Shulou(Shulou.com)06/03 Report--

Artificial intelligence is groundbreaking and sometimes even shocking. We keep getting amazing stories about efficiency, automation and intelligent prediction. However, artificial intelligence is not perfect. For every success story, there is also a chaotic or wrong story-the situation did not go according to plan.

Although we are loyal followers of AI and have seen the power of AI in our businesses, we sometimes see the other side of the coin and remember that we are all trying to promote greater development and better things. But along this road, there will be friction and disturbance. How we deal with these anomalies and defects ultimately defines the direction we start from here.

Example of AI failure

We don't want artificial intelligence to be perfect. At least not yet. The fact that AI still experiences failures with unintended consequences is refreshing. But even for those on the verge of mistakes, it can be used as a lesson for growth and improvement. Today, the more failures there are, the more growth and improvement there will be-which will eventually lead to a better and more complete industry.

Here are some good examples of AI failures over the past few years:

Microsoft AI Chatbot learned some unpopular languages.

Artificial intelligence chat robots have become the norm on social media and other websites. Facebook has an excellent tool built into Messenger that is used as a powerful business tool for providing customer service and quick conversations with potential customers. But the AI chat robot is not perfect. Take Microsoft's AI chat robot, for example, which was briefly called "Tay".

Tay was released in March 2016 and deployed for Twitter users and is programmed to use typical millennials to have casual and natural conversations. But it only lasted 24 hours. What happened? Well, the site targeted its loopholes and manipulated Tay to make very sexist and racist comments.

Microsoft AI and Peter Lee, vice president of research, had to apologize publicly for not anticipating the possibility.

An athlete or a felon?

Amazon has a project called recognition. This is an AI-based facial recognition software that has been sold to police agencies for investigation. In essence, it should cross-analyze images and direct law enforcement officers to potential suspects. The problem is that it is not very accurate.

In a study by the Massachusetts branch of the American Civil Liberties Union, dozens of Boston area athletes were run through the system. At least 27 of these athletes (or about 1/6) were mistakenly matched with facial photos. These include three-time Super Bowl champion Duran Harmon of the New England Patriots.

Can you say it's not pretty?

Users find defects in the Face ID of Apple

Apple has been coming up with excellent technology. They have set standards in the smartphone and mobile device industries. In most cases, they will get things done. But sometimes they may be too brave in marketing. One example is the release of iPhone X. Before the release, Apple had invested a lot of time and marketing resources in its frontal facial recognition system to replace fingerprint readers as the main way to access phones. Some people claim that AI components are so smart that readers can wear glasses, makeup, etc., without affecting the function. This is essentially right. The problem is that Apple has also made it clear that Face ID technology cannot be deceived by masks or other technologies.

A security company based in Vietnam sees it as a challenge. For just $200, they made a mask out of stone powder, pasted it on some printed 2D "eyes", and unlocked the phone. This is just a reminder that bold statements can sometimes get people's attention again!

The robot dog has a fatal ending.

Who doesn't like the idea of robot puppies? You'll get a cute little machine without barking, walking, defecating, eating or expensive veterinary bills. However, if you are looking for a life partner, you may not need this robot dog.

In 2019, Spot, a robotic dog from Boston Robotics, died on a dramatic and anachronistic stage when he was demoted by the company CEO at a conference in Las Vegas. The walking task made him stumble slowly and eventually fell to the ground as the audience gasped and giggled uncomfortably.

IBM's Waston is a great technology. The smart supercomputer has made many achievements under his leadership, including beating some of the smartest people in the world in dangerous games televised. But as far as Watson knows, he hasn't been trusted to be a doctor.

In 2018, IBM Watson tried to activate the medical AI system to make recommendations for the treatment of cancer patients. The goal of IBM is nothing more than "eliminating cancer". But hospitals and oncologists soon discovered major flaws. Watson has suggested that patients with excessive bleeding should be put on drugs that cause more bleeding, which may kill patients.

IBM blamed its engineers for programming Watson with hypothetical and fictitious cases rather than relying on actual patient data and historical medical records. 、

Voice deceives the CEO of AI software Cons

Fakes are becoming a serious (and shocking) problem. Hackers have found ways to forge sounds, pictures and even videos. In some cases, the consequences are catastrophic.

In March 2019, the chief executive of a British company called his boss from his German parent company. He was instructed to transfer the equivalent of $243000 to Hungarian suppliers. The request was marked as urgent and asked the CEO to execute it immediately. The only problem with the request is that the other end of the bank is not his boss. This is an AI-based software that mimics the voice of the boss.

When we call it AI failure, the reality is that AI software wins.

We can see clearly that artificial intelligence is not without problems. As far as business is concerned, the implementation of artificial intelligence still faces some challenges. They include:

The processing capacity is limited. Although AI and ML have great potential, they take advantage of a lot of processing power. Most calculations are not that advanced at all. As a result, it is difficult to make full use of these technologies outside of a very specific environment. Knowledge is limited. Only a few people really know enough about AI to explain it to the market. This has prevented the adoption rate from reaching its due level and is slowing growth. Lack of trust. There is always a certain degree of mistrust between people and computers. Although the relationship has improved over the past few years, failures such as those outlined in this article have not helped much. Data security. In order for AI applications to work, these systems need to access millions of data points. This creates possible weaknesses and vulnerabilities for hackers to target and compromise.

Despite these challenges, the beauty of AI lies in its intelligence and brilliance. Not only are many of the world's greatest minds committed to improving and perfecting technology, but when used in conjunction with machine learning (ML), you will have a self-refueling growth cycle, with each problem and mistake eventually paving the way for greater development. Efficiency, accuracy and opportunity. Speaking of opportunities, here are some of the most important opportunities I have seen:

Make it easy for big data. One of the problems companies have with big data is to find ways to understand it. Because there is so much information to organize, it is difficult to find out how to explain and apply discovery. Artificial intelligence can simplify this process and lead to better and more efficient processes. Superior analysis. Some AI systems can be used to monitor, detect, and analyze changes in various settings. Companies can use it to keep a close eye on competitors and understand price changes, public relations activities, social participation, and so on. Smarter search. The Internet makes the world work. In particular, search engines can identify trends and control the flow of information. Search engine companies such as Google are using AI and ML to change the industry and provide customers with more accurate and timely results.

Is AI perfect? As the story in this article shows, this is unacceptable! But is it powerful and constantly improving? Right. So when we look forward to the next one, three and five years, please remember that artificial intelligence is a work in progress. In order to achieve the goal we want, we must constantly solve some frictions in the process.

Https://ai.51cto.com/art/202008/622744.htm

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.