Podcast 93: AI and Machine Learning for Lawyers, with Noah Waisberg

On this week’s podcast, Sam and Aaron talk about how going paperless can mean different things to different people, and that’s just fine. Then, Sam talks with Noah Waisberg about Artificial Intelligence (AI) and machine learning for lawyers.

Noah Waisberg

Noah Waisberg is the founder of Kira Systems, proprietary machine learning technology to analyze contracts.  Prior to founding Kira Systems, Noah practiced at the law firm Weil, Gotshal & Manges in New York, where he focused on private equity, M&A, and securities. Noah is an expert on contract analysis, legal technology, and artificial intelligence; has spoken at conferences including SXSW Interactive, ILTACON, and ReInvent Law; and was named 2016 ILTA Innovative Thought Leader of the Year. Noah holds a J.D. from the NYU School of Law, an A.M. from Brown University, and a B.A. with honors from McGill University.

You can follow Noah on Twitter and LinkedIn.

Thanks to Ruby Receptionists and Xero for sponsoring this episode!

Listen and Subscribe

To listen to the podcast, just scroll up and hit the play button (or click the link to this post if you are reading this by email).

To make sure you don’t miss an episode of the Lawyerist Podcast, subscribe now in iTunes, Stitcher, or any other podcast player. Or find out about new episodes by subscribing to the Lawyerist Insider, our email newsletter. We will announce new episodes in the Insider, and you can listen to them right here on Lawyerist.

Transcript

Sam Glover: Hi I’m Sam Glover.

Aaron Street: And I’m Aaron Street and this is episode 93 of the Lawyerist podcast where we talk with Noah Weisberg about machine learning and AI as it relates to law practice.

Sam Glover: Today’s podcast is sponsored by Xero. Beautiful legal accounting, simplified. Find out more at xero.com. That’s X-E-R-O.com

Aaron Street: Today’s podcast is also sponsored by Ruby Receptionists. Ruby answers our phones at Lawyerist so we don’t have to worry about getting interrupted when we’re being productive and we think they are awesome. You can visit Ruby at callruby.com/lawyerist to get a risk-free trial with Ruby.

Sam Glover: So Aaron, today’s interview is with Noah Waisberg of Kira Systems and we’re going to talk about machine learning and AI. One of the things that is going to be immediately obvious is that taking advantage of, really, the amazing technology out there really depends on getting your documents in digital format. I am constantly amazed at the resistance that I find to going paperless when I’m out and meeting lawyers, the lawyers who aren’t sure that it’s a viable thing yet, which baffles me because I first went paperless over a decade ago. The bottom line is, it’s actually quite simple. You just need to buy a scanner and start scanning things and getting them into digital format and then you can start doing amazing stuff, like feeding them to a machine learning system that pulls all of the relevant provisions out of contracts. Or you can just do small amazing things like back them up so that if your office burns down you don’t lose them or encrypt them so that they aren’t vulnerable if you leave your computer at a coffee shop.

Aaron Street: It is absolutely true that in order to digitize a document, all you need to do is buy a scanner and scan stuff. That is not, though, the resistance lawyers have to converting to being paperless. Certainly there are, in my experience, a handful of lawyers who otherwise see the benefits of paperless, but just think it’s not good. I think 99% of lawyers who aren’t yet paperless, it is neither of those issues.

It’s not that they don’t realize that you just need a scanner and scan stuff. It’s that they don’t have the existing capacity to work through all their archives of file cabinets full of paper because they’re too busy to do it themselves and it would cost too much to outsource the weeks and weeks of scanning it would take to get all those into digital form. They aren’t trained in proper naming conventions of files and, therefore, fear that they won’t be able to find the exact document they need when they need it. Whereas, right now, they know exactly where it is in their paper folder. They haven’t yet thought through how they would need to adapt their workflows so that right now they know how to take the next step in a particular case, but if there’s a new workflow they need to develop, they don’t want to take the effort away from existing client work to think through how to do workflows.

I think those are the major resistances. And probably sinking trust of the cloud or whatever the backup option would be to make sure you don’t lose your files. I think that’s the hold up. It’s not that they don’t know that they could buy a scanner and scan stuff. It’s then what?

Sam Glover: Yeah, you’re probably right about some of that. I mean, I find that one of the hold ups, for example, is that lawyers who question everything in the world except this believe that going paperless means that you’re not allowed to have paper anymore. It, of course, doesn’t mean that. Of course you can use some paper. I think that same, sort of, going paperless must mean this thing and I disagree with what I believe going paperless means seems to be one of the major obstacles. You can reconfigure it however you need it to be. Print out all the paper you want. It’s fine.

I met a lawyer just last week. When he discovered that, was like, “Oh! Oh, I didn’t realize I could keep paper.” Well, of course you can. Maybe backing it up a little ways. Going paperless is one of those things that requires you to work on your business instead of just working in your business serving clients. Which it’s always easy to prioritize the client work that you have in front of you, but every once in a while, you do need to take a little bit of time to think about workflow.

Aaron Street: I think that’s absolutely right and, to be clear, I am essentially paperless and advocate going paperless. I just think the barriers to doing it for most actively practicing, existing, small firms are higher than just simply buy a scanner and scan stuff.

But you’re absolutely right that there are best practices for things like file structures and naming conventions, but there aren’t rules or paperless police that you have to adhere to. I think that’s legitimately one of the concerns is someone hears you make it sound so simple and they realize that they don’t want to do things exactly the way you do and therefore, who are the police that tell me how paperless should work?

It reminds me, my wife and I last night were talking about a dinner party and the conversation was something like, “Well, do vegetarians eat this?” And then we said, “Well, you know, there aren’t vegetarian rules. Each person decides which things they eat or don’t eat and you would have to ask them to know for sure.” Similarly, there aren’t vegetarian police and there aren’t paperless police. If you want to keep some files, and you should probably always keep originals of important stuff where original signatures matter. You get to decide what your rules are.

But for sure, this conversation we’re about to have with Noah will indicate that getting your stuff in digital form has a tremendous number of benefits around the security of your documents, around sharing with opposing counsel and clients, around making sure that you have the ability to future-proof yourself. Lots of benefits.

Sam Glover: Yeah, I guess I would close this conversation, maybe, with a quote from David Allen’s “Getting Things Done”, which is that you have to think about this stuff a little bit more than you are now, but not nearly as much as you think. Once you get your documents into digital format, you can do much more cool and amazing things with them and it increases the value of them.

So let’s talk to Noah and find out what some of those things are.

Noah Waisberg: Hi, I’m Noah Weisberg and I’m a co-founder and the CEO of Kira Systems. Prior to co-founding Kira, I was an associate at Weil Gotshal practicing in mergers and acquisitions in New York.

Sam Glover: So you’re a bit of a big law ambassador to our podcast.

Noah Waisberg: Yeah, I guess so.

Sam Glover: We like to shake it up and bring in big law folks every once in a while.

Noah Waisberg: It’s my group. What can I say? I’ll never get away from them.

Sam Glover: So the first time you and I talked, Kira Systems was called Diligence Engine.

Noah Waisberg: Yes.

Sam Glover: Why the name change?

Noah Waisberg: Well, when we started the company, it was based on my experience as a mergers and acquisitions lawyer. In that job I spent a lot of time reviewing contracts and then as I got more senior, supervising people reviewing contracts. I came to realize that there’s a lot of room for improvement in this process and that technology could maybe help.

So we started out heavily focused on mergers and acquisitions. We continue to get a lot of use around mergers and acquisitions work, but we came to realize that there were a lot of other people who reviewed contracts for things that weren’t to do with mergers and acquisitions. Diligence Engine was such a name that was connected with mergers and acquisitions, so it literally got to the point where we’re finishing half of our demos with people saying, “Wow, I could use this in something else.” It’s like, “Yep, Diligence engine, we’re not just diligence.” It was almost a company tagline. When your company tagline is like “We do lots of other stuff, too” maybe we should adjust the name to be something a little bit broader.

Sam Glover: Time to re-brand.

Noah Waisberg: We still do do a ton of mergers and acquisitions related work, but we do a lot of other stuff. For example, Deloitte, which uses our system in their audit and consulting and some other parts of their business. Not necessarily doing diligence, but doing things with extracting data or contracts. They’re a big customer of ours. They have more than 3,000 people using the software in the US and so it’s really not just Diligence Engine. Even though we do do a bunch of that work, too.

Sam Glover: The first time you showed it to me, I had a light bulb moment where I realized that I’m always getting reprint requests and I always have to check and see whether that author has an agreement, has assigned us the copyright to their work or merely a license. So do I have to go back to the author and get their permission before we share it or do I have the right to decide that on my own? When you showed it to me, I just dropped all of our writer contracts in there and I was able to immediately grab the IP assignment terms out of all of those contracts and it would make it a piece of cake. That’s not diligence, it’s just common, everyday utility actually.

Noah Waisberg: Yeah, I think there are a lot of different … there are maybe three different situations when people review contracts, like doing it for diligence, trying to find information, which might not even be diligence. What you did is arguable that first use case, where it’s just extracting a bunch of data out of contracts. Second use case would be trying to find out what’s market. You would look at a whole bunch of different podcast and publisher contracts just to see what is appropriate. Again, there you’re in that data extraction piece.

Sam Glover: Oh that’d be so cool because everybody’s like, “This is a standard term”. I could actually spend some time digging out as many contracts as I can find and find out, do I really think this is standard term.

Noah Waisberg: You could. In a large law context, that is the kind of thing, or maybe not even large law. If it’s something where it’s a high enough value context, that might be something you would decide to do.

A third time you would review contracts is in connection with negotiating where you’re thinking, “Is this a good deal or not?” Really, we are pretty focused on the first two of those use cases with helping people extract data out of a lot of contracts. We realized, like you’re saying, that there are just so many situations when people do that work and diligence was just really constraining us and constraining our thinking and making it hard to convey to customers what we did.

Sam Glover: I guess if I can abstract it a little bit, what it really is is a way to look at anywhere from a small to a huge body of documents, contracts, with sort of an overview. My wife is a lawyer for the educator’s union in Minnesota and they have contracts all over the state. Maybe it would be useful for them to at a glance see what kind of similar provisions look like across the state. What if there’s a big project where they’re trying to get the same type of provision into every contract around a new law or something? It sure would be nice to figure out what’s already in those contracts, find out if we are normalizing them. Right now, I imagine that’s a pretty time-intensive, manual process.

Noah Waisberg: Yeah, there were some hard surprises with starting up this company. It turned out it took a lot longer than expected to get the tech to work. It took us a while to figure out how to explain to lawyers why they should pay us money. Then we had this really nice surprise, which is that there are tons of people who spend tons of time reviewing contracts and it has nothing to do with the M&A or … Another thing we were thinking about when we started out was maybe people pulling data out of their contracts out of leases, but it’s like so much broader than that. That it really is something for transactional lawyers that they spend huge amounts of time doing.

Sam Glover: We’ve been talking about what it does, but not the way in which it does it, which gets us closer to the subject of our podcast today. When you were initially showing me Diligence Engine, you very casually were like, “Let’s see what’s in this giant … what all the choice of law provisions in this giant pile of contracts.” You just clicked a button and there they were. It was like such a boring thing. I’m used to things happening that way now. Then I think there was a pause of a beat or two or three and then I was like, “Whoa, wait a second. What you just did used to take an army of clerks weeks or months to do and then deliver me a manual document and if I wanted that document to have anything else in it, I had to send it right back down to that army of clerks to start over.”

It’s that teaching the computer how to read the document and pull out those provisions is actually the holy shit, awesome part of this that is so smooth it’s boring, but what’s going on under there? That’s machine learning, right? Am I using that term correctly?

Noah Waisberg: You are. Maybe we should step back for a second and just describe to people in case they have forgotten, because no doubt they read every single thing that you write, but just in case they don’t. A quick explanation is our software will read through contracts and find stuff that you tell it to in there. So, if you felt like finding the assignment clause or the choice of law clause or the arbitration section or even what arbitration rules the arbitration section points to, the system could automatically go through and find that information and dump it into a chart or into Excel or into an XML output that you could then pull out using an API. Basically the system will find stuff automatically in contracts and pull it out.

What you’re asking is, “How does it do that?” There’s two ways that … a bunch of ways, but two basic ways that you might get software to do this and the one that I think a lot of people would think of initially is just a rules-based approach. So if you think about searching Lexis or Westlaw or some other research database, you might use a terms-and-connectors type search. That’s pretty close to a rules-based search. If I was trying to find an assignment clause, I might say, “I’d like to find all sentences that have the word ‘assignment’ within five words of the word ‘agreement’.” That would theoretically show me all assignment clauses. That’s what you call a classic rules-based approach.

In recent years, people have come to think that rules-based approaches aren’t an especially accurate way to find information. That they may work if you know what you’re looking for. If you know that every assignment clause in your pile of documents has the word ‘agreement’ within five words of it, then you’re good. In fact, lots of the time you don’t necessarily know what you’re looking through. For us, our system needs to be able to work even if we haven’t seen the contracts in advance and even if they’re in the format of poor-quality scans.

What we did instead is we use what’s called machine learning. With machine learning, you take special algorithms. An algorithm is just a fancy name for a computer program. We have special algorithms that are really good a learning word-to-sentence paragraph-length text and we spent years at building out these algorithms. My partner in the business has a Ph.D in Computer Science, we have several other people with Ph.Ds in Computer Science on the team. They have spent years finely tuning these algorithms to work while learning word-to-sentence paragraph-length text.

What we do is give our system examples of what we’d like it to find looks like and from that, it learns what these clauses look like. I spent the first year, year and a half of our company’s existence, ten hours a day, six days a week just reading random contracts and saying, “This is an assignment clause. This is a change of control clause. This is a confidentiality clause. This is both assignment and change of control.” We fed those examples into the system and based on those examples our system was able to build up a model of what assignment, change of control, confidentiality, and a whole bunch of other clauses look like.

Sam Glover: Gotcha. You’re teaching the computer how to understand, basically?

Noah Waisberg: Mm-hmm (affirmative)

Sam Glover: That’s machine learning in a nutshell, right?

Noah Waisberg: Exactly. A really good machine learning example, if you’d like one more, is around translation programs. People remember translation programs from like ten years ago, maybe had one in your PDA, your Palm Pilot or whatever, and it was probably a rules-based system. If you think about translating from French, you would think it would be amenable to a rules-based system. You would say, “Chez equals chair and tabla equals table or ordinateur is computer.” And anytime you saw ordinateur, you would put down computer in the app. Those systems actually worked pretty poorly if you remember. I think I had one in my … I can’t even remember what device it was. It was like pre-Palm Pilot. It was quite mediocre.

People who have used Google Translate know that now translation systems are pretty okay. How Google Translate works is not at all a rules-based thing. Instead, what they did is they built algorithms that are good at learning and they gave them tons of text that was equivalent. They would do something like, “Take the transcripts of the European Parliament.” And they would say, “Here is the same phrase in English and Latvian and French and Italian.” And give just tons and tons and tons of examples. After heaps of these examples, the system would actually learn what these different sentences and words looked like in real life.

Sam Glover: Gotcha. So if you feed enough examples in, you can, in theory, train a system to learn just about anything.

Noah Waisberg: Machine learning is behind a lot of things in the world today. The reason that our inboxes are not filled with spam or that cars might drive themselves or that Netflix gives us pretty okay recommendations, that’s all machine learning. I think we’ll start to see lots more.

One of the interesting things about machine learning, though, is often, we just see the outputs. Like, what you were talking about where a system just shows you the governing law, but you don’t actually see the system learning governing law or spam doesn’t make it into your inbox, or that sentence on that Lithuanian webpage you’re checking out just happens to be translated, but it’s not necessarily clear to you how that occurred. So many of these wow moments today are powered by machine learning.

One of the things that I think is really cool about our system, not to be all hype here, but is that you can actually train it for new stuff. I could teach it to find a further assurances clause or a revenue recognition clause or clauses that are relevant to revenue recognition. To me, one of the really amazing things about that, I think it’s like one of the only times that you actually get to see machine learning in action. I know when people do, they often really smile at it.

Sam Glover: Very cool. So machine learning is one thing, what’s AI? Artificial Intelligence / AI gets thrown around a lot and I’m tempted to suggest that it is becoming quickly robbed of all meaning. But maybe not.

Noah Waisberg: Yeah. Maybe. We never used to use AI in our marketing stuff, and then eventually we almost had to just because that’s what people understood. The idea with artificial intelligence is it is some kind of machine intelligence ability to do something that’s not human driven. That would broadly be artificial intelligence.

Within artificial intelligence, there’s kind of a strong AI. That is like a machine that actually has consciousness. Or artificial general intelligence is …

Sam Glover: A brain.

Noah Waisberg: A brain. Like an ability …

Sam Glover: Like you and I can go do math one day, write a book the next day, learn physics another day. We can do all kinds of different things with the same brain.

Noah Waisberg: Exactly. Whereas weak AI is something where a system is able to do one task well in a way that approximates or maybe betters human performance. Our system is really good at finding information in contracts. Car driving systems are really good at driving cars. If you had our system try to drive a car, I would not get in that car.

Sam Glover: Gotcha. You wrote a pretty neat article about that where, at least where we don’t have strong AI yet. That’s the singularity, that’s we’ll all be out in space and uploaded to computers or we’ll be serving our robot overlords. We don’t have that yet.

Noah Waisberg: Happily for all akin and a world free of work or who knows.

Sam Glover: Watson is really good at Jeopardy and now at diagnosing medical patients, but Watson, the IBM “AI” is not built to do legal research necessarily, although the guys at Ross are working on that.

Noah Waisberg: Yeah, I think if you are hearing about AIs right now, they are weak AIs. They’re good at specific tasks. I think a lot of people … AI is an area that seems to generate a ton of hype. I think we can all imagine, there are tons of tasks in our everyday life that probably a computer could do and might even be able to do better than us. That’s a really appealing thought. The thing is is that the AIs that exist right now, and most of the good ones are machine learning systems, the things that you’re hearing about today are subsets of machine learning, but there are other things that fall within AI that are not machine learning. Like a rules-based expert system could be an artificial intelligence and there are neat things going on there, too.

Many of the kind of things that are getting attention today are machine learning-based. They’re very specific and good at certain tasks. Watson, the Jeopardy thing, super, super impressive. Does that mean it would be good at doing e-discovery? Probably not. Could IBM get it to be good at doing e-discovery? Maybe. They have a lot of money. Certainly with money and time, you can solve many problems. But there are specific systems that are really great at deciding whether documents are relevant or not relevant, privileged or not privileged.

Sam Glover: Gotcha.

Noah Waisberg: Ditto for many other tasks within law. Even though there are these things that seem like really impressive machine learning or AI systems, it doesn’t necessarily mean they’re that good at anything beyond the specific tasks that they’re really good at. That’s not a knock on them, it’s just a thing.

Sam Glover: Let me take two minutes so we can hear from our sponsors and when we come back, I want to talk a little bit more about the relationship between machine learning and AI and, more importantly, talk about what this means for lawyers and how lawyers should read news about the future of law and robot lawyers and all that kind of stuff. So we’ll be back in two minutes.

Aaron Street: Billable hours are the lifeblood of a successful law practice. Problem is you still have to bill those hours. Even if your law firm has an accountant, tracking hours, clients, rates, preparing invoices, and collecting on those invoices is time you never get paid for. Writing notes to yourself in court or on the road is inefficient and error prone.

Run your legal practice better with cloud accounting software and see why over 600,000 small businesses love Xero, including Lawyerist. Get a free trial at xero.com. That’s X-E-R-O.com. Beautiful accounting software.

Sam Glover: This podcast is supported by Ruby Receptionists. As a matter of fact, Ruby answers our phones at Lawyerist and my firm was a paying Ruby customer before that.

Here’s what I love about Ruby. When I’m in the middle of something, I hate to be interrupted, so when the phone rings, it annoys me and that often carries over into the conversation I have after I pick up the phone. Which is why I’m better off not answering my own phone. Instead Ruby answers the phone and if the person on the other end asks for me, a friendly, cheerful receptionist from Ruby calls me and asks if I want them to put the call through. It’s a buffer that gives me a minute to let go of my annoyance and be a better human being during the call.

If you want to be a better human being on the phone, give Ruby a try. Go to callruby.com/lawyerist to sign up and Ruby will waive the $95 setup fee. If you aren’t happy with Ruby for any reason, you can get your money back during your first three weeks.

I’m pretty sure you’ll stick around, but since there’s no risk, you might as well try.

Sam Glover: Okay, if you can sum up for me how should we think about the relationship between machine learning and AI? Machine learning, it sounds like, is a form of weak AI and it doesn’t necessarily have to be AI, it’s just a thing where you can teach computers how to do things.

Noah Waisberg: Yeah, I would think about AI as an overarching category and machine learning as a sub-category within AI. Like one form of AI, many forms of AI are machine learning-based.

Sam Glover: Gotcha.

Noah Waisberg: But there could be other forms of AI that are not machine learning-based. As far as the weak versus the strong AI, a strong AI system could be and almost certainly would be machine learning-based, but that doesn’t necessarily mean that, like today, we’re not there.

Sam Glover: What we have mostly today is a bunch of machine learning systems.

Noah Waisberg: We have a bunch of machine learning or other expert systems, which can be fine in certain areas that solve specific tasks well.

Sam Glover: Gotcha. Okay, so every day, basically, there’s an article that comes out about artificial intelligence in law. Maybe it’s an opinion …

Noah Waisberg: Robot lawyers are coming to get us.

Sam Glover: Today, I sarcastically observed on Twitter that the last sentence in every … Somewhere in the first paragraph of every story about AI and law, you have to either state that it’s coming for lawyers’ jobs or that it’s not going to take over lawyers’ jobs because that is the ultimate question, I guess. It’s the 42 for lawyers.

How should we be thinking about this news? What’s a more intelligent way to look at this news and to understand the development of AI and the intersection with law practice and law that’s a little bit better than the binary, “Are they taking our jobs or not?”

Noah Waisberg: A couple different things to think about. One is there’s a ton of hype around AI. There are some solutions that are making an impact today that are kind of AI solutions. To me, prominent examples, the e-discovery technology-assisted review or predictive coding systems. Those are, I think, an AI and I think those have made a pretty significant impact on large-scale litigation. I think that will, over time, flow down to all litigation discovery.

Secondary would be contracts where we have a system that we believe to be pretty strong in this area, but there are other people with other systems in this area and people are actually using them. We’ve got lots of, mostly big firms, but not all that use our system to actually make a real impact in the process. Then there are expert systems, too, which are … Things like [inaudible 00:28:15], where people have been able to build apps, including for access to justice purposes that can make an impact in those areas. There are other things, too, like quantitative legal prediction.

The first thing from that is there are specific areas that artificial intelligence or machine learning, in some cases, is being applied to and it is having an impact in those areas. As a lawyer, one of the things you should be thinking about is: Is there work that you’re doing that just seems high-volume, highly repetitive, actually needs to be done accurately because, in some cases, a computer may be better at doing a high-accuracy demanding task than a human is, especially where it’s high-volume. If there are attributes like that of the tasks that you are doing, you should think about the fact that they probably will get automated at some point and maybe not base your career around those things.

Sam Glover: It’s starting to sound like we should not be thinking about this in terms of, “Is it going to replace my job”, but more in terms of, “Hey, here’s a really cool tool that maybe I can use in my practice.”

Noah Waisberg: Yeah, to the extent your job is doing a really repetitive thing and that is your practice, maybe you should be thinking about … Maybe stop or maybe do way more of it, too. AI can be real opportunity.

Sam Glover: Right. It’s a chance to scale.

Noah Waisberg: Yeah. The example that I really like is refrigeration. In the early 1970s, an average refrigerator sold in the US. took about 2,000 kw hours a year of electricity to run. Today, it’s closer to like 500 kw hours a year of electricity to run and you’re average fridge is 20% bigger and costs 60% less. You would think, based on the fact that it takes about 25% of the electricity to run a refrigerator today as it did in the 70s, that we would use less electricity on refrigeration. But that is not so. In fact, apparently we use more electricity keeping stuff cool than we did back then.

That analogy can sort of flow through to people and AI. A user of our software, our clients tell us that they can complete contract reviews in 20-60% and sometimes even 90% less time and do it as well or better than they could without the software. That doesn’t necessarily mean that they’re doing 20-90% less work. It’s that they can review an individual contract in less time. That actually opens up possibilities to review more stuff.

Sam Glover: I read this fascinating piece by Stephen Wolfram where he pointed out that as we can automate things, we can do more complex things. The complexity of the things that we need to do increases. Those of us who think that automation means we need fewer lawyers may be totally off. We may need twice as many lawyers because the relationships that we are contracting for have gotten so much more complicated that we need more people to do it. Or that just that people are able to contract in more relationships or they’re able to do more law and so there need to be more lawyers delivering services. Although the nature of those services may look a little different.

Noah Waisberg: Practice in 20 years may be very different than it is now, but it does not necessarily follow that there will be no practice because computers can do some stuff that lawyers currently do.

Sam Glover: Yeah. There was another interesting piece from that where he points out, it used to be that in order to get anything done, you had to create a checklist or procedure and then have humans go do it. Think about how you would print something out before there was even a typewriter. To print something out meant writing it out in long hand, sending it down the street to the printers where they would carefully typeset it. You couldn’t print something. Now you just send it to your printer.

Noah Waisberg: And yet, life goes on.

Sam Glover: Right. Contracts are essentially a set of directions to humans. As you no longer need a set of directions to humans to print something out, maybe we can see a different way of doing contracts going forward, which is a different way, not the end of lawyers. I think.

Noah Waisberg: To me, the really interesting thing is what are some of the businesses that are going to be created by and services offerings that are going to be created by efficiency.

Sam Glover: Mm-hmm (affirmative)

Noah Waisberg: Right now, what we’re thinking about is, “What does it mean for my business or for clients of our business if they reviewed contracts in 20-60% less time?” How does that fit into their financial model? Over the medium term, I think the question’s really going to change to be, “What are new things that we can offer because of the fact that we can do this more efficiently?” I don’t know what the answers to those are, but I think there are some really exciting businesses and service offerings …

Sam Glover: Cool shit is what we can offer. We can offer cool shit.

Noah Waisberg: Yeah. Seriously. Think about how many under-served legal consumers there are right now. So many people who don’t get lawyers. It’s like, “Well, we can’t really afford to offer them services at the prices that they’d be willing to pay.” Okay, so now maybe there’s technology coming along that will make it affordable to provide services to people that otherwise, right now, would look at getting a lawyer and be like, “Uh-uh, not going to pay for that.” I think it’s really exciting. I don’t know what those businesses will be, what those offerings will be, but I think they’ll be really exciting and I think they apply across the spectrum. I think they apply whether we’re talking about individuals who definitely are under-served by lawyers today, but I think it goes all the way up to the largest clients, too, where there are large corporates that have all these legal needs that aren’t being met because it’s just not either packaged or priced in a way that works for them to get these problems solved.

Sam Glover: On the optimistic note that lawyers should see opportunity in news about AI and stop asking the question, “Are we going to lose our jobs or not?” I think that’s a great time to say thank you so much for giving us a basic grounding in the machine learning and AI and helping us think through all of this news that’s coming out every day that is either doom and gloom or the sky’s the limit. Thank you so much, Noah.

Noah Waisberg: Thank you. I enjoyed this. Great to be on.

Sam Glover: Make sure you catch next week’s episode of the Lawyerist podcast. Subscribe to the Lawyerist podcast in iTunes or in your favorite podcast app. You can listen to it at lawyerist.com/podcast. You can also subscribe to the Lawyerist Insider, our weekly newsletter. Just go to lawyerist.com and look down the sidebar or click on Newsletter up at the top.

We’ll remind you where to find the podcast whenever we release a new episode. Thanks for listening.

Podcast 93: AI and Machine Learning for Lawyers, with Noah Waisberg was originally published on Lawyerist.com.