AI is always calm. It doesn't have working hours. And that makes it the ultimate tool.

When people ask questions from the AI it doesn't need to go anywhere. That system will never need to go home. It doesn't afraid that something happens to its family. And it will not need to drink or eat. The AI never goes to meetings. So that means the AI has endless patience. That makes AI an excellent doctor. If a patient asks for something unnecessary from the AI that system will not need to take another patient in its room. The AI might be connected with a diagnosis robot that has X-ray and ultrasound systems in its hands. 

And if the AI requires the room, it can offer the link to the patient's mobile telephone or guide that person to another room where it can continue the discussion. It might say "I'm sorry but I need that equipment with another patient, but we can continue our discussion outside the room". Or the AI can send a robot to talk with patients. In that case, the robot would be the external body of the AI. The AI runs on supercomputers. And the robots are connected to it by using WLAN. Those AI-based systems robots are external sensors that the AI can use for making things. 

AI played doctors and made the job better than humans. Then AI played judge, and the results were similar. The reason why AI is a better doctor than human is simple. The AI is never tired. The AI can follow the patient, and AI has time to answer those 200 questions without the need to visit the toilet. That thing means that AI makes a better job than humans. Another thing is that AI can see that people are using the right datasets when they make some decisions. 

That means the AI will not let emotions affect judgments. That means the AI should follow the voice and maybe sweat more effectively than humans. And the AI can also be neutral if somebody tries to make judges angry. This thing is a great point for AI. The computers will not get angry, and they have no secret relationships with the people, who are in the courtroom. 

"Machine-learning models designed to mimic human decision-making often make different, and sometimes harsher, judgments than humans due to being trained on the wrong type of data, according to researchers from MIT and other institutions". (ScitechDaily.com/When AI Plays Judge: The Unintended Severity of Machine Learning Models)

"The “right” data for training these models is normative data, which is labeled by humans who were explicitly asked whether items defy a certain rule. However, most models are trained on descriptive data, where humans identify factual features instead. When models are trained on descriptive data to judge rule violations, they tend to over-predict these violations, potentially leading to serious real-world implications".(ScitechDaily.com/When AI Plays Judge: The Unintended Severity of Machine Learning Models)

One thing that we must remember is that machine learning is not human. And creators of the AI are not professional judges. So that's why the judgments are overestimated and too strong. The thing is that the AI cannot operate as a "judge". It can make notifications like is the person nervous? Or is there some evidence that is not effective to questions? It can also see things like if some paper is left without notifications. In the worst case, the attorneys not even opened some pages of police reports. 

"Researchers have found that machine-learning models trained to mimic human decision-making often suggest harsher judgments than humans would. They found that the way data were gathered and labeled impacts how accurately a model can be trained to judge whether a rule has been violated. Credit: MIT News with figures from iStock"(ScitechDaily.com/When AI Plays Judge: The Unintended Severity of Machine Learning Models)

The AI-human combination is the ultimate tool. The AI is not advanced enough that it can make the same type of decisions as humans. But if humans are using the AI as a tool, that can help them to see if something is not as it should be, which makes the AI effective assistance. 

The AI can search for things like does the judge look at all evidence. And another thing is that the AI can make notifications like the witness seems to be angry or upset. It also can search for similarities with answers like some phrases from other courtroom records. If different people use precisely the same words in different cases, that is the mark that some of those testimonies practiced before the court. 

We all read the tests where U.S Marines used cow skin and cowboy boots for bluffing the AI. The reason why the AI didn't react to that thing was that it didn't have the image of a cow with boots. The "UFO algorithm" would make that thing impossible. That means if the AI sees something that is not stored in its database, it can call a human operator to see what it sees. 

Nobody predicted that somebody wears cow's skin and boots, and that image didn't store in the database. That's why the AI didn't recognize those men. But there is a possibility that if the AI has an algorithm that alerts human operators to see if there is something unknown that thing is easy to fix. The human operators should always look at the screen. 

And the AI might have a dataset about things like most animals and humans. That system can recognize armed persons and persons who wear clothes like camouflage that are making them suspicious. The system can also use infrared cameras and other tools that can see hidden weapons through clothes. The thing is that the AI is an effective tool for security. But it requires human support. The AI can mark persons who put things in their pockets in the shops. 

https://scitechdaily.com/when-ai-plays-judge-the-unintended-severity-of-machine-learning-models/