[Ed Note: We were holding off on publishing this episode because we were waiting on the official announcement that Kara Peterson and Rich DiBona of Descrybe.ai were the Silver Anthem Award winners in Best Use of AI from the International Academy of Digital Arts & Sciences! Congratulations to Kara and Rich for the prestigious award! – GL]

In this episode of “The Geek in Review,” we welcome Kara Peterson and Richard DiBona, the dynamic married duo behind Descrybe.ai. Descrybe is an AI-powered legal tool aimed at democratizing access to legal information, particularly by providing summarized legal decisions that make case law more accessible to everyone. The conversation centers around how Descrybe was created, its mission, and the challenges and opportunities involved in developing a legal tech tool that truly serves its users.

Kara and Rich share the story of how Descrybe began as a personal project sparked by a challenging legal issue they faced themselves. Rich, a software engineer, started experimenting with AI summarization of judicial opinions and was struck by how well it worked. Over time, this experiment evolved into Descrybe—a platform that not only delivers quick case law summaries but also enhances accessibility through simplified language and multilingual support. They delve into the technical challenges of managing massive datasets of judicial opinions and ensuring that their platform remains accurate and reliable.

The development of Descrybe has been heavily influenced by user feedback, which Kara and Rich emphasize throughout the episode. They explain how advisory feedback has been instrumental in shaping key features, such as Spanish translations and simplified summaries, which make the platform more accessible to non-native English speakers and those with varying levels of reading proficiency. Their genuine commitment to evolving the product based on user needs is a core aspect of Descrybe’s mission to democratize legal information.

A particularly intriguing segment of the episode focuses on Rich’s perspective on the legal tech market and his skepticism regarding companies that label their offerings as “LLMs” (large language models) without meeting the threshold of a true LLM. Kara adds her insights on the broader mission of Descrybe: not only to innovate within legal tech but also to address access to justice as a public health issue. They both underscore the importance of responsible AI development, ensuring that new technology genuinely benefits society rather than exacerbating existing inequalities.

Towards the end of the episode, Kara and Rich discuss the future of AI in the legal industry and the role Descrybe aims to play. They express optimism that AI can be a powerful force for improving the legal system, especially if it is used to automate lower-level tasks and free up lawyers to focus on more impactful work. They also highlight the potential for AI to facilitate alternative dispute resolutions, diverting some cases from the courts altogether. Kara concludes on a hopeful note, emphasizing that if the right choices are made now, AI could be a driving force for positive systemic change within the legal field.

This conversation with Kara Peterson and Richard DiBona offers a candid, behind-the-scenes look at building Descrybe—a legal tech solution driven by a mission to improve access to justice. Their commitment to making complex legal information understandable for everyone shines through, making this episode a must-listen for anyone interested in the intersection of technology, law, and social good.

Listen on mobile platforms:  ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Apple Podcasts⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ |  ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Spotify⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ | ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YouTube⁠⁠⁠⁠⁠⁠⁠⁠⁠

Twitter: ⁠⁠⁠⁠⁠@gebauerm⁠⁠⁠⁠⁠, or ⁠⁠⁠⁠⁠@glambert
⁠⁠⁠⁠⁠Email: geekinreviewpodcast@gmail.com
Music: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Jerry David DeCicca⁠⁠⁠⁠⁠⁠⁠⁠

TRANSCRIPT

Marlene Gebauer (00:08)
Welcome to The Geek in Review, the podcast focused on innovative and creative ideas in the legal industry. I’m Marlene Gabauer.

Greg Lambert (00:15)
And I’m Greg Lambert. And Marlene and I just got back on Friday from Florida. And I wanted to talk about that, but instead of just the two of us, I’m actually going to bring in our guest, because I think this is something that is going to be interesting for them as well. So today we are joined by Kara Peterson and Richard DiBona.

Marlene Gebauer (00:28)
Yeah, bring in our guests.

Greg Lambert (00:40)
co-founders of describe.ai, a justice tech startup that aims to democratize access to law through AI-powered legal research tools. So Kara and Rich, welcome to the Geek in Review.

Marlene Gebauer (00:47)
It’s easy for you to say.

Kara Peterson (00:55)
Hey, thanks. We’re so excited to be here.

Greg Lambert (00:58)
All right. So, so last week, Marlene and I went on an advisory panel with about what, about five other profession. Yeah. And so it, it was really good. It was, it was in Fort Lauderdale, Florida. We were right on the beach. Although I have to say, even though I was 50 feet away from the beach, I never actually put my feet in the sand the entire time. And, and we were in a windowless room.

Richard DiBona (00:58)
Thank you.

Marlene Gebauer (01:07)
I think they were like all told there were 20 people there, but some, know, so probably about 10 clients.

Greg Lambert (01:28)
Most of the day so, you know before before the meetings after the meetings it was work So just just so you know that but We’ve done advisory panels before and this is why I wanted to bring Rich and in Cara on to talk about it One of the things I thought was really interesting this time around Marlene was that this one felt like we were being asked for our opinions on things instead of being

Marlene Gebauer (01:29)
Ha

It was work. It was work. Yes.

Greg Lambert (01:58)
kind of given a dog and pony show on products that the vendor was coming up with. And I felt like that was the best use of our time and of the vendor’s time. So Marlene, I’ll start with you first on this and get your opinion. What did you think of that?

Marlene Gebauer (02:16)
Well, I do like that and I think for this particular advisory board they really do focus on that. They really do want to hear what we have to say about how they’re doing things. there is a small amount of sort here’s where we’re moving with product. But it’s much more a strategic type of discussion as opposed to this is how it works and how do you like this and how do you like this.

this type of thing and that type of thing for each product. Because that’s really the benefit of this. mean, they could do that kind of demo anyplace and have different audiences and give that type of feedback. This is really more, they’re bringing in leaders from different firms to basically hear what they have to say about what they’re doing and how they could be doing things differently or better.

Greg Lambert (03:15)
Yeah, and Rich and Kara, I’d like to, this is why I brought you guys on. I mean, if you could sit down with eight of your customers, high-end customers, and talk to them, how would you, what do you think would be the most value that you could get out of that meeting?

Kara Peterson (03:37)
Yeah, that’s such a great question and I love that the folks down there flip the script kind of like that. It’s almost like flipping the classroom, right? Like that, you know, the movement that kind of happened in higher education. And I think they were really smart to do that because when you have people like yourselves or other advisors or users sort of in a space where you can pick their brain, it’s…

really, really fortunate and that’s like golden time and you shouldn’t waste it. So exactly like you’re saying doing the dog and pony show and like here’s what we have like that’s gonna put people kind of in a box about what they’re reacting to but if you give people more of a space of let’s kind of dream this up together I think that’s really valuable and that’s actually how we approach our advisory board, our advisors at Descrybe I’m happy to talk more about that if you’re curious.

Richard DiBona (04:25)
I, yeah. And that brought up a thought like early on. I remember a few months ago, we had a call with an advisor and I’ve gotten better about this, but we had, calls with people and they would, they would be like, I don’t like the way you do this. Or I don’t like the way you do this, or maybe you should do this. And at first I would be like, you know, like kind of like you’re making fun of my baby or whatever. Kara helped me.

Greg Lambert (04:50)
Yeah.

Kara Peterson (04:51)
You

Richard DiBona (04:53)
actually realize like no these calls are so important and this feedback is so important and you know it’s good if somebody’s blunt with you because if you go out there with some feature and you have some experts saying maybe this isn’t the most useful feature you shouldn’t just disregard it so i i think it’s a super important thing it’s not even a nice to have

Greg Lambert (05:14)
Yeah, yeah, and I thought, you know, the questions that they kind of asked us were open-ended questions that they didn’t necessarily know how we were going to answer them. And some of them were nice and some of them were like, hey, you really could, you know, improve in this area. And it was really interesting because there was one specific area that I think everyone thought

we could improve on and the person that was over that piece of it was sitting right there in the room. But I think handled it very well. It wasn’t meant to be insulting. It was meant to be, hey, we’re here because we want you to succeed as well. And so this is honest feedback where we think we would love to see some improvement here.

Marlene Gebauer (05:59)
trying to help.

Kara Peterson (06:01)
Yeah. Yeah.

Richard DiBona (06:01)
Yeah.

Greg Lambert (06:11)
because it helps us back in our office, you know, kind of get people to buy into your product that we’re spending a lot of money on.

Kara Peterson (06:20)
Right, like leave the ego at the door, you know, and if it’s a gift that you can hear that feedback immediately or before you go to market with somebody because the market will give you that feedback. And it’s better to have it ahead of time. My only criticism of this team, and I don’t know who they are, so sorry if I’m offending anyone, but you should have let your advisors go take a swim in the afternoon. That’s the only thing, you know? Okay.

Marlene Gebauer (06:32)
Right.

Richard DiBona (06:44)
haha

Marlene Gebauer (06:45)
We did get some networking time to be fair.

Greg Lambert (06:47)
Yeah, we did and good food and we actually got to see Gary Oldman, the actor at the end of a bar on our way out from our dinner. So, and I dragged Marlene back into the restaurant and made her verify that is Gary Oldman, right?

Kara Peterson (06:50)
Okay.

very cool.

Marlene Gebauer (06:57)
Yeah, we embarrass ourselves highly. It’s good.

to verify.

Kara Peterson (07:05)
Okay, what makes up for it?

Richard DiBona (07:06)
That’s so funny.

Marlene Gebauer (07:06)
I will point out one other thing about this advisory board. It was also nice because we did do some design workshops in terms of brainstorming. I think they were very effective because I think a number of ideas came out of that that they hadn’t considered at all or they were going down a different route and were like, no, do it this way. And they’re like, this is great.

Kara Peterson (07:29)
yeah.

Marlene Gebauer (07:36)
They were so, so happy and grateful about that type of feedback because it’s going to have them change direction.

Kara Peterson (07:43)
That’s great.

Richard DiBona (07:43)
And it’s important, like Cara and I are not in the legal industry. And so I think part of the reason we’ve done well is because our naivety about it and building something that we think would be good, but that naivety, like you don’t know every everything. And so there’s people who are experts and you have to listen to them. So I think it’s, super important.

Greg Lambert (08:05)
Yeah.

Kara Peterson (08:05)
Right, yeah, we approached it like, so we kind of approached it in two ways, like as an engineering problem and just sort of a communications problem, because those are both our areas. So when he says we’re not in the legal industry, he means we’re not attorneys. So we jacked our advisory group up with the most underrated and the most important professionals in the field, legal librarians, law librarians from higher education, like they are worth their weight in gold.

So.

Greg Lambert (08:35)
Yes, they are. And I did want to do one more thing on the design sprint, because I’ve done design sprints with others where the outcome was almost predetermined on how they set up the sprint. And I thought this one was much more reliant upon the individuals in the groups that were set up to come out with the outcomes. So there was very broad inputs that were given.

Kara Peterson (08:41)
Mm.

Mm-hmm. Yep.

Greg Lambert (09:04)
for instructions, but the outputs were kind of left over to or left to decide from the people that were actually doing the design sprint. again, that’s

Marlene Gebauer (09:13)
And then the entire group had the opportunity to weigh in on each of these. And then everybody got to do their stickers in terms of what’s critical.

Kara Peterson (09:14)
cool.

Greg Lambert (09:23)
Yeah, yeah, worked out really well. So, all right, well, enough talk on other people and events. Let’s get to our guests and let them talk about their products. So, Cara and Richard, you both come, like you said, not necessarily from a legal background. So, do you mind describing, describe AI and kind of walk us through the experiences and, you know,

Kara Peterson (09:24)
Nice.

Greg Lambert (09:52)
marketing, public health, and software engineering that is your background to help you develop this AI-powered legal tool. whoever wants to kick it off.

Kara Peterson (10:02)
Sure, so I’m happy to start and people who may not know us, we are actually a married founders. So we are lucky in that we share a common sort of ethos or like a way to think about the world and kind of what we would want to contribute back to it. So I have a lot of experience in communications largely in higher education, but.

a space I spent a lot of time was in public health and it forever changed me sort of learning about public health in terms of looking at systematic issues and in sort of what levers there are to change those such things as disparities in health outcomes or racism or whatever, things like that, gun violence. So I see access to justice as really part of the public health continuum in a lot of ways, like the out.

the outcome of our sort of unfair and unjust legal system that nobody likes, right? Nobody in the law thinks that we’ve got it set up the right way. I think that there’s just a whole lot of overlap with sort of public health areas. So that’s, think, where it got really interesting to me. Plus, from a more sort of just business side perspective, the inefficiencies are so astronomical and insane. When you start to look at it from an outside perspective, I think…

People in the law might be a little bit of boiled frogs and just kind of used to it. But I think a fresh perspective is able to find some really obvious low-hanging fruit, especially with AI. So that’s my perspective.

Greg Lambert (11:40)
Rich, what in your background helped you with this?

Richard DiBona (11:44)
Yeah, I think I was a software engineer at like traditional large companies and dealt a lot with data architecture and scaling large systems and a lot of just traditional backgrounds. And then what happened was we had our own legal issue, like a, you know, nothing. We didn’t rob a 7-Eleven or anything. was, but, but you realize how

quickly it is to access the law, but you don’t know how hard it is until you need to get at it. So if something happens to you and it’s not criminal, but civil, it’s not like you automatically get a lawyer or anything. So I had a hard time researching this. And so what I did was when OpenAI about two years ago now came out with their APIs as a software engineer,

started playing around and I had found some judicial opinions that were, that I had been reading and that I found on my own and I decided to have it summarize them just for fun. And I was amazed at how well it did. And so I started summarizing more and more and we ended up where we are now. So.

Marlene Gebauer (13:10)
interesting.

Richard DiBona (13:11)
Yeah.

Marlene Gebauer (13:12)
So Richard, I have a spicy, a little spicy question for you. So you have expressed skepticism about some recent projects claiming to be LLMs. So can you elaborate a bit more on the key differences between these high functioning LLMs and smaller projects in the tech space, the legal tech space? And how does that impact the development of tools like Descrybe AI?

Richard DiBona (13:16)
Sure.

funny. Did you find that on LinkedIn or something? Did I make a comment somewhere? I think. Yeah, there you go.

Marlene Gebauer (13:45)
We do do some research.

Kara Peterson (13:46)
very good. Sneaky.

Greg Lambert (13:47)
Yeah, I think this was from our email conversation a few weeks ago.

Richard DiBona (13:53)
yeah, yeah, yeah, yeah, yeah, yeah, yeah. Okay. So I think what happened there was there was this company that was saying they created a new LLM for law and, whatever companies I kind of look at whatever’s out there. Like if Bob Ambrosio talks about something new or you see it on LinkedIn somewhere, I’ll go and look. And to me, it looked like they had a legal Q and a chop out, which is great.

Kara Peterson (13:55)
He’s already forgotten.

Richard DiBona (14:20)
but I didn’t understand like what the threshold was for them to be calling themselves an LLM. Like to me, an LLM is like your open AI or Anthropic or Cloud or some other, some other thing that is, you know, brings in billions of tokens and makes that transformer and all that. So I just didn’t know what the threshold was for them to call themselves an LLM. So that, yeah.

Marlene Gebauer (14:50)
I think it’s…

Greg Lambert (14:51)
I think you find that in the market, especially whenever there’s something new, whether it’s extractive AI or generative AI. Even maybe, now you’re seeing things like agents in this. It seems like you can kind of make your own definition of what you are for a little while.

Marlene Gebauer (15:12)
I was going to say there’s a lot of discussion in the community about definitions, about what something specifically means. And I know, like, I put out a cheat sheet about what, you know, different types of rag. And I know other people have put out, you know, here are definitions because, you know, I think, you know, maybe the vendors, you know, sometimes, you know, keep that sort of loose in terms of what the definitions are and then.

The rest of us are kind of like, okay, well, what exactly does that mean? And, you know, is this the same thing or not? So, I, you know, I commend you for basically calling it out and questioning it because, we all want to get it right. So, you know, I think it’s important.

Richard DiBona (15:55)
Yeah. Yeah. And it seems like the, it’s changing so rapidly. Like I saw an interview with Sam Altman from just last week and he was talking about like a SI, like super intelligence. So now we have to think about this. And so it’s like, like, like it just changes so fast.

Marlene Gebauer (16:10)
get that’s another one, okay. Thanks, Sam.

Kara Peterson (16:12)
Bye.

Well, and it’s like the wrappers, right? I mean, we’ve all spent a lot of time talking about wrappers. Like, and what does that mean? Is it thin wrapper? It’s a thick wrapper? Is it a wrapper? You know, and that’s, it’s a similar thing. And so we can’t really be too surprised when the marketplace is really confused and it’s a lot to put on the market, right? To try to make these decisions properly for their firms or for their own use when there’s so much confusion.

Greg Lambert (16:17)
Yeah, it’s…

Kara Peterson (16:42)
and also in such high-stake cases as legal. And don’t get me started on this, because I’ll take the whole hour, but the idea of how are we assessing how well these things work, that’s like an entire other deep rabbit hole that needs to be explored someday.

Greg Lambert (16:59)
Yeah, and you have to be careful because a lot of times you get, especially in the legal industry, AI’s given a little bit of a honeymoon period on this, but a lot of times you get one bite at the apple. And if you’re not what you’re advertised that you are, it’s hard to get that second bite once you’ve kind of disappointed.

the audience and you don’t necessarily want your marketing. Yeah, I think a lot of times you’d like to under promise and over deliver than the other way around.

Marlene Gebauer (17:32)
and

Richard DiBona (17:32)
just like the new restaurant down the street exactly with the bite. Like if they, if they’re not good, you may not go back.

Kara Peterson (17:32)
with that.

Yeah.

Marlene Gebauer (17:39)
Exactly.

Kara Peterson (17:39)
Well, there’s a conversation, I don’t remember which show it was, but there was a lot of discussion about this. Somebody came out and said they were the first XYZ, and I don’t remember what the XYZ was in particular, but this panel really had a field day with that because it’s like, were you really the first? And it’s really dangerous when people say that, and as journalists, we immediately are questioning. So yeah, I think you’re right. You gotta be careful what you’re telling the market about who you are if you can’t back it up.

Marlene Gebauer (18:06)
And to go back to your, know, are these things really as good as they say they are? mean, again, lot of discussion about standards or accuracy and like some vendors are kind of doing their own thing and saying, okay, here’s our accuracy, you know, rate or, you know, others are like, well, we, you know, we go back to citations and that’s how you determine it. And that, and yet there’s even more conversation about, about like third party groups trying to set some sort of standards. you know, everybody’s sort of struggling with that to try and just get some purchase.

Kara Peterson (18:12)
Right.

Mm-hmm.

Marlene Gebauer (18:36)
on what accuracy is enough that we’re comfortable with it. I don’t know the answer to that.

Kara Peterson (18:45)
I don’t think anyone does yet. Yeah.

Greg Lambert (18:47)
It kind of reminds me, every Mexican restaurant that you go to here in Texas is the very first one that ever served a frozen margarita. I have seen that claim at so many Tex-Mex places.

Marlene Gebauer (18:53)
You

Kara Peterson (18:57)
Of course.

Marlene Gebauer (19:01)
That’s true. Home of the original frozen margarita.

Kara Peterson (19:03)
Wow, I believe it. Right. It’s like, wait, I just had that yesterday somewhere else.

Greg Lambert (19:11)
Well, let’s talk about the mission that you have there at describe.ai. So I know that there’s a big push towards democratizing access to legal information. And Kara, recently you’ve included upgrades that feature things like Spanish search and simplified summaries.

Can you talk to us a little bit about what the mission is there and what were some challenges that you faced when trying to implement, putting these things going from mission to actual reality?

Kara Peterson (19:51)
Yeah, that’s great. as things evolve, this is happening so fast. And even when Rich said two years ago, that’s 1,000 years ago, right? In AI time, in terms of how long we’ve been around. But we still feel like a pretty new company, even though in AI years we’re old. We’re long in the tooth. So we, opposed to or opposite of what we were just talking about, are

from the beginning was we’re going to do one thing and we’re going to do it really well. So we’re not trying to be everything to everyone. What we’re trying to do is play at the end of the continuum of access to justice in the access to law space. So when you think about sort of the whole ecosystem.

Companies like ours, where we’re really dealing with access to law or access to publicly available information in a way that makes it more accessible to people, either through the language or the summarization or just the fact that it’s free, there’s no login, those kinds of things. All the way on the other end of the continuum are people who are working in…

probably even a much more interesting space, but also much riskier in the actual delivering of legal services, right? Using AI and direct to consumer kind of work. we believe that the changing access to justice in this country will be something that takes all of us, and it’s going to take a continuum. So we’re excited to be playing in this foundational part. We’re standing on the shoulders of others, like the free law project, that’s where we get our data. And then we hope that over time other people will stand on our shoulders and we have an incredible trust

of data that we’re going to be doing things with and perhaps we’ll find partners that can do things with the data as well. But right from the beginning, it was exactly that. We are going to make case law available easily and in ways using natural language for people who don’t have access to it right now. And that’s where we spend our time. Lots of challenges, but maybe I’ll let Rich talk about challenges because there’s some from the tech side and just some from the…

universe side but Rich you want to talk tech?

Richard DiBona (21:53)
Yeah, mean, unless there’s a follow-up question to your… Let’s talk tech. Yeah, I think the challenge for us is what we decided to do was pre-summarize everything and get everything searchable instantly so that when you go into our thing, it’s not actually calling some AI search, which helps with the hallucinations and all that too, but…

Greg Lambert (21:56)
Nope, let’s talk tech.

Kara Peterson (21:58)
Hehehe.

Richard DiBona (22:22)
The scale of that was just, and is so massive, like dealing with millions of traditional opinions that are each like 40 pages long or, you know, there was one in Pennsylvania that the first time I ran it, a year, whatever it was, a year ago, whenever I was running Pennsylvania the first time, this thing was 750 pages long, this traditional opinion. I’m like, my God.

Kara Peterson (22:48)
Someone had to write that, that poor person.

Richard DiBona (22:49)
Yeah, it’s like somebody wrote war and peace as a judicial opinion. so it’s just massive amounts of data. So what we’ve ended up with like 100 million summaries of these things when you include the translations and the simplified. But we think it’s worth it because we’re able to get more accurate searching and just really fast results.

Marlene Gebauer (22:52)
Hahaha

Kara Peterson (22:53)
Right.

Greg Lambert (23:16)
Well, describe, Rich, you don’t mind, describe what your typical customer, who they are and why they use describe.

Richard DiBona (23:29)
Sure, I’ll describe with a why it. Yeah. So at first I would say that Kara and I, or maybe just I, but I think Kara too, we expected this to be like a B2C thing where people would go in and type in like my dog bit, my neighbor’s kid, you know, or.

Kara Peterson (23:32)
There you go.

Richard DiBona (23:53)
I slipped and fell outside of a store or my landlord evicted me for no reason or whatever, base, not basic, but consumer level searches that happened to the public every day. And what we’ve found is that it seems to be mostly used by attorneys and we don’t track everything, but just from people talking to us. it’s a good place to start for attorneys. If a client walks into your.

office into your law firm and has an issue and you just quickly want to find out some relevant case law. It’s the fastest and cheapest way to do so in natural language. So maybe somebody walks in and they say, I own a store and my employee accidentally knocked over a customer and they got hurt. Who’s liable or you know, you may not know that off the top of your head, but

If you just type that exact sentence into our described AI tool, it’ll give you relevant case law in less than a second. So you could go and move on with your research. And so it’s not the end, it’s legal information, but what it does is it gets you started towards further research where maybe you take those cases and you shepardize them or do whatever and see if you could actually use them.

Also with the summaries that are kind of like cliff notes of these opinions, you could read through them in seconds to see if it’s a relevant case rather than having to read the whole 40 pages and try to stay awake and see if it’s relevant.

Greg Lambert (25:37)
I think that leads into, we’d asked if you guys were willing to do a little demo on this. So Rich, would you mind just kind of walking us through kind of a typical use case?

Richard DiBona (25:48)
sure. Yeah. Let me me hit share share screen.

Kara Peterson (25:53)
While he’s doing it, I’ll just quickly say back to challenges. We actually rewrote the entire week. Rich actually rewrote the entire platform using the latest models. And that’s the other challenge when you’re in this space is things are improving so fast. Yeah, we rewrote it from the ground up, which he’ll explain.

Marlene Gebauer (26:07)
Keep it up.

Richard DiBona (26:14)
Yeah, Carl, do you want to come up with a search here or should I just type in something?

Kara Peterson (26:20)
Yeah, just whatever, one of our usual ones is fine.

Richard DiBona (26:24)
Okay, here’s…

Kara Peterson (26:26)
And remember to explain what you’re doing.

Richard DiBona (26:28)
Yeah, I am, yeah, so basically all anybody has to do is go to our website and there’s a search box here just like Google and then the rest of the website if you want to read about us and figure out what we’re doing. And then you could go to the search tab up here and you could basically pick your state to search in and you could type in legal terms, concepts or case facts. I…

I just put in is Miranda necessary if out of custody. So I’ll hit search on that and then it comes up and you see this is doing all of our stuff. So like this one’s Massachusetts, this one is New Hampshire, this one is North Carolina and then you could narrow it if you want and then this is you get matching scores so this

Marlene Gebauer (27:05)
And it’s one jurisdiction at a time.

Richard DiBona (27:24)
the AI, this first one, it gave a match score of 96. So the AI feels that this excerpt from this Commonwealth versus Smith in Massachusetts Supreme Court is probably a good match for you. And the first sentence is, and this is a summary again, for people who are audio only, Miranda warnings are mandated only during custodial interrogation. So.

you notice it didn’t, it’s not like a keyword search where you have to type in the exact words to get this. It knows the concept that you’re trying to do. And then what we added here is there’s four buttons, one for Spanish and all the results translate instantly to Spanish. And I won’t try to read one of those. And then simplified English for people who have lower reading proficiency.

or even school-age kids for Supreme Court cases or whatever, can, this is the case at a fifth grade reading level, which is pretty amazing. If I do say so myself. Yeah, I tend to read these a lot too. It’s actually pretty good. And then sometimes it’s funny, like I do a lot of testing every time I’m adding new features and I’ll type in like murder weapon was thrown in the dumpster or something.

Greg Lambert (28:33)
This is what I would pick for me.

Kara Peterson (28:35)
Yeah, right.

Richard DiBona (28:50)
Like to hear the simplified English to describe crime scenes is kind of something you could do on your own, but it’s pretty wild. And then we have simplified Spanish also, and you can see how quick these are. So this is the exact same summaries, and then I’ll just go back to English for a bit.

Greg Lambert (29:10)
Yeah, and I noticed that also that it was, for example, in the English version versus the simplified English, that the layout was also a little bit different, that even though it was talking about the four factors, the four factors in the more complicated one were all kind of.

Kara Peterson (29:27)
Mm-hmm.

Greg Lambert (29:33)
pushed together as one paragraph, whereas this one lists out so it’s a little bit easier to read as well. I don’t know if that just happens to be this one example, but…

Richard DiBona (29:43)
Yeah, so down here, it seems to break it up into paragraphs more, but yeah, here’s another one with bullet points. Yeah, so I think it just makes it overall easier to digest shorter sentences, easier words. Yeah, so it’s good about that. And then I could type in, I mean, I could click through, I’m clicking the first one, Commonwealth versus Smith.

And then you get the whole opinion. This is the summary of it. So even the summary is long. So then below you could see the original opinion. And this is the long thing that lawyers have to deal with currently or had to before our tool came along. You’d have to read this whole thing to see if there was anything relevant. And I think that’s why a lot of times lawyers tend to use the same

cases over and over to site because first of all, they’re good precedent. But second of all, it might be harder to surface other ones that might be good for your case. And then the same feature we have on here, if I click Spanish, it instantly translates the whole opinion, simplified English. So again, we have the translations here. And then one last thing before we…

It’s a quick demo. I’ll go back to here. I did a Spanish language. My landlord evicted me. you can’t see this, right? Can you see? Okay. So what I did here was I’ll do a new search. I made a Spanish language translation. This says my landlord evicted me for no reason. And so I’ll go back here. And so we could search in Spanish also. And we get back.

Kara Peterson (31:19)
we can’t see.

in

Richard DiBona (31:39)
results that so and we got them in English but we could translate them again to Spanish simplified English and Spanish so you could search and read in Spanish now which will help a lot of people.

Kara Peterson (31:57)
Can I say one thing just back to the top of the conversation we were talking about the advisory board and so you know feedback and all that and it wasn’t from an advisor but this

idea to translate into Spanish and particularly into simplified level for readers came directly from someone who’s very prominent in the legal aid world and was suggesting that you know something like this would be an incredible help. It didn’t really exist right now so we went away from that meeting saying well we could do this let’s do this so it came directly from consumer needs so that feedback loop is so critical.

Richard DiBona (32:38)
Yeah, they actually said, we like the summaries that you have, but they might be a little bit too complex. And we were like, wow. And because I originally didn’t want to, for some people, yeah. And so I didn’t, like, you don’t want to do things where you think you’re being condescending to anyone or anything. So they made the suggestion that maybe a fifth grade level might be appropriate.

Kara Peterson (32:48)
for some clients. Yeah.

Marlene Gebauer (33:06)
Are there plans to add other languages? Sorry, I’m losing my voice here a little bit. Are there plans to include other languages at all in the future?

Kara Peterson (33:15)
I mean, it’s certainly an option and something we’re really interested in. And again, we would talk to the community. We’re not sure, you know, if we went for like, I don’t know, Mandarin or something like that. We’re not exactly sure how the AI would approach it. know, languages like English, French, Spanish, know, Portuguese, whatever might be a little more fluid. But I think it’s a really, really interesting.

Marlene Gebauer (33:36)
easier.

Kara Peterson (33:41)
question and we start to think about like what are the most spoken languages in the United States where we don’t have an official language, right? Spanish is the obvious one, but then from there it’s a really interesting question.

Marlene Gebauer (33:53)
Yeah, mean the reason I was thinking… go ahead.

Richard DiBona (33:54)
We could certainly do it. I mean, it’s super expensive to do it. And so that’s part of it too. Like all these things, every time you’re every like this new language, this was four translations of 3 million opinions. So broken into yet. So our open AI loves us. We’re like tier five, but we think they, love our money, but they haven’t done anything for us. So, yeah.

Kara Peterson (34:00)
Mm.

You’re welcome, Sam. be nice. If they’re listening, call me. No, but yeah, but I do think like, I think that’s the excitement about this kind of technology. You know what, where, the access can become so meaningful, right? It’s pretty cool. And there’s no reason that this tool is only usable in the United States. Like people use it all over the world.

Yeah, so and it can be applied to any jurisdiction, and it can be applied to any complicated document. You know what I mean? It’s not just for legal. could be for…

Richard DiBona (34:49)
If I mean, presumably it could not out of the box. Like we’d have to do work to apply it to any.

Kara Peterson (34:55)
No, no, I know. See, this is the difference between a marketer and an engineer. The engineer is like, whoa. And the marketer is like, we can translate everything in the universe. But somewhere between the two is the truth.

Richard DiBona (34:59)
Yeah.

Marlene Gebauer (34:59)
Share can be double, wait a minute.

Richard DiBona (35:05)
You

Greg Lambert (35:05)
This is when the marketer says that you’ve created your own legal LLM.

Kara Peterson (35:09)
That’s right. Exactly. And this is usually when Rich will tell his favorite joke about what it’s like to be married to your co-founder.

Richard DiBona (35:09)
Yeah, there you go. That’s pretty funny. That’s funny.

yeah, mean, well, it’s not really a joke. It’s actually true. think marketers, I think to have a 24 seven IT person is pretty amazing for a marketer. And I remember like when the Celtics were up in Boston and when they were going to the finals and they had months, two months of playoffs and Carl would be like, we need this on the website. And I’ll be like, I’m watching the Celtics playoffs. You know, it’s at like nine at 10 at night.

Marlene Gebauer (35:20)
Yeah, let’s hear the joke.

Kara Peterson (35:21)
Yeah.

Tough. You know they’re going to win anyway.

Richard DiBona (35:49)
So I think that’s, yeah, but we’ve figured out how to balance all that. It’s actually good.

Kara Peterson (35:56)
Yeah, so the joke was that it’s every marketer’s dream to have a 24-hour IT person that can’t say no. And that’s pretty much true.

Richard DiBona (36:04)
That’s true, I can’t say no, yeah.

Greg Lambert (36:04)
That is true.

Marlene Gebauer (36:06)
That’s fair. That’s fair. Well, I think this kind of moves into a conversation, into a question that I have for Kara. And you’re involved in various AI and legal tech organizations. So what trends or developments in terms of responsible AI or access to justice do you see as most crucial for the legal industry to address?

Kara Peterson (36:12)
Yeah.

That’s great. So there’s so many issues around bias and even in the foundational models. And we even talk about this ourselves. And I was talking to someone about it recently who’s a pretty well-known person talking about bias and AI. But it was a private conversation, so I won’t share the name. how do you deal with this if the things you’re building suppose

tools you’re trying to make ethically on top of data sets that are considered in some ways unethically sourced. So it’s a very complicated sort of dance you have to do with yourself. Like, OK, but if we only had people who didn’t care if the data was not ethical building things on top of it, then we don’t end up with people building unethical things. So it becomes very tricky and slippery slope and confusing. But I think the most important thing is to keep talking about it.

because we don’t really know, but we have to keep questioning ourselves and we have to keep questioning what we’re building and we have to keep questioning if we’re building it the right way and if we’re making the right decisions. Because again, we have to think about this now because it’s happening so fast. And one of the things I really worry about, and this is not just in legal tech, but it’s certainly going to be super important in legal tech, but it’s for all of us and all AI.

areas is that if we follow our traditional models of how we build and decide what’s valuable and important, we’re going to end up with the same problems we already have. They’re just going to be even worse. And what I mean by that is if we only fund white dudes who look like they can turn into unicorns, you know, we’re going to be building tools that serve white dudes who turn into unicorns, right? And so we have to find other models in this society to support.

companies like ours and despite other groups in the Justice Tech Association and things like that who are trying to use AI to solve big societal problems, maybe the potential for personal income growth isn’t as high as just saying, we’re going to do whatever we can to make as much money. So we need to make a lot of noise. And I think legal tech is doing a pretty good job about that, to be honest, compared to some other sectors. I think there’s a lot of women in particular in legal tech who are making a lot of noise about this and, you know.

we need to keep speaking up.

Marlene Gebauer (38:45)
Yeah, I mean, I think you raise a really good point that making the noise, because we live in a capitalist country and everybody’s chasing the dollar. And this might not be an area where there’s necessarily all the dollars, but if we make enough noise, it continues to be developed.

Kara Peterson (39:07)
Right, and we can’t be surprised then if we don’t, right? And if then, gee, we just made all of the divides worse. Because if you imagine like a dystopian future, right, where things, we don’t correct for that, basically AI is just going to make the already well-served better-served and be able to get one over, so to speak, on everyone else even more easily. So it’s not gonna naturally.

Market forces don’t work that way. We wish they did. We always say, self-regulation, it’ll be fine. And we know that doesn’t work. So there has to be purposeful attention paid to spreading the benefit across. It’s hard, though. It’s really hard.

Marlene Gebauer (39:47)
It is hard. This kind of leads me into the next question, and I think I heard dystopian, so now I’m a little worried. It’s like, we’ve got an uplift here. So what change or challenges do each of you see on the horizon over the next couple years in the legal industry that the legal industry is going to need to face? And Richard, I will start with you.

Kara Peterson (39:53)
Sorry. I’m in a mood lately. What can I say? Yeah. All right.

Greg Lambert (39:59)
Yeah.

Richard DiBona (39:59)
Yeah.

Yeah, that’s interesting. I was thinking about this in terms of using AI to help me program. So I was formally trained in software engineering, computer science, and learned how to debug and learned how to do all these algorithm developments and write all this code. And I found that you can ask the AI to write code for you.

but it’s not always precise and right. Even though the AI says like this code, here’s the code you’re looking for. You could ask it to help you. And you have to know that it’s not right in order to know it’s not right. Like if you just use the code that it gives you, you could end up in a bigger hole than where you started. Even though it does do a great job. And I was actually thinking about this and I mentioned this to Kara on one of our dog walks in the last week or two, like,

I imagine it’s similar for lawyers where if you rely totally on AI, it might get you into a issue where it like does something wrong and talks of it confidence, like confidently tells you something that’s wrong. But if you, so you kind of have to monitor what it’s doing and then it’s powerful. Like if, as long as you don’t get lazy and just take its output and just submit it, I think it’s great.

but I think you do have to monitor all of its outputs and proofread it as if maybe a lower level associate gave it to you before you just go and submit it.

Greg Lambert (41:56)
Yeah, I’m one who has used it to help create some simple scripts and things like that. And again, you’re right. I think one, if you keep it simple, it does okay. And usually with scripts, if you run it, unless you’re doing something dangerous, typically if there’s an error, it will bounce back and then you can kind of debug that. It’d be really cool if you could do that in legal, if you could just submit it into…

a system that would check it first and then I guess in a way maybe shepardizing is one of those steps. But as far as bias and some of the maybe even incorrect answers that you get back, that’s definitely something that you’re going to spend a lot of time legally debugging the answers that you get as well. I think that’s a…

Richard DiBona (42:50)
Like if you, yeah, mean, along those points, if you have it write a brief for you, you can’t just submit that brief, right? You have to proofread it as if it was, you know, like you’re a professor or something. Like I think it could lead to danger there, but it is super powerful in how far it could get you along the road, but you just can’t blindly trust it.

Marlene Gebauer (43:12)
Yeah, Richard, I mean, your point sort of raises the…

People who know the content can use it effectively because they can check. But what happens to people who don’t know the answer? And how are they going to check it and use it and be effective? I think it’s something we’re wrestling with now in terms of different people in the profession. Like how should we use this appropriately?

Richard DiBona (43:21)
Yeah.

Kara Peterson (43:21)
Okay.

Richard DiBona (43:26)
Yeah.

That’s a great point, Marlene. Yeah.

Marlene Gebauer (43:47)
and continue to have people to learn. I don’t think anybody knows at this point.

Kara Peterson (43:54)
Yeah.

Richard DiBona (43:55)
No, but I mean that, that, that it’s a good point that people like good lawyers and good programmers and all that their jobs aren’t necessarily going to go away because you do need this human in the loop for anything it does pretty much. Except for art. can correct it’s art. Those pictures.

Kara Peterson (44:14)
No. No. Yeah.

Marlene Gebauer (44:14)
Yeah, that’s still pretty bad. Kara, how about you?

Kara Peterson (44:20)
So this is where I get really excited because I think the future is possibly, if we make good decisions now, like super, super, super bright. in all of my experience, and I did have a stint working at Suffolk University Law School as a marketing director there, so I do have some tangential like relations with the law. That sounds strange, but you know what I mean. And.

All the time that I’ve been talking to people in the legal profession, nobody has said to me, boy, things really work great. We’ve set up such a good system that’s so efficient. So people are ready to fix this. so AI opens a possibility to just change so many things. So some of it is those, you know.

Maybe we think about the things that are around people in the law that are maybe not that high level functioning stuff that’s needed. So just the amount of time you’re spending on sort of rote activities that AI could really help with, that would free up lawyers to do more lawyering, right? And so you do more high level cognitive thinking and to work with their clients more and to have more time and to frankly have a better life and to have a better work-life balance and all these things. So the possibilities are endless in terms of the efficiencies it could

it can make and if you look at the current system right like what are people now doing that don’t have like they don’t have anything anyway that’s good in terms of getting wrong information or knowing how to process certain things so pro se people

you know, it’s not their fault. They’re doing the best they can, but it’s causing all kinds of, you know, gunk in the system and slowing things down. And it’s just, so I look at it as the area of efficiencies that apply like, applied to a systemic way across, across the whole system. can just change the nature of way things happen in the legal profession, which would be good for everyone. Cause there aren’t enough top level or even just any lawyers to serve the need, right? We know that. So there’s going to have to be some parts.

between automation and human to actually solve the need. And I don’t know if this will come to pass or not, but one of my biggest hopes actually…

And things that I get really excited about is how AI can divert a lot of things from the courts in the first place. So we already know a lot of like it’s a very small number that gets in front of judge in the first place. But think about the alternate ways we could have to settle disputes. know, I think AI in the alternate dispute resolution sort of space is going to be fascinating to keep an eye on and to watch and to see where there may be different ways to deal with these problems that actually help create a more efficient court sort of

legal system overall. That’s very, you know, Pollyanna, but I look at this in a very Pollyanna way. I think if we make good decisions now, we can create that future.

Greg Lambert (47:03)
I would say that the best outcome for a legal issue is to settle it before it needs to get to court. All right, well, Kara Peterson and Richard DeBona, I want to thank you both for coming in. You’re the co-founders of Descrybe, and I want to make sure that’s described with a Y, .ai. So thank you very much for coming in and talking to us.

Kara Peterson (47:13)
yeah. Yeah. Yeah.

Marlene Gebauer (47:13)
Yeah, true.

Kara Peterson (47:27)
That’s right.

Thank you so much. That was a lot of fun.

Richard DiBona (47:31)
Thank you so much. This was great.

Marlene Gebauer (47:35)
And of course, thanks to all of you, our listeners, for taking the time to listen to the Geek in Review podcast. If you enjoy the show, share it with a colleague. We’d love to hear from you, so reach out to us on LinkedIn.

Greg Lambert (47:46)
and Kara and Rich. We’ll put links to in the show notes, but for the listeners, what’s the best way to find out more about Descrybe or to reach out to you guys directly?

Kara Peterson (47:59)
Yeah, come to our website. It’s DESCRYBE.com. So please, or .ai, sorry. We do own .com too, sorry. But come visit us and we’re on LinkedIn. We’re very active on LinkedIn. So do find us there.

Richard DiBona (48:05)
We do own.com too though.

Marlene Gebauer (48:08)
Ha ha ha.

And as always, the music you hear is from Jerry David DeCicca Thank you, Jerry.

Greg Lambert (48:20)
Thanks Jerry. Talk to you later Marlene.

Marlene Gebauer (48:22)
Okay, bye.

Kara Peterson (48:23)
Bye, thank you.