Stanislaw discusses characterization of intelligence by starting with thoughts about information. He characterizes language, signs, and their composition. The next section, Doubts and Antinomies, covers the creation of novel ideas in computing, like the Perceptron. He describes a virtuous cycle of creation of systems to interact with existing systems as a way of more deeply understanding said systems. This leads to the establishment of a psychological division between symbolic problem solving, like language processing, and real-time decision-making, as in games like Chess. He presents an ironic point that it’s possible to create systems which are optimized for rapid arithmetic, but doing so as a human, or mathematical processing as a whole, can be quite difficult. Extending this, considering the perceptron, the author analyzes whether certain types of tasks or analysis or certain types of data can easily be automated or learned. The next clear question is how this compares to the existing limits of human intellectual capacity. An important point, which is highly relevant to the ideas being considered in this quarter’s Augmented Intelligence seminar (CSE 590A), is of collaboration between humans and machines, where the author states that machines are better at search and selection, an important component of the human process, though not all of it. Another component is the claim that human society will continue to become increasingly complex as associated with technology. Finally, a comparison is made involving the social nature of humans: the idea that society introduces more through social contracts than could be easily simulatable or replaced in a computer system.
The Piraha paper, as referenced in lecture, focuses on the idea that the Piraha language does not fit Chomsky’s Universal Grammar due to a wide array of violations. This is important because it frames certain concepts to which we frequently allude as societal, not inherent. One particularly relevant reference may be the fact that Piraha minimizes ambiguity, which may be useful for computer understanding, especially if posed in translation from standard languages. References to origin stories mention that the lack of these in concrete form stems from a perception of stories as tools, not foundations. In the comments, a constant criticism is the idea that the paper either doesn’t analyze the Piraha sufficiently deeply, especially relative to other similar cultures, or that it provides too little contextualization and expression of exclusion from universal human language.
The Graziano piece cites itself to uniquely discuss consciousness in terms of human society and interpersonal interactions. One example of this interplay is the idea that is presented that a similar part of the brain handles self-perception as handles perception of others. A fascinating analogy, Graziano paints a picture of a patient who imagined he had a squirrel inside his head, illustrating the perception of consciousness that each of us have. The description that Graziano poses of attention makes RNNs make vastly more intuitive sense. As a final point driving intrigue, a comparison is made to spirituality, which sets even more unique context for consciousness.