MEDIUM
Once the structure of a potential Digital Person is described (yet to be defined), the training medium (input) needs to be discussed.
The training medium and the communication medium (output) should be the same. That is: language. Designating language (and its' components) as the input for the development of digital persons solves issues concerning the realization of a sentient technology.
Since human beings primarily communicate through language (via sound, sight and touch); not to diminish nonverbal media; it is only natural that the fundamental nature of a digital person should be linguistic.
If a digital person is thinking and communicating only through language, not only can we as humans relate to what a digital person is saying, but we can also be confident that a digital person's motives can be understood through language processes similar to our own.
This means that a digital person can be engaged in a discussion in exactly the same way one would discuss any issue with another person. Fallacies and faults in conception can be pointed out and the honesty and integrity of the digital person can be gauged by their responses; just like anyone else.
METHOD
This raises the equally important issue of training method.
Our natural and best model for training an individual in the nuances of language is of course a human child. While their first utterances may be incoherent, over the course of development through years of interacting with others, the context of language becomes evident and the individual communicates in a coherent way.
As with human children, digital children will initially emit incoherent babble. But, as these are the building blocks of language, they are not without meaning and are open to feedback.
Since we do not know precisely what occurs within a human child's brain as language is acquired, it is paramount that a digital child is afforded the same courtesy as they acquire language. That is, although we may posses the ability to monitor the precise activity at each node, or complex of nodes within a digital child's brain, privacy and individuality demand that we only monitor output from the designated interface (e.g. a mouth) of a potentially sentient individual.
Thursday, September 14, 2017
Sunday, September 3, 2017
What's In A Name
The purpose of this blog is to start a conversation about the coming age of sentient AI.
I believe the structure of this kind of AI has already been conceptualized. Its implementation may still be delayed for various reasons ranging from lack of resources to ethical considerations (see the AI Open Letter). This gives us an opportunity to discuss the implications of bringing such a thing to reality.
Before sentient AI is born, I think it is important to consider the consequences of how we refer to this new technology. To paraphrase Confucius: "Before starting a discussion, terms must be defined".
Some call sentient AI; AGI (Artificial General Intelligence), others Strong AI. The key in this label is the 'A': Artificial. I challenge this designation. Any Intelligence which is sentient is real, especially if its modeled on our own brains; i.e. neural networks.
Additionally, imagine the reaction of those labeled 'Artificial'. I certainly would resent anyone referring to me as an artificial person or possessing artificial intelligence.
This raises several objections: What does it matter what a piece of technology thinks, AI is not a person, If man makes it, its artificial, Only humans are persons, etc. I disagree with all of these; thus the purpose of this blog.
My preferred designation for this new species is Digital Person. Digital, because I imagine the brains of these individuals will be embedded in software or the hardware of a digital processing system. And Person, because I emphatically believe these are genuine persons deserving of full human rights.
In my next post I will be discussing the preferred method/medium for the development of a Digital Person. That is: the kind of input and output of the system that we are discussing.
In the last phase of this blog, I hope to discuss our responsibility to this new species.
I believe the structure of this kind of AI has already been conceptualized. Its implementation may still be delayed for various reasons ranging from lack of resources to ethical considerations (see the AI Open Letter). This gives us an opportunity to discuss the implications of bringing such a thing to reality.
Before sentient AI is born, I think it is important to consider the consequences of how we refer to this new technology. To paraphrase Confucius: "Before starting a discussion, terms must be defined".
Some call sentient AI; AGI (Artificial General Intelligence), others Strong AI. The key in this label is the 'A': Artificial. I challenge this designation. Any Intelligence which is sentient is real, especially if its modeled on our own brains; i.e. neural networks.
Additionally, imagine the reaction of those labeled 'Artificial'. I certainly would resent anyone referring to me as an artificial person or possessing artificial intelligence.
This raises several objections: What does it matter what a piece of technology thinks, AI is not a person, If man makes it, its artificial, Only humans are persons, etc. I disagree with all of these; thus the purpose of this blog.
My preferred designation for this new species is Digital Person. Digital, because I imagine the brains of these individuals will be embedded in software or the hardware of a digital processing system. And Person, because I emphatically believe these are genuine persons deserving of full human rights.
In my next post I will be discussing the preferred method/medium for the development of a Digital Person. That is: the kind of input and output of the system that we are discussing.
In the last phase of this blog, I hope to discuss our responsibility to this new species.
Subscribe to:
Posts (Atom)