Synthetic Intelligence

= "the synthesis of human and machine intelligence in pursuit of insights or creativity that would be impossible for either on their own". [1]

Description

Benjamin Bratton:

"Synthetic Intelligence refers to the emergence of artificial machine intelligence in both anthropomorphic and automorphic forms as well as a complex and evolving distributed system. In contrast with many other approaches to AI, we emphasize ① the importance of productive misalignment and the epistemological and practical necessity of avoiding alignment overfitting to dubiously defined human values and ② the eventual open-world artificial evolution of synthetic hybrids of biological and non-biological intelligences, including infrastructure-scale systems capable of advanced cognition."

(https://research.antikythera.org/)

Discussion

Benjamin Bratton:

"It is almost certain that today the growth of machine intelligence is hamstrung by various ideologies of “artificial intelligence,” which are in turn hobbled by misconceptions about what is and is not artificial and what is and is not intelligent. Foremost among these is the presumption that machine intelligence must be recognizably “human-like” to qualify as intelligence. Multiple anthropomorphic biases and presumptions have left us with inadequate allegories for the remarkable things that machine intelligence does accomplish. Most of these look nothing like human thought — though some do, like the very large natural language processing models.

Recently, researchers at the Moscow-based Strelka Institute and I have been revisiting the distinction between the “artificial” and the “synthetic” posed by the economist Herbert Simon half a century ago. The artificial refers to something that merely resembles an original (such as a cheap plastic “diamond”) whereas the synthetic is a genuine and meaningful version of something that was deliberately created (such as a laboratory-grown diamond identical to a “natural” one at the molecular level). Thus, artificial intelligence merely seems smart, but synthetic intelligence really is. We should be pursuing synthetic intelligence, not artificial intelligence.

There is another connotation of synthetic intelligence that is perhaps even more important: the synthesis of human and machine intelligence in pursuit of insights or creativity that would be impossible for either on their own. A now-famous example of this occurred in the Go match between Lee Sodol and AlphaGo in 2016. The AI’s move 37 in the second game was one that Go experts have said no human could have imagined. But in the next game, Sodol’s move 35 was equally unexpected and creative. If the first move proved that AlphaGo was in some way not just “smart” in a narrow sense, but also capable of creating novelty, the second move proved that in response to this, a human saw the game differently and so produced a brilliant move that also never would have happened otherwise. This is a synthesis of intelligences, a glimpse of what a general sapience may look like."

(https://www.noemamag.com/planetary-sapience/)