AI may not steal your job, but it could change it

(This article is from The Technocrat, MIT Technology Review’s weekly tech policy newsletter about power, politics, and Silicon Valley. To get it in your inbox every Friday, register here.)

Advances in artificial intelligence tend to be followed by worries around jobs. This latest wave of AI models, like ChatGPT and OpenAI’s new GPT-4, is no different. First we had the launch of the systems. Now we see the automation predictions.

In a report released this week, Goldman Sachs predicted that advances in AI could somehow automate 300 million jobs, or about 18% of the workforce. world. OpenAI also recently published its own study with the University of Pennsylvania, which claimed that ChatGPT could affect over 80% of jobs in the United States.

The numbers sound scary, but the wording of these reports can be hopelessly vague. “Affect” can mean a whole range of things, and the specifics are murky.

People whose jobs deal with language might, unsurprisingly, be particularly affected by large language models like ChatGPT and GPT-4. Let’s take an example: lawyers. Over the past two weeks, I’ve spent time examining the legal industry and how it’s likely to be affected by new AI models, and what I’ve found is as much a source of optimism than concern.

The antiquated and slow-moving legal industry has been a candidate for technological disruption for some time. In an industry with labor shortages and dealing with tons of complex documents, technology that can quickly understand and summarize text could be extremely useful. So how should we think about the impact these AI models could have on the legal industry?

First, recent advances in AI are particularly well suited to legal work. GPT-4 recently passed the Universal Bar Examination, which is the standard test required to license lawyers. However, this does not mean that the AI ​​is ready to be a lawyer.

The model could have been trained on thousands of practical tests, which would make him an impressive candidate but not necessarily a great lawyer. (We don’t know much about GPT-4’s training data because OpenAI hasn’t released this information.)

Still, the system is very good at parsing text, which is of utmost importance for lawyers.

“Language is the coin in the realm of the legal industry and in the realm of law. Each path leads to a document. Either you have to read, consume, or produce a document…it’s really the currency that people trade in,” says Daniel Katz, a law professor at Chicago-Kent College of Law who took the GPT-4 exam.

Second, legal work involves many repetitive tasks that could be automated, such as researching applicable laws and cases and extracting relevant evidence, according to Katz.

One of the bar exam researchers, Pablo Arredondo, has been secretly working with OpenAI to use GPT-4 in its legal product, Casetext, since this fall. Casetext uses AI to perform “document review, legal research memos, deposition preparation, and contract analysis,” according to its website.

Arredondo says he has become increasingly enthusiastic about the potential of GPT-4 to help lawyers the more he uses it. He says the technology is “incredible” and “nuanced”.

AI in law is not a new trend, however. It has already been used to review contracts and predict legal outcomes, and researchers have recently explored how AI could help pass laws. Recently, consumer rights firm DoNotPay considered arguing a case in court using an AI-written argument, known as a “robot lawyer,” delivered through an earpiece. (DoNotPay failed the stunt and is being sued for exercising the law without a license.)

Despite these examples, these types of technologies are still not widely adopted in law firms. Could that change with these new big language models?

Third, lawyers are used to revising and editing work.

Large language models are far from perfect and their output should be tightly controlled, which is cumbersome. But lawyers are very used to reviewing documents produced by someone or something else. Many are trained in document review, which means using more AI, with a human in the loop, could be relatively easy and practical compared to adopting the technology in other industries. .

The big question is whether lawyers can be convinced to trust a system rather than a junior lawyer who has spent three years in law school.

Finally, there are limits and risks. GPT-4 sometimes composes very convincing but incorrect text, and it will overuse the source material. Once, Arrodondo says, GPT-4 made him doubt the facts of a case he had worked on himself. ” I told him, You are wrong. I argued this case. And the AI ​​said, You can sit there and brag about the cases you’ve worked on, Pablo, but I’m right and here’s the proof. And then it gave a URL to nothing. Arredondo adds: “It’s a bit sociopathic.”

Katz says it’s critical that humans stay informed when using AI systems and stresses lawyers’ professional obligation to be specific: “You shouldn’t just take the results from these systems, don’t examine them, then give them to people.”

Others are even more skeptical. “It’s not a tool I would trust to ensure that important legal analyzes have been updated and appropriate,” says Ben Winters, who leads the Electronic Privacy Information Center’s AI and human rights. Winters characterizes the culture of generative AI in the legal field as “overconfident and irresponsible”. It has also been well documented that AI is plagued by racial and gender bias.

There are also the high-level, long-term considerations. If lawyers have less practice in legal research, what does this mean for expertise and oversight in the field?

But we are far from that – for now.

This week, my colleague and Tech Review editor, David Rotman, wrote an article analyzing the impact of the new era of AI on the economy, particularly on jobs and productivity.

“The optimistic view: it will prove to be a powerful tool for many workers, enhancing their abilities and expertise, while boosting the overall economy. The Pessimist: Companies will simply use it to destroy what once looked like automation-proof jobs, high-paying jobs that require creative skills and logical thinking; a few high-tech companies and tech elites will get even richer, but that won’t do much for overall economic growth.

What I read this week

Some bigwigs, including Elon Musk, Gary Marcus, Andrew Yang, Steve Wozniak and more than 1,500 others, signed a letter sponsored by the Future of Life Institute that called for a moratorium on major AI projects. Many AI experts agree with the proposition, but the reasoning (avoiding AI armageddon) has been the subject of much criticism.

The New York Times has announced that he would not pay for Twitter verification. It’s yet another blow to Elon Musk’s plan to monetize Twitter by charging for blue ticks.

On March 31, Italian regulators temporarily banned ChatGPT over privacy concerns. Specifically, regulators are investigating whether the way OpenAI trained the model with user data violated GDPR.

I’ve been drawn to longer cultural stories lately. Here’s a sampling of my recent favorites:

• My colleague Tanya Basu wrote a great story about people sleeping together, platonically, in VR. It’s part of a new era of virtual social behavior that she calls “comfortable but scary.”
• In the New York Times, Steven Johnson published a charming, if haunting, profile of Thomas Midgley Jr., who created two of the most climate-damaging inventions in history.
• And Wired’s Jason Kehe spent months interviewing the most popular sci-fi author you’ve probably never heard of in this sharp, deep look into the mind of Brandon Sanderson.

What I learned this week

The “news snack” – browsing headlines or teasers online – seems like a pretty poor way to learn about current events and political news. A peer-reviewed study by researchers from the University of Amsterdam and Macromedia University of Applied Sciences in Germany found that “users who ‘snack’ more than others gain little from their high levels of ‘exposure’ and that ‘snack’ results in ‘much less learning’ than more dedicated news consumption. This means that how people consume information is more important than how much information they see. The study builds on previous research showing that if the number of “encounters” people have with news each day increases, the time they spend on each encounter decreases. Turns out… it’s not great for a mature audience.

Tech