There is still a long way for AI to gain human tru

  • Detail

How can artificial intelligence gain human trust? There is still a long way to go

China] although artificial intelligence is booming, the executives of major companies still take a cautious attitude towards this technology. One area worthy of attention is the ethics of artificial intelligence and its potential business value. 70% of companies even have special ethics training for IT personnel engaged in this technology

recently, Forbes, together with SAS software research institute, Accenture and Intel, investigated the attitudes of 305 enterprises towards AI ethics, confidence and business value. Just according to technicians, 51% said that the impact of AI based technology deployment on their company's operations has been "successful" or "very successful.

the survey report defines AI as" a training system science that automatically simulates human tasks through learning ". AI adopters said that today's moral constraints are relatively strong. 63% of them affirmed that they "have an ethics committee to review the use of AI", and 70% said that they "provide moral training for their technicians". According to the report, 92% of the companies that have achieved real success through the deployment of artificial intelligence said that they provided training for technical experts on ethical testing machines as a necessary equipment to control product quality, while 48% of the companies that have not achieved real success in the deployment of artificial intelligence

nevertheless, as Thomas Davenport said in the current MIT Sloan Management Review, AI needs greater transparency. He advised AI implementers to be cautious: "whether using internal or external cognitive technology, we need to cover everything."

artificial intelligence

the key is that the development of artificial intelligence needs to be open and transparent, Davenport explained, "when introducing new functions, 9. Outlet of exhaust fan: diameter 150mm (exhaust fan is installed at the back of the box) Launch it as a beta product and convey the use goal of the technology. Don't cancel the alternative solutions to solve the problems of employees or customers, that is to say, the role of human beings cannot be cancelled. Over time, as technology matures and AI solutions improve, the situation will become better and better. "

Davenport even proposed a third-party artificial intelligence algorithm certification method, just like "the efficacy of drugs certified by the food and Drug Administration (FDA), the financial process certified by auditors, and the safety of products certified by underwriter laboratories". It is time to independently certify "the reliability, replicability and accuracy of AI algorithm", he urged

in a word, the success of AI depends on the trust of stakeholders. Forbes and SAS are globally renowned specifications that are revised every year. Accenture, Intel's joint survey results show that: "Business and government leaders will do this, but many of them naturally worry about their employees' lack of trust in AI, which is reflected in their concerns about the impact of AI on their companies. There are even signs that some of them have doubts about the extent of AI output. Trust may be more useful for the development of AI than any technology or progress."

trust in AI "requires education, transparency, high accuracy and sensitivity in the measurement and control of load, deformation and displacement, clear moral standards, and patience"

Copyright © 2011 JIN SHI