Artificial Intelligence Suffers from The Biases of their Human Creators, that Causes Problems Searching for ET – Science Times
One structure on the dwarf planet Ceres made big news but there is a hitch. It seems that the square-shaped form inside a larger triangle, located in a crater. Everyone else saw it, but the use of artificial intelligence might be a square peg not fitting a round hole. This remark by a Spanish neuropsychologist is questioning the veracity of depending on AI, which might be unsound by SETI.
Ceres is located in the main asteroid belt, a dwarf planet, and the biggest object too. One of its craters, Occator had bright lights which lead to several ideas of what it was. Nasa sent the Dawn probe to go close enough to capture visual evidence of what these lights were and solve the mystery. These lights were from volcanic ice and salt eruptions, nothing more.
It gets more interesting as researchers based in the University of Cadiz (Spain) have examined images of these spots. Areas like them are called Vinalia Faculae, in an area where geometric contours are very evident for observers. It now serves as a template to compare how machines and humans perceive images on planetary surfaces in general. Tests like these will show artificial intelligence can see technosignatures of other lifeforms besides human-civilization.
During the test, more than one individual saw the squarish shape in Vinalia Faculae, and it was a perfect chance to test artificial intelligence on a human. Subjecting a human subject with what AI sees, is a comparison to see the result. In searching for ETs, radio signals aren't the only consideration but also captured images not like before.
To see further what their hypothesis would bring about. These neuropsychologists made more modifications from the previous experiments to dig deeper by adding another layer to it. Another batch of volunteers was conscripted, but this time were amateurs in astronomy to analyze what they see in the Occator image.
By comparison to the artificial vision system that is grounded on the convolutional neural networks (CNN), an AI taught to see squares and triangles to identify them. From this point on, the experiments got interesting as it progressed.
Researchers ran the experiment and this what the people saw, some peculiarities were observed, as the images were perceived.
a. AI and people saw the square structures and did not miss it.
b. Another thing is the AI did not fail to notice the triangle as well.
c. Whenever the triangle was pointed out to people, more mentioned seeing it.
d. Visually the square was inside the triangle as it was visually represented too.
This is what the neuropsychologists drew from the results of the experiment on the "amateurs". It was published in the "Acta Astronautica journal" for reference.
a. The application of artificial intelligence for use in finding ETs is not as foolproof. Just like human intelligence the AI can be mistaken, confused and make false perception as well.
b. AI can be applied for some tasks to find technosignatures and ETs in some exceptions. Overall, AI will not be implemented but with caution, especially in SETI.
c. The presence of biases in the programming of AI is that of their creators, which is unavoidable. The best move is to study artificial intelligence, when under human stewardship.
Originally posted here:
Artificial Intelligence Suffers from The Biases of their Human Creators, that Causes Problems Searching for ET - Science Times