0
$\begingroup$

Richard Feynman has made in 1974 a famous speech about cargo cult science.[1] He described a traditional tribal society who has no access to modern technology but imagined how it feels to manage an airplane landing field. The tribe has build planes out of wood, which weren't able to fly but are utilized in a ritual in which science was only simulated.

The topic of aviation isn't very interesting because the technology is available. In contrast, Artificial Intelligence is something which isn't ready today and the only opportunity to become familiar with it, is doing some kind of cargo cult science. That means, AI isn't the reality, but it's imagined with the help of prototypes build out of wood who aren't working in reality. One example in doing so is a teleoperated robot hand. Teleoperation means, that no AI software was utilized but the signals from the human operator's dataglove are transferred as raw data to the robot hand. If the robot hand grasps a ball it's not real artificial intelligence because the human has made the task. He is simulating the overall domain and nobody knows how to build a software which can do the task autonomously.

Is a teleoperated robot hand similar to an tribe's airplane landing site? Is it some kind of proto-science which imitates the desired technology?

[1] https://en.wikipedia.org/wiki/Cargo_cult_science

$\endgroup$
  • $\begingroup$ I would actually upvote except for the statement "Artificial Intelligence is something which isn't ready today" which is counter-factual and wholly relative. (In fact, contemporary AI is demonstrating utility in a host of real world applications.) $\endgroup$ – DukeZhou Jun 24 at 21:06
3
$\begingroup$

I do not see how that would equal cargo cult science. Cargo cult science is emulating the symptoms without understanding the cause, in order to get the desired result.

In Feynman's example, the landing strips, planes, and watchtowers were not the cause of the cargo being delivered. It was the technologically advanced culture that brought the cargo. The tribal society did not understand that the planes were just a means to an end (it could just as well have been a boat).

Teleoperation is simply controlling a machine from afar. Popular examples include ROV's, UAV's and remote surgery. These pieces of technology do exactly what they were designed for; there's no gap in understanding between the cause, means, and the end. This is where it fundamentally differs with cargo cult science.

$\endgroup$
0
$\begingroup$

A key distinction is that AI methods demonstrate utility.

While it's true that many research endeavors could be seen as akin to "building a landingstrip (a new AI) in hope of attracting a plane (a real world application)", a more suitable analogy would be building new airplanes that perform better than previous models.

Specifically--a cargo cult building a landingstrip has no connection to the flow of commerce, and will not magically produce cargo. But a new type of AI can produce utility if it is applied to the right kind of problems.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.