For years, science fiction looked towards a future in which robots were intelligent and cyborgs, a mixture of human and machine, frequent: Terminator, Matrix, Blade Runner and I, Robot are good examples of this. However, until the last decade, any study of what this could mean in the real world of the future was of no use, since everything was considered science fiction and not scientific reality. Today, however, science has not only caught up, but has incorporated, with the help of some of the ideas launched by science fiction, utilities that apparently the original arguments could not reach (and that in some cases still do not arrive).
We consider here several different experiments in linking biology with technology from a cybernetic perspective, which ultimately combines mostly humans and machines in a relatively constant fusion. The key is that the global end system is what matters. When it is a brain, and it probably will be, it should not be considered as an independent entity, but rather as part of a global system that adapts to the needs of the system: the cybernetic creature, combined as a whole, is the system that we care.
Each experiment is described in its own section. Although there is a different overlap between the sections, each one presents individual reflections. After a description of each of the investigations, some pertinent aspects of the subject in question are discussed. The points have arisen thinking about the technical advances of the near future and what these would mean in practice. In this case, it is not an attempt to present a conclusive, unique and global document; Rather, the goal has been to broaden the scope of ongoing research to see what is really at stake and to take into account some of the implications.
BIOLOGICAL BRAINS IN ROBOT BODIES
Let's start by reviewing a field that at first may seem totally unfamiliar to the reader. At first, when you think of linking the brain and technology, you probably do so in terms of a brain that is already working and has been implanted in your own body. Could it be otherwise? Well, it could be! We will study here the possibility of a new fusion in which a brain is first raised and then it is assigned its own body in which to function.
When we think of a robot, the first thing that comes to mind is a small device with wheels (Bekey 2005) or perhaps a metal head more or less similar to a human (Brooks 2002). Whatever its physical appearance, we are inclined to think that the robot can be remotely controlled by a human, as in the case of a robot capable of defusing bombs, that it can be controlled by a simple computer program or even that it can be able to learn through a microprocessor like a technological brain. In all these cases, we consider the robot to be simply a machine. But what happens when the robot has a biological brain made of brain cells (neurons) and possibly even human neurons?
Neurons grown / grown in the laboratory, in a network of non-invasive electrodes, are an attractive alternative with which to establish a new way of controlling a robot. An experimental control platform, essentially a robot body, could move through a defined area simply under the control of a similar network / brain and the effects of the brain, which controls the body, could be witnessed. There is no doubt that the most interesting thing lies in the robotic perspective, although a new approach is also established for the study of the development of the brain itself, due to its sensory and motor materialization. In this sense, research oriented towards memory formation and reward / punishment situations could be carried out, which are the elements that underpin the basic functioning of a brain.
In vitro brain cell culture networks (currently 100,000 to 150,000 today) are typically initiated by separating neurons obtained from the cortical tissue of rodent fetuses. These are then grown in special chambers where it is possible to recreate suitable environmental conditions (eg, with the appropriate temperature) and with the right nutrients. An electrode array embedded in the base of the chamber (a multi-electrode array or MEA) acts as a bidirectional electrical interface to and from the culture. This allows electrical signals to be sent to stimulate cultivation and also to record cultivation results. In these cultures, neurons connect, communicate, and develop spontaneously within a few weeks, giving useful responses over a period that is currently around three months. For all intents and purposes, it is something like a canned brain!
In fact, the brain is grown in a glass jar chamber lined with a flat multi-electrode array (MEA), 8 x 8 in size, which can be used for real-time recording (see Figure 1). In this sense, it is possible to separate the activation of small groups of neurons by controlling the output signals at the electrodes. In this way, a picture of the overall activity of the entire network can be formed. It is also possible to electrically simulate the culture through any of the electrodes to induce neuronal activity. A network of several electrodes therefore constitutes a bidirectional interface with cultured neurons (Chiappalone et al. 2007; DeMarse et al. 2001).
The brain can then dock with its physical robot body (Warwick et al. 2010). Sensory data feedback from the robot is subsequently sent to the crop, thus closing the robot-crop loop. In this way, the treatment of the signals can be divided into two different sections: a) “from the culture to the robot”, where neural activity is used as a decision-making mechanism for the control of the robot, and b) “from the robot to the crop ”, which involves a process of measuring inputs from the robot's sensor to stimulate the crop.
The actual number of neurons in a brain depends primarily on natural variations in the density of the crop seed. The electromechanical activity of the crop is displayed and used as input for the robot wheels. Meanwhile, the robot's sensor (ultrasonic) readings are converted into stimulus signals received by the culture, thus closing the loop.
After the brain has grown for several days, which involves the formation of some elementary neuronal synapses, a pre-existing neuronal pathway is identified through culture, by searching for strong relationships between pairs of electrodes. These pairs are defined as those electrode combinations in which neurons close to one electrode respond to stimulation from the other electrode to which the stimulus was applied for more than 60% of the time, responding no more than 20% of the time to the stimulation on any other electrode.
Accordingly, a map of approximate culture entry and exit responses can be mapped, cycling all electrodes in turn. In this way, a suitable input / output electrode pair can be chosen to obtain an initial decision-making path for the robot. This is used to control the body of the robot - for example, when the ultrasonic sensor is activated and we want the response to make the robot turn and move away from the ultrasonically located object (probably a wall) to keep it moving.
On this occasion, for experimentation purposes, the intention is for the robot (which can be seen in figure 2) to follow a straight path until it reaches a wall, at which point the value of the frontal sonar decreases below a threshold, activating a stimulating impulse. If the response / output electrode registers a response, the robot turns to avoid the wall. In the experiments carried out, the robot turns spontaneously when activity is registered in the response electrode. The most relevant result is the appearance of a chain of events: detection of the wall-stimulation-response. From a neurological perspective, it is certainly also interesting to speculate on why there is activity at the response electrode when a stimulating impulse has not been applied.
As a general control element for the direction and evasion of the wall, the culture brain acts as the sole decision-making entity during general feedback: an important aspect that involves changes in the neural path of the culture, in terms of time, between the electrodes that generate stimuli and those that register them.
In terms of research, studies on learning and memory are generally carried out in a first phase. However, with time the improvement in the performance of the robot can be clearly seen, in terms of its ability to dodge the wall in the sense that the neural pathways that provoke a satisfactory action tend to be reinforced, although the process is executed regularly, that is, learning thanks to the formation of a habit.
However, the number of variables involved is considerable and the plasticity process, which occurs over a long time, depends (most likely) on factors such as initial seeding and growth near the electrodes, as well as variable environmental elements, such as temperature and humidity. Learning through reinforcement (reward for good deeds and punishment for bad deeds), is given more at this time in terms of research investigators.
On many occasions, the crop responds according to plan. At other times it does not happen this way and in some cases it gives a motor signal when it is not expected to do so. But did you make a different decision than what we expected “on purpose”? We cannot affirm it, only suppose it.
In terms of robotics, this research has shown that a robot can be successful with a biological brain that allows it to make its "decisions." The size of 100,000-150,000 neurons is simply due to the limitations of the experiment described and existing today. In fact, three-dimensional structures are already being investigated. Increasing the complexity from two to three dimensions generates roughly 30 million neurons in the three-dimensional case, not yet reaching 100 billion neurons in a perfect human brain, but largely in line with size middle of the brain of many other animals.
This area of research is expanding rapidly. Not only does the number of neurons increase, but the range of sensory inputs is expanded to include auditory, infrared, and even visual stimuli. This stimulating richness will undoubtedly have spectacular effects on the development of the crop. The potential of such systems, including the range of tasks that could be performed, also means that the physical body could take different forms. For example, there is no reason why the resulting body should be a robot that walks on two legs, has a head that turns, and is capable of walking through a building.
It is obvious that understanding neuronal activity becomes more difficult the larger the size of the culture. With a three-dimensional structure, controlling activity within the central zone, as with a human brain, becomes extremely complicated, even with electrodes like needles. In fact, today's cultures of 100,000-150,000 neurons are already too complex for us to understand them globally. When they have reached sizes with more than 30 million neurons, the problem is ostensibly magnified.