AI defends defendant in court New Scientist

Soon an artificial intelligence system will tell the defendant exactly what to say in court. It is probably the first case ever where an AI takes care of the defense.

In February, an artificial intelligence system will advise a suspect in court for the first time in history. The AI ​​will run on a smartphone and listen to everything said in court. Based on this, the system will instruct the defendant via an earpiece what to say.

The location of the court and the name of the accused are kept secret by DoNotPay, the company that created the AI. It is clear that the suspect has been accused of speeding. The suspect will only say what the AI ​​transmits through the earpiece.

READ ALSO

Is there a diet that will make you live years longer?

The advice about healthy food is flying around you. But is it really possible to live longer by following a strict diet? If yes, what can y…

DoNotPay sees the case as a test. The company has pledged to pay any fine that may be imposed if the defense is unsuccessful, founder says Joshua Browner.

Hearing aid

In most countries it is illegal to use a smartphone or computer connected to an earpiece in court. However, DoNotPay has found a location where this system is classified as a hearing aid, so it is allowed, Browder said. “It’s technically within the rules, although I don’t think it’s in the spirit of the rules,” he says.

Browder recently had the AI ​​talk directly to a bank’s customer service team using an artificial voice. In this way, the system has reversed several imposed costs on its own. “It’s the most amazing thing I’ve ever done,” says Browder. “We only saved $16, but that’s a perfect job for an AI; who has enough time to waste to be on hold for 16 bucks?’

truth value

DoNotPay was launched in 2015 as a relatively simple chatbot that provided legal advice around consumer issues. The system relied heavily on pre-entered conversation templates. In 2020, the company started to focus more on AI. At the time, OpenAI released a programming interface that allowed anyone to leverage the capabilities of GPT-3, its language-processing AI.

According to Browder, it took a long time to train DoNotPay’s AI with numerous statements from previous cases. The AI ​​app now covers a wide range of topics, including immigration law. The company claims the system has impacted about 3 million cases in the United States and the United Kingdom.

Read also: AI can create almost any picture: fascinating and worrying

The AI ​​had to be trained to stick to statements of fact, rather than saying anything that could win a case, regardless of the truthfulness. “We try to minimize our legal liability,” says Browder. “The system is not meant to distort facts or be too manipulative.”

The audio tool has also been tweaked to not respond to every statement. “Sometimes silence is the best answer,” says Browder.

Legal

Browder’s goal is that the software will eventually replace some lawyers. “It’s all about language, and that’s what lawyers charge hundreds or thousands of dollars per hour for,” he says. “There will always be many good lawyers who plead before the European Court of Human Rights. But now a lot of lawyers just charge way too much money for copying and pasting documents. I think they will certainly be replaced, and that would be good too.’

Computer scientist Nikos Aletras from the University of Sheffield (UK) has developed an AI that can accurately predict the outcome of human rights litigation. He sees an increasing use of the AI ​​machine learning technique in the legal system, but warns that its application must be carefully considered.

“When your lawyer says ‘okay, let’s do A’, we trust that they have the expertise and knowledge to advise us,” says Aletras. “With AI it is very difficult to rely on predictions.”

According to Aletras, it is providing audio via audio real time legal advice in the courtroom is still a technological challenge. There are also ethical and legal issues involved, such as whether the use in court is legal at all.

ttn-15