PALO ALTO—An underserved legal market. Inefficiency at law firms. Bias and amorality in artificial intelligence. These are some of the themes that were percolating Thursday at the Stanford CodeX center's FutureLaw conference, an annual event focusing on tech innovation in the legal industry.

The tenor of the event was one of both excitement and urgency, with the occasional dour warning that technology may actually end up putting some lawyers out of work. It was hard to escape the feeling that technological change is coming to the legal profession—and fast.

Here are some highlights from the daylong event:

1. “We can't all work at Big Law charging $800 an hour.”

That was the assessment of American Bar Association president Hilarie Bass, who is also co-president of Greenberg Traurig. She delivered the keynote address at FutureLaw. “The market for those legal services will remain,” she added, “but it's going to get more narrow than ever.”

Bass described a legal industry, education system and regulatory structure that is fundamentally mismatched with the market needs in the United States. Roughly 80 percent of Americans don't have access to legal services, she said, and that's in large part due to the fact that they're so expensive.

State bars require general legal education, even though the clients are demanding increasing specialization, Bass noted. “We don't ask our neurosurgeons to take a test on pediatrics. But that's how we test our lawyers.” And she observed that few laws schools are teaching students how to use AI or do legal project management, even though those are increasingly valued skills in the field.

Bass also warned that client-side pressure for fixed cost arrangements will force firms to adopt technology and become more efficient—or else lose to competitors who do.

2. AI has an “underspecification” problem.

“We have to remember that algorithms are good at what they're good at and incredibly bad at everything else,” said Sharad Goel, an assistant professor and executive director of the Stanford Computational Policy Lab. In a panel on fairness, accountability and transparency (FAT) in artificial intelligence, Goel highlighted how skewed datasets can lead to biased AI output, and the limitations of software in solving underlying policy problems.

Meanwhile, CodeX fellow Bryan Casey discussed the horrifying consequences that could ensue if self-driving cars are one day programmed only to reduce the liability of the manufacturer—a coldly logical end goal if machines are wired to internalize tort law and profit maximization. Casey warned that would lead to robot cars that accord less value to the lives of pedestrians in poorer areas and more to those in an “upper-crust” Silicon Valley neighborhood, for example, even though human drivers (hopefully) wouldn't act the same. (He also wrote a paper on this.)

It's a problem known as “underspecification,” or machines not being told what the real end goal is or how to get there, explained Google Brain fellow Been Kim. She also talked about how to shed more light on how AI models reach their conclusions, using a method called “testing with concept activation vectors (TCAV).” Short version: It's peeling back the layers of how a machine reached its decision, to be able to assess its biases and the validity of its results. That could become even more important with the new European Union data privacy law which establishes a “right to explanation” for AI.

3. Legal tech is proliferating.

Pieter Gunst, CodeX fellow and co-founder of Legal.io, gave an overview of how the legal tech industry has grown using metrics gathered through a project called the CodeX TechIndex. Among its findings was that document automation is the most crowded corner of the industry, with nearly 200 companies in that space, followed by legal marketplaces, practice management and analytics at a distant fourth.

While many in the industry have said that legal technology and AI will free up lawyers to do only the interesting part of their jobs, Larry Bridgesmith, the founder of a company called LegalAlignment, was more somber. “I'd love to say AI's not going to put any lawyers out of work. That'd be a lie,” he said. “If 60 to 90 percent of your day can be automated, what are you going to do when it is?”