The claim that many of the services that call themselves `` AI '' are false things



There is criticism that some of the services provided by companies as artificial intelligence (AI) have not been examined to see if they work well. Irvin Narayanan, a computer scientist at Princeton University, explains why this is happening.

How to recognize AI snake oil
(PDF file)

Mr. Narayanan uses “AI system for evaluating candidates for recruitment” as an example of artificial AI provided by companies. This system can evaluate a person by analyzing the characteristics of the body language and conversation, not the content of the candidate recruited at the interview. However, Mr. Narayanan has criticized, “It is only an elaborate random number generator”.

`` Companies that provide AI-based personnel systems are raising hundreds of millions of yen together and actively targeting customers, '' said Narayanan, who seeks job seekers with fake AI I am worried that will get worse.


Waseem Farooq

Narayanen pointed out that 'AI' is only used as a comprehensive expression of a series of related technologies. While some of the technologies that have been named AI are actually using artificial intelligence, some companies have made use of the fact that the general public has little understanding of AI. It is criticized that there are a lot of things that are labeled.

For example, DeepMind ’s Go AI “ AlphaGo ”, which won the top players, is a brilliant artificial intelligence, a level of research that most experts do not think would be possible 10 years ago. However, AlphaGo's achievements have nothing in common with a tool that predicts the candidate's ability to work.

Narayanan said companies that promote 'services using AI' as a solution to all problems have been helped by generous media. As a result, Americans even think that 'all work will be automated in 10 years'. AI experts conservatively estimate that it will take about 50 years for powerful AI to appear, but `` Looking at history, even experts are very optimistic about predicting artificial intelligence ”Says Narayanan.

by deepak pal

AI is already expected to be able to work with higher accuracy than humans in face recognition, medical diagnosis, transcription from voice to text, deep fake, etc., and research continues to improve technology every day. AI works without any uncertainty or ambiguity, and AI can learn to distinguish human faces if there is sufficient data and computational complexity. Although the accuracy and reliability are still not completely high, Narayanan evaluates that further research will improve the accuracy of AI. But “In some tasks AI is better than humans, but social outcomes cannot be fully predicted,” Narayanen claims.

“In order for AI to develop correctly, we must resist tremendous commercial interests. In most cases, evaluating someone with human hands is more accurate than AI. It ’s much more transparent and worth considering. ”

in Software, Posted by log1i_yk