Consciousness - An Introduction (ch18)

How should a conscious machine be built? - Chapter 18

Many robot and computer technicians ignore all arguments against designing machines with a consciousness and go stiffly with trying. There are two ways to tackle this:

1. make a conscious machine, or

2. make a machine that seems to be conscious .

We must remember that it may seem as if machines have intentions to achieve goals, while that may not be the case at all. In addition, it is surprising that we do not think that other objects experience subjectivity, while we treat them as if they did.

Kismet was the first robot to look like a human being. He had control over attentional processes and vision and watched movement and color. The mood of Kismet was a combination of three variables: happiness, alertness and openness to new stimuli. Kismet could hear and make sounds, but do not understand words. You might think that Kismet has no consciousness because it consists of metal and performs simple routine actions.

Yet it is true that there is no place in Kismet where 'everything comes together' (as is sometimes said about consciousness). In short: Kismet did not have a Cartesian theater . Dennett believes that this also applies to people. Suppose a new robot is created that recognizes emotions in people, laughs, can cry and can respond to the person he is conversing with. Would you still think that he has no consciousness? We can say that it only looks like this robot has a consciousness. Another possibility is to believe that he really has a consciousness. Which vision is correct and how do we find out?

Convictions and beliefs 

In 1979, McCarthy, one of the founders of AI, claimed that thermostats have beliefs: they can believe it's too hot, too cold or the temperature is right. A thermostat can observe the environment and respond to the environment on that basis.

It seems strange that McCarthy thinks that thermostats have convictions, but his example invites us to think about what real intentionality is exactly. You might think that the convictions of thermostats are as real as the beliefs of people, but that they are simpler. You might also think that there is no such thing as real intentionality and intentionality that seems real . Other people again make a clear distinction between intentionality and consciousness. If you believe that robots do not have real intentionality, then you should be able to explain what real intentionality means to people and whether it can be implanted in a machine.

X

Suppose people have an X that ensures that they have a consciousness. If we want to make a conscious robot, then we have to figure out what this X is. McGinn wonders whether this X (which he calls C *) can exist in objects. He concludes that we can never know that. He believes that we will never know how the brain and consciousness are connected. He therefore believes that we can never find out what C * is.

Other scientists are less pessimistic. Chalmers believes in artificial intelligence and rejects the thought experiment of the Chinese room from the previous chapter. Chalmers is a dualist, but believes that every system that is organized in the right way can have a consciousness like ours. According to him, having the right system is not enough. Implementing this system is very important. He states that we must find the X of human consciousness for this. One way to do this is to make a list of possible Xs

Stuart proposes ' engaged embodiment' ; purposeful animation, perception, imagination and the ability to recognize experiences as own experiences. Aleksander takes phenomenology as a starting point. He comes to the following criteria to make a conscious robot:

• perception of himself in an ' out there world' ;
• imagination of the past and fiction;               
• attention;             
• plan;       
• emotions.

Aleksander devised the Kernel Architecture (KA) for these five criteria. The key mechanism in this is depicting: creating a direct representation where elements of the world are located that ensure that attention can be directed in a suitable manner. According to Aleksander, KA also contains an image of the self in his world. He therefore concludes that a robot can be aware if he has a KA.

Another approach is to start from existing theories of consciousness. The ' Global Workspace Theory' (GWT) states that the content of the consciousness consists of representations that are processed by the ' global workspace' . The 'global workspace' is itself a large network with cooperating neurons. The content of these neurons would be aware and they would make this content available to the entire system that is unconscious. In this theory, the x factor stands for general availability. A robot would therefore have a consciousness if it had a ' global workspace' that could ensure that the content would be available for the entire system.

Franklin designed IDA (Intelligent Distribution Agent ) as software for the US Navy. This software is based on GWT, but according to Franklin, this software only has a consciousness and no self-awareness. In addition, the software would not have subjective experiences (' phenomenal consciousness' ). Franklin therefore draws the conclusion that IDA has no real consciousness.

Edelman and Tononi say that the consciousness is the result of a dynamic core, with neurons in the brains. This nucleus would have all kinds of connections with the rest of the brain. It might be possible to design a machine that also has such a dynamic core.

Hameroff and Penrose argue that the implementation of a dynamic core would not provide a real form of consciousness. They state that consciousness arises through coherence in microtubes of cells.

There are also scientists who think that consciousness is a big illusion. It would only seem like we have conscious experiences. We would have this illusion because we have language skills. We quickly think in terms of 'me' and the 'self' and that would create a sense of consciousness.

Talking machines

Many attempts have been made to design speaking machines. In the past, the GOFAI approach was mainly used . This approach has been used to program computers with the right rules. The GOFAI approach has also tried to teach robots language. The problem here is that natural languages ​​do not always follow certain rules. For example, words have multiple meanings and sentences can be interpreted in several ways. Machines can weigh different meanings and choose the most likely ones, but they will not really understand the languages ​​they learn.

A new approach is the ' evolutionary theory of memetics' . According to memetics, if organisms imitate each other, they then evolve. This could mean that if imitation occurs, languages ​​can arise spontaneously. The same thing would happen if machines can imitate each other. There is evidence from computer simulations and robot studies that there seems to be an evolutionary system. Steels has, for example, designed robots that can identify and imitate each other's sounds. Language could also be seen in the same way: a changed system in which words and meanings emerge spontaneously. This would mean that robots can learn language as children do. What's more, this means that robots that imitate each other can quickly come up with their own language.The memetics approach states that machines that can imitate each other are clearly different from machines that do not, just like people differ from animals. For example, robots could design their own culture, as we do.

Immortality

Kurtzweil believes that it is a matter of time that humans become immortal. We could do this by putting parts of brain functions on a computer. The first question is whether you still have a consciousness and the second question is whether you are the same (conscious) person as before. The answers to this depend on how you define consciousness and whether you are a supporter of bundle or egotheories.

Resources:Blackmore; Susan. (2010). Consciousness, Second Edition An Introduction. Abingdon, Oxon: Taylor & Francis.