Why Artificial Intelligence Can't Compete With Humans, and Vice Versa

Jun 5, 2017 | 2463 Views

Lawyers today can be forgiven for thinking artificial intelligence (AI) is taking over their industry. From AI contract review squeezing the demand for outside counsel to governments calling for laws to regulate and limit how AI is deployed, there is much to fuel this apprehension. Not to mention that oftentimes, it seems that the capabilities of AI are limitless. While AI solutions such as ROSS market themselves as an "artificially intelligent attorney," others are popping up that are capable of predicting the outcome of trial cases with up to an 80 percent accuracy.

But what is commonly missing from the conversation over AI is the context of how it compares and contrasts to human intelligence. While overlapping in some regards, the two intelligences hold vastly different and unique strengths which, for the time being, cannot be matched by one another.

AI in legal most commonly encompasses machine learning, the ability for software to "learn" and understand a task or a data set, far better and more efficiently than any human practitioner can. It is why AI is being relied on more than attorneys to manually review contracts, or scour through databases to make connections in data.

But as anyone who went to law school will know, being an attorney encompasses far more advanced and complex work than just shifting through data. And it is this higher level work-from creating and executing legal strategies to interpreting the changing nuances of laws and regulations within certain contexts and situations-that machines are ill-equipped, if not outright incapable, to learn.

Writing in Legaltech News, Rees Morrison, partner at Altman Weil, explained that this is because AI's learning power is limited by its grounded, math-based computational intelligence. The algorithms that power AI's machine learning, he said, "do not develop concepts and abstract ideas or create anything new, but they can extraordinarily rapidly discern patterns in data that humans simply cannot possibly match."

And though at times it may seem that machines can read documents similar to the way humans can, Morrison adds that is not exactly the case. "Machine learning algorithms can handle what are called factors, such as practice group names, because essentially the algorithms turn those factors into numbers."

So while AI in e-discovery and contract review can find all documents in a data set that relate to a specific party, it will not be able to understand if those document ultimately assist in building a case for or against that party. But this is often a common misperception of what AI can achieve. Farid Vij, lead information governance specialist at ZL Technologies, noted that "a lot of the time we are looking for machine learning to tell us what is important, what do we need to look at."

Yet "in many, many scenarios in our space that is an almost a false premise," Vij adds. "What machine learning actually does a brilliant job of is [telling you] what is the stuff you should ignore, what is the junk that is just going to district you, and is actually taking away from the actual value of the content."

AI in legal therefore only acts as a type of filter, able to parse through data with unparalleled ability, but not to intellectually conceptualize or interpret what the data means for a specific problem, an area still exclusively reserved for the human mind.

"They say if you take all the super computers and tack them from here to the moon, you still can't replicate the human brain, and that's because there is a certain context that we as humans can interpret that a machine today cannot," Vij adds.

Of course, modern innovators are always looking to push the envelope, and it may be only a matter of time until that saying is proven wrong. But Vij believes that such a possibility is far off in the future. "I still think we are a ways away to where machine learning starts to take over some of the contextual human thinking we do." Should that day every come, however, that may be much more pressing considerations, than robots practicing law. Continue Reading>>

Source: Law.com