Los Alamos National Laboratory researchers have developed a novel method for comparing neural networks that looks into the “black box” of artificial intelligence to help researchers comprehend neural network behavior. Neural networks identify patterns in datasets and are utilized in applications as diverse as virtual assistants, facial recognition systems, and self-driving vehicles.[1]

Figure 1. New Method Exposes How Artificial Intelligence Works

Figure 1 shows “The artificial intelligence research community doesn’t necessarily have a complete understanding of what neural networks are doing; they give us good results, but we don’t know how or why,” said Haydn Jones, a researcher in the Advanced Research in Cyber Systems group at Los Alamos. “Our new method does a better job of comparing neural networks, which is a crucial step toward better understanding the mathematics behind AI.” [2]

Neural networks are extremely highly effective however fragile. For instance, autonomous automobiles use neural networks to acknowledge indicators. They’re fairly adept at doing this underneath the proper circumstances. Nonetheless, the neural community can falsely detect indicators and can by no means cease if there’s even the slightest abnormality, resembling a sticker on a cease signal. [3]

There has been extensive effort in industry and in the academic community searching for the "right architecture" for neural networks, but the Los Alamos team's findings indicate that the introduction of adversarial training narrows this search space substantially. As a result, the AI research community may not need to spend as much time exploring new architectures, knowing that adversarial training causes diverse architectures to converge to similar solutions.

"By finding that robust neural networks are similar to each other, we're making it easier to understand how robust AI might really work. [4]

References:

1. https://scitechdaily.com/new-method-exposes-how-artificial-intelligence-works/
2. https://headtopics.com/us/new-method-exposes-how-artificial-intelligence-works-30989577r
3. https://naijaonpoint.com.ng/new-method-exposes-how-artificial-intelligence-works/
4. https://www.sciencedaily.com/releases/2022/09/220914102256.htm

Cite this article:

Thanusri swetha J, New Method Exposes How Artificial Intelligence Works, Anatechmaz, pp.185