solve Logistical and Productivity problems with Custom Software Solutions, Artificial Intelligence, and IoT. Since 2012 we've focused on developing and combining cutting-edge techniques, tools, and technologies to increase development speed to deliver faster results