Neural networks often make mistakes and fully relying on them can be dangerous, according to Vladislav Tushkanov, head of Kaspersky Lab’s machine learning technologies research and development group, RIA Novosti reported on February 29.
“Now the user may encounter a situation where the device speaks in an understandable language, with beautiful words, in a confident tone, but the information itself is distorted or incorrect. We, as a human species, were not prepared for this. If something speaks well, speaks beautifully, especially if it is a computer, then it is a reasonable thing and we tend to believe it, but it can be wrong, and often so.”– said Tushkanov.
The problem described by the expert is called overreliance (excessive trust or dependence). At the same time, people begin to rely too much on the help of neural networks to optimize processes and facilitate some of the routine intellectual work, and mistakes are made.
“These are probabilistic algorithms, they imitate human speech, the text written by a person, but this is not always objectively correct and the system still confidently presents it to us as the truth; These situations are called hallucinations. And the more accurate and useful neural networks are, the more we will use them, but we must always remember that they can and will make mistakes; trusting them too much is the main risk.”says Tushkanov.
The expert doubts that people will ever be able to trust neural networks 100%.
Source: Rossa Primavera
I am Michael Melvin, an experienced news writer with a passion for uncovering stories and bringing them to the public. I have been working in the news industry for over five years now, and my work has been published on multiple websites. As an author at 24 News Reporters, I cover world section of current events stories that are both informative and captivating to read.