In this episode of The Geek in Review, hosts Marlene Gebauer and Greg Lambert sit down with Otto von Zastrow, the founder and CEO of MidPage.AI, an AI-native legal research platform. With a recent $4 million seed round and an ambitious mission to rival legacy research tools, MidPage is drawing attention across the legal industry. Otto shares his unconventional journey from AI-powered lawn robotics to transforming how litigators interact with case law. His pivot into legal tech was fueled by a combination of technical curiosity, the rise of language models, and firsthand insight from his lawyer friends overwhelmed by inefficient research workflows.
Otto walks listeners through the core of MidPage’s offering, which includes the usual suspects—case law, statutes, regulations—but with a twist: smarter search tools, intuitive UI, and features like a proprietary citator and their newly launched Proposition Search. This feature aims to solve the long-standing “needle-in-a-haystack” problem by surfacing judicial language that matches precise arguments, accompanied by contextual metadata and filters. Otto highlights that the goal isn’t just to match or mimic tools like Lexis or Westlaw, but to rethink what legal research should feel like when modern AI capabilities are built in from the ground up.
One of the more unique aspects of MidPage’s product development is their internal “kangaroo court”—a monthly teamwide challenge where employees, regardless of role, must conduct legal research using MidPage or traditional tools. Otto notes that this process not only improves product design but builds real empathy for the user experience. Engineers and designers are encouraged to think like litigators, helping identify pain points and close functionality gaps. As a result, the product continually evolves based on firsthand user scenarios, not just speculation.
The episode also delves into the data-side challenges that have historically prevented innovation in legal research. Otto explains why now—thanks to improved AI models and open access to data—is a rare inflection point for startups. He emphasizes the strategic importance of MidPage building its own case law dataset to avoid being beholden to incumbents. This independence allows them to innovate more freely, enhance precision, and lay the groundwork for broader API access that could empower the next generation of legal tech tools.
Finally, the conversation looks ahead. Otto predicts that AI will amplify the capabilities of individual lawyers, enabling them to process more data at greater depth. In a world where clients are increasingly self-educating with tools like ChatGPT, MidPage aims to provide lawyers with the means to maintain credibility and efficiency while ensuring accuracy. As AI models grow more capable and agentic, Otto sees an evolution not just in how legal research is conducted, but in how lawyers interact with knowledge, data, and ultimately their clients.
Listen on mobile platforms: Apple Podcasts | Spotify | YouTube
[Special Thanks to Legal Technology Hub for their sponsoring this episode.]
Blue Sky: @geeklawblog.com @marlgeb
Email: geekinreviewpodcast@gmail.com
Music: Jerry David DeCicca
Transcript
Greg Lambert (00:00)
Hi, I’m Greg Lambert with the Geek in Review and I’m here with Nikki Shaver from Legal Technology Hub and you’re going to talk to us about transaction management, so tell me more.
Nikki Shaver (00:10)
Yes, I mean, some people might hear that and not think it’s an exciting topic. To me, it is. This is actually one of the hot areas of legal technology at the moment. Transaction management platforms have been around for a while, but they’re really gaining traction. And in fact, Legal Tech Hub has just added another premium category on our site for transaction management. So what is transaction management or what is the technology?
These products are solutions that facilitate and streamline the processes involved in a corporate transaction, allowing them to run more efficiently and effectively. The platforms support complex deal work, but in fact, they can also templatize transactional matters, which means there are really great benefits as well for simpler deals, especially when they are run at high volume. Some of the processes that can be run in a transaction management platform include issues lists, conditions, precedent checklists,
warranty disclosure exercises, real estate diligence processes, and steps plans. The technology available often supports deals all the way from the opening of a matter through to closing and even post-closing activities and obligations tracking. Some of the available solutions only address one aspect of a transaction like closings, a real pain point, and others are end-to-end. And of those end-to-end, some platforms are modular,
which allows firms to start small and grow into the platform. Perhaps most critically, many of these solutions are collaborative, which means that clients are able to communicate with their outside counsel through them and self-serve information about where a deal is up to, their WHIP on a matter, and see clearly what the next deliverable is and who’s responsible for it. If your firm or legal department is interested in streamlining deal practices, check out the premium category on Legal Tech Hub, which provides you
with a full overview of the area, including deep dives on the providers, checklists that you can use to evaluate products, and a business case that will help you explain to leadership why this kind of investment is worthwhile. You can check it out on LegalTechnologyHub.com.
Greg Lambert (02:16)
who knew there was so much involved in transactional management. So thanks for bringing awareness to that.
Nikki Shaver (02:22)
Thanks, Greg.
Marlene Gebauer (02:30)
Welcome to The Geek in Review, the podcast focused on innovative and creative ideas in the legal industry. I’m Marlene Gabauer.
Greg Lambert (02:36)
And I’m Greg Lambert and this week we are diving into the rapidly evolving world of AI powered legal research with Otto von Zastrow, founder and CEO at MidPage AI.
Otto (02:49)
Thanks for inviting me, nice to be here.
Marlene Gebauer (02:51)
So Midpage is an AI native legal research platform that’s making waves by aiming to provide litigators with tools to conduct better legal research faster. recently announced a significant $4 million seed funding round and launched an intriguing new feature called proposition search. So I guess we’re going to talk about that.
Greg Lambert (03:13)
And just for a little background, Otto brings in quite a unique background to the legal tech space with experience in AI robotics and a master’s thesis from ETH Zurich focusing on applied AI language models to legal documents. So Otto, I think this will be interesting.
Marlene Gebauer (03:35)
Yes, definitely. Otto, your path to founding mid page AI is quite interesting. you started a company called Navstack AI, developing AI powered lawn robots. And then your academic work at ETH Zurich focused on using AI to help lawyers find relevant cases more efficiently.
Otto (03:35)
Thank you. ⁓
Marlene Gebauer (03:58)
you know, what was the spark or, or, or the key insight that led you to make that pivot from AI and robotics to tackling the complexities of legal research? that, that, seems like, it seems like a hard left turn somewhere.
Greg Lambert (04:13)
Doesn’t look, yeah, it doesn’t look like a straight path.
Otto (04:16)
you
⁓ It always feels different on the inside of course. I had a bunch of friends at law school or starting their first jobs and so I was acutely aware of the things that they get confronted with every day. But when I was doing robotics before, I was really focusing on the deep learning part. At the time we were using these low-cost AI models to see the surrounding world. And so for our robots it would be understanding where am I allowed to drive? Is this a good surface to drive on?
And the research side there was evolving extremely quickly and at the time I also noticed that not only is consumer robotics very difficult field to bring to market but also the language models were evolving at a much more rapid pace than the vision models. And so
and now much of the language model research has moved over to vision models as well. So was really interesting for me to work on language because that was the more interesting architecture from a technical standpoint. But then it also sort of just made me try to find out which fields are to be most impacted by language models being really good at reading and searching text. And because I already had lots of contacts to law everywhere, I thought this would be
probably the best area to apply this to.
Greg Lambert (05:26)
Yeah, we’ve had a few guests that talk about the language models affecting two markets specifically, and that’s law and healthcare. Did you have any desire to go into the healthcare aspect of it?
Otto (05:41)
I feel, I I didn’t it myself and I feel there are some others too. I mean obviously as a programmer that was actually the first contact one. around two and a years ago every programmer suddenly got this say 20 % boost in productivity because suddenly from one day to the other there was this tool called GitHub Co-Pilot by Microsoft and OpenAI where every programmer suddenly in their text editor had this
really clever autocomplete. And that was actually what we sort of something like what we started with too. So we said we’ve got to transfer this amazing experience for programmers into what lawyers do because they come from a field where there’s fewer technologists around so somebody’s got to give them the push.
Greg Lambert (06:19)
Well, MidPage brand itself as providing ⁓ everything you expect in a legal research platform and more. ⁓ So for listeners who might be new to MidPage, do you mind just kind of stepping back and give us a big overview of the core functionalities and describe some things like the research agent or the citator?
Marlene Gebauer (06:30)
And more.
Otto (06:45)
Yeah, of course. So like with any legal research platform, we have all case law and we have statutes and regulations. And of course, every litigator, but also many other kinds of lawyers, would use that as a daily driver to just do research and find precedents. And of course, as we all know, for most of the last 100 years, this has been exclusively Lexis and Westlaw.
And there were a couple of reasons for that, mainly because those were the only ones that had the data. But now that it’s possible to create datasets without being dependent on those two, and we’re not the only example, but I think it’s going to become much more accessible over the next couple of years. We might be sort of one of the furthest along for fully utilizing this new data availability. And so we can actually have a really different search experience that allows you to do some searches in just a couple of minutes that would have taken you maybe an hour before with
the older methodology. And so right now we’re being used by a couple hundred firms from small law of big law but really sort of focusing on making it easy to use and accessible. And the users that use us today, some of them use it as a replacement and others sort of just as a compliment to the platforms they’re familiar with. when we say sort of everything and more, what this really means is…
People have learned to use Lexis at law school and then 30 years later they’re still using it, right? So there’s a whole list of things that people are expecting and need from a research platform. So even if it’s better at one thing, it doesn’t mean that you can replace your old ways of doing things until you have all those other items on the checklist too. And this would, for example, be a good citator where you can see was this case overruled? What are all the other opinions that cite this case? What did the later court say about this?
What is the subsequent case history? And all of those things are now part of the MidPage platform.
Marlene Gebauer (08:27)
So one feature that caught our eye is the grid-based search, which, and we’ve seen some praise for this, its intuitive way of organizing and comparing case details. What was the thinking behind this design, and how does it help litigators, especially when they’re juggling multiple cases or complex legal research issues?
Otto (08:51)
Yeah, great question. So when you do a search today, you get this list of links and then you have to take it from there as a human. Maybe you try a couple different searches, but really what you’re doing is you’re opening a hundred different tabs just to start reading every single case that appears relevant. And what we do instead is basically take that as a starting point and allow you to really narrow down these potentially hundreds or thousands of search results into something that is
is
much more transparent. So you can say, you can turn this whole classical search into a spreadsheet. every column can be a question that you care about.
So you have your classical search query and then you just add all the things that you want to find and then it will break down the like that. So maybe you can ask the outcome should be a certain type of motion that you want to be granted. You can ask something about the fact pattern. It should include some clause or some type of car accident. And with all of this, you’re taking your initial search and really narrowing it down so that the AI shows you how does each of these results fall.
all your questions. And that, I feel, is a way to give more control back to the lawyers without them having to, you know, open up a hundred tabs and read every case.
Marlene Gebauer (09:57)
it’s interesting, like when you talked about the spreadsheet. So would you say this works in sort of a similar fashion to some of the tools that have been more in the transactional space where it’s like, OK, we’re highlighting certain things and then we put it into a report that you can review more easily?
Otto (10:14)
Absolutely. I think it’s a common new UI pattern that is visible not only in transactional law, but also in fields that have nothing to do with law. It’s like having a Google search, but instead of just this one-dimensional list, you now have a multidimensional list. It just happens to be a very useful way to display data and to let the users refine it over time.
Greg Lambert (10:35)
Well, we had a couple of guests on a couple of weeks ago from Definely that talked a lot about the user experience and the user interface. And I know that that’s one of the things example on results. I’m curious to learn more about the philosophy that you worked with in creating the user interface.
you leverage attorneys to kind of help you design that and then once you have it built, how do you get your feedback to make sure that what you’re designing is actually accomplishing what you’re hoping for?
Otto (11:14)
Great question, yeah. one of the…
unexpected things was that we hired a lawyer that used to work in Big Law as a litigator and he turned out to be our best product manager because he was the target user. He was using, you know, CaseText and Lexis and Westlaw before he also, you know, had a computer science background so he kind of came from both worlds and so he had a really strong sense of ⁓ what the ideal research experience should be like and that was one part, so having him really guide the whole
process. And the second part was we have this internal kangaroo court which I think is a fun way… ⁓
to teach everyone how to do law. So in every engineer and every, regardless of what the employee does, once a month we give them a legal research task where they have to use either classical tools or a mid-page to do research like they were a litigator. And that really helps you see and understand all the little problems that litigators have while reading cases. I think there were several features that probably got a lot better just because the engineers building the features and
to the designers and everything, they just knew what it meant to do research.
Marlene Gebauer (12:20)
Yeah, that’s interesting.
Greg Lambert (12:21)
The kangaroo court,
I like that.
Marlene Gebauer (12:24)
It’s an interesting concept because you’re giving access to people who do legal research who maybe aren’t as familiar with just the law and how that’s structured and what they should be looking for. And I could see where that would highlight some things where it’s like, we could have a better user experience that way. Because you really don’t know the level of understanding that you might be getting for a user. It could be somebody who’s very new to it.
someone who’s not an attorney, but it could be someone who’s very sophisticated. it’s, I would imagine it’s hard to satisfy that range of people.
Otto (13:02)
Absolutely,
Greg Lambert (13:04)
I’m just wondering, do any of them sneak off to the local law library and do their research, they do it all on
Marlene Gebauer (13:09)
And
come back.
Otto (13:12)
We certainly had the discussion telling whether people are allowed to use mid-page to answer the kangaroo court tasks because it would have been so much easier. then sometimes we forced them to do it the old way so that they know, you what are these features for? So yeah, that was quite funny.
Marlene Gebauer (13:27)
So Otto, I want to talk about the proposition search. you know, you’ve mentioned that this was inspired by CaseText’s parallel search, but it aims to pick up where parallel search left off by addressing, some of its perceived shortcomings. So what in your opinion are those shortcomings that proposition search allows litigators and researchers to do that they couldn’t easily do before?
Otto (13:53)
Yeah, great question. I would say there’s probably three parts. One is that the technology got better. Now three years later, the AI that is powering the search engine under the hood is just a lot stronger than it used to be. We have a really strong engineering team, of an engineering first company in a we’re proud of the search engine that is powering this, just means the results are going to be even better ranked. The second part is…
the interface where I really like this spartan approach because now when you do a proposition search you can you know let’s say you have a you know that there’s a certain argument that courts probably made many times over the years and you want to find the strongest you know
I don’t know, a couple of citations that say exactly what you want them to say. And so that’s when you use proposition search. And with our interface, you kind of get this very clean side-by-side view of here’s exactly the phrase that the judge said or wrote. And next, you get this summary of the procedural history of the case, what is the case about, roughly how does it compare to your facts on your search. And this is just the most condensed view that you could have to sift through these results.
And that’s just enabled by taking this great search and moving it into our spreadsheet-like view. And then the last part is all the other features that we have on the page that really allow you to filter down. So if you can now say, please only show me cases where
this was a non-compete, a certain condition about the non-compete should be true, maybe it has to be the duration of the non-compete was over five years or something. You can now sort of take all the search results and filter them down to only those that natural language filters. And this just gives you level of control and precision that you didn’t have before. So it’s few things. It’s a better search engine, it’s a great interface, and it’s sort of these powerful filtering options.
Marlene Gebauer (15:42)
I want to follow up on that. I’m a litigator, right? I’m not a tech person. And I want to understand how MidPage is going to help me better prepare my arguments and prepare for cases. What do you envision is the difference between what they’re doing now and what MidPage can offer?
Otto (16:07)
So I would say most of the time during research is really just spent being overwhelmed by this huge mass of information and at the same time being really afraid of overlooking some very important precedent. And because we have a better search and a really, you know, a search that makes it really easy to boil the ocean and to know what did I see before, you what is new, how does this compare to each other. It’s just much…
easier to get a clear, organized view of all the relevant case law for your research questions. And once you have done a couple of searches and you’ve found all these good cases, we also help try and sort of organize that by giving you some tools where you can chat with all the opinions that you’ve already found because they’re all part of your same notebook, your research repository.
And that just gives you a good way to, you know, after several days of research, your brain starts spinning because you’ve seen so many things. It helps you stay on top of everything and keep everything structured.
Greg Lambert (17:01)
speaking of the data set, you guys made a pretty strategic decision there at MidPage to building your own case law data set. And I I saw somewhere where the ambition is to be the largest case law data set ever. So I was just curious on, you mentioned that this would help you stay more independent. And I’m assuming that
that means against the two big vendors that we all know. you also this week, you you kind of back that up with the $4 million seed round, which was led and I think the official word by an unnamed legal publishing house. So, so Otto, without naming this, ⁓ this publishing house, you know, what, what kind of strategic importance do you have in having a legal publisher as the, the, you know, one of the lead investors at the stage?
Otto (17:53)
I mean it’s always great work together with people who really understand their market you’re in. We have VCs from previous rounds as well and they’ve been really great at knowing how to hire great engineers and how to do marketing and so would say…
It’s been such a fun experience to meet everybody in the legal tech industry because many of the people have been around for two, three decades and the same names start reappearing everywhere. And so it becomes much clearer to understand which strategy makes sense and which one doesn’t make sense. And yeah, you should make sure to align yourselves with the right players so that they can help you grow.
I can’t say too much more.
Greg Lambert (18:35)
Yeah, it’s a big market, it’s kind of the run into the same people over and over again.
Marlene Gebauer (18:42)
a lot of competition in this space. you have the big information providers are working in this space. You have smaller startups that are trying to crack the legal research solution.
You know, and then you have your buyers who are somewhat inundated by different solutions and generally have some of the larger resources, the larger tools. you know, what are the key differentiators that you think make, MidPage, their offer is superior compared to some of these other players and the, you know, it seems like ever evolving group of AI tools.
Otto (19:27)
So I would say, know, the competitors for us are Lexus, Westlaw, Vlex, and that’s it. And there might be more in the future. There’s definitely lots of big legal tech companies who’ve raised a lot of money and they can always, know, companies always moving and evolving and expanding to different fields.
And at the same time, there’s always this sort consolidation where every couple months or years, one company goes bankrupt and the other one merges. I think we’re now in this consolidation phase where probably the number of players is going to get smaller, not larger. And legal research has been really difficult for startups to enter. That’s why most of our new legal tech startups were actually in the transactional side or contracts. ⁓
they were helping with, sometimes more specifically with dockets, but really legal research because you need to have this huge legal data set that is very hard to build.
and then you also have to compete with these very two established players, like Lexis and Westlaw. It’s just been such a hard thing to get into. But there have been some precedents. I K6 is an example that is very top of mind for people where they built an amazing product. They only had half of the essentials that people really needed for legal research because they didn’t have,
an army of 2,000 people that are marking cases as overruled or not, like what the older players have, they just couldn’t afford that. But at the time, they also didn’t have this really powerful AI that we have today that can replace that same job. And so I would say three years ago, would have been impossible to build a true legal research platform that can compete with But now it’s doable. It’s still a lot of work. But now you can create this data set and you can create all the
things like the citator, the head notes, the case history where you sort of see how do you, what was the procedural history of a case and so on. And ⁓ now it’s possible. So I feel that now this data will become way more abundant and there’ll be more players like us that build new interesting services on top of it.
Greg Lambert (21:14)
Yeah, we’ve had this discussion before where tends to be, well, and we’ll back this up. A few years ago, I would say there tended to be a ceiling that a lot of startups would hit with data. That became a point know, they built a great tool, but they didn’t quite have the data there. And it was either spend $50 million to
get the data or be acquired by someone who already has the data? Are you thinking those days are over that AI tools can help you with establishing the data sets that you need in order to continue to grow?
Otto (21:58)
Yes, yes, we think they’re over. And it’s an interesting time to then enter the market because…
Now it’s no longer impossible, even though it’s still a lot of hard work to create a dataset. But also now the bar is rising, right? For example, Lexis and Westlaw, they have state trial courts, but they have never sort of fully gotten them to the same level. They don’t have the same quality citator. They don’t have full coverage in the state trial courts. And this is mainly because their mostly manual approach of doing things with this huge, partly offshore human labor team.
It didn’t scale to the 100 times larger volume of cases coming from the state trial courts and also people, you know, didn’t find the state trial courts as important for research and they have no presidential power and so on. But now that it’s possible to do it, you know, we’re just going to raise the bar. We’re going to add all those cases too and then suddenly that is the new minimum. when you add more and more things on top, it will actually start getting difficult again for new entrants. So this is probably a time where there’s more shifts happening at once.
then there will be in a couple years when it comes to the data landscape. But yeah, I feel like now there will still be only a few people that have all the data because there’s so much work to maintain at a high level of quality. You know, obviously most of the work goes into ensuring the quality, not the completeness, but ensuring that everything is structured cleanly.
but I think it will no longer be just two players. And so as soon as there’s competition, people will be forced to make this data available via APIs. And then suddenly there’s this whole range of new companies building maybe side check tools or drafting tools that all rely on the same case of data that can be coming from us or some of the other new players.
Greg Lambert (23:34)
Well, it’s hard to talk about legal and AI and not bring up something that I have lawyers knocking on my door asking about. that’s, you know, when I use AI, how do I make sure that my research is correct? It’s not making stuff up or hallucinating. So with MidPage, what are you doing to look at these issues where, you
AI is a great tool, ⁓ may also, since it’s generating information within that generation, it may make stuff up. So how are you leveraging things like your proprietary data set to play a role?
Otto (24:14)
So the easy answer is for every feature that we have, we have an evaluation pipeline. So we have some, know, handcrafted set of a couple hundred examples, and maybe like around 10 different features that use AI. And for each of those, we use a slightly different model based on which performs best.
based on our own benchmarks. And so whenever there’s a new model that comes out that is very promising, maybe OpenAI released a new one, we can run our evaluation on that feature, on that model. And then we get a score. And that score includes things like hallucination. And so it’s become very interesting to see how those, the numbers got better and better over time. Hallucination.
It used to be something that can happen anywhere. It’s now almost impossible to get hallucination unless you’re really encouraging the model to do so, or you have a really weirdly designed system that is just asking the model things that it can’t know. Then it will still try to make up something. But of course,
We’re not just a chat, right? We have these very specific ways in which we use AI, and those are, of course, designed in mind with what AI can do. And so all of the features that we have, it always has some grounding. So it always bases it on the original opinion. And then it always quotes from that opinion and gives you a link so that you can open the opinion of that section. And so both you can verify, and also it’s near to impossible to get our systems to hallucinate now, even if you try really hard on purpose.
Marlene Gebauer (25:27)
You may have answered my question then. So, you know, if you’re saying that the hallucinations are pretty rare, that is good to know because we continue to see in the news about people who are filing documents in court with citations that are made up cases and they’re not sort of checking their work. you,
working to educate users on the responsible and effective use of of mid page and you know what’s the reception like been how has that been so far
Otto (26:04)
Yeah, on the hallucination side, if they happen now, then it’s not really a totally made up wrong answer, but it’s instead like a human that is kind of avoiding your question directly and not giving you the best possible answer, but instead a bit too much of a vague ⁓ way out. And humans do this all the time when they’re not quite sure.
And it’s not as helpful, but it’s also not as wrong as having just a totally made up that any legal tech product, if it still does that today, then they are clearly doing something very wrong because it’s actually quite easy now to avoid this. This was more a thing from a year ago where the AI wasn’t as good.
Now, the problem is that every lawyer now, almost every lawyer, has tried out ChatGPT once. And ChatGPT does not have access to a caselaw database So then it tries to answer those things from memory without having any place to check or to look it up. And then it can still have those errors. And I’m sure that ChatGPT is going to fix that soon. We actually even have a plugin where you can connect our database to ChatGPT. And then it has better access to all those cases.
It’s not quite as powerful of using our actual interface as well, but I think this issue is going to become fixed quite soon and for now I would just recommend everybody to only use ChatGPT for questions where all the material needed to answer those questions is uploaded directly in that chat.
Greg Lambert (27:24)
Yeah, you’re ⁓ talking about the human evading the answer to the question reminded me of a uncle who’s a Pentecostal preacher and he was laughing and said, know, when I get to a point where I’m not really quite sure what I’m talking about, I just turn the volume up and I talk louder. And so it just makes me sound like I know what I’m doing. So it almost sounds like what both.
insecure humans and AI do at the same time. Otto, we’re at a time for our crystal ball question. So we’re going to ask you to pull out your crystal ball and look out just a couple of years for us. And what do you see as the biggest change or advancement in how you think AI will impact?
Marlene Gebauer (27:56)
Think about that.
Greg Lambert (28:18)
the legal research area specifically and perhaps even law more broadly and where do you think MnPage’s role is going to be in
Otto (28:29)
Yeah, so I think one powerful mechanism is going to be that the same litigator will have more more in-depth work product and output. I think everybody is going to get more powerful and more effective. These tools are like a multiplier on your existing talents, I would say.
So if maybe in the past a litigator was able to, in a week, maybe read and assess 100 cases, which would be a lot, next year or the year after, you’ll be able to process 1,000 cases or 10,000 cases with the same level of…
of depth because the tools are just giving you many more pairs of eyes and helping you focus on the most important parts. And I think this is basically as if the guns on both sides are getting more powerful. So the other side has it And I think they’ll continue to be sort of a similar amount of competition before. think small teams can handle maybe new types of cases that they wouldn’t be able to handle before. But I think
More importantly, when you have a big litigation in the future, we will just be really impressed by how they just turned every single stone on the world. They found every case that was ever just mentioning this concept. And they could do this much quicker than before. So I feel that’s where we’re headed more depth. And then, of course, also maybe a slightly easier access.
For the small law firms where you really care about efficiency and the clients are pressing you to like, please don’t even do any legal research. They’re just asking you to give them the simplest possible answer. I feel that you will have clients that are trying to educate themselves. They will use ChatGPT like they’re using it for their doctor as well. they will come up with some wrong things that they think you ought to do as a lawyer. it will certainly be the case that they…
that they have a way of understanding your work much better than before. And maybe that they can even solve some of easier questions themselves, even though you as a lawyer might be afraid of letting them do that.
Greg Lambert (30:24)
I want to go back just to one thing you said right at the beginning when we talking about robotics. ⁓ You had mentioned that you wanted to come into law because the advancements in the language models were pretty much more advanced than the vision models. But like you said, they’re adding the vision models in. So I’m curious from your perspective.
What are some of the, I think what are some of the multimodal aspects of bringing in vision, bringing in audio, things like that, maybe even sentiment analysis. Do you see that as something that you’re gonna see more and more of in the legal field?
Otto (31:05)
Well, would say at least for legal research, being able to understand images is not going to have a huge effect. Of course, it is, everything becomes easier when they can listen to the position transcripts instead of having first the tool that translates the text and then the AI reads the text. but instead of just focusing on the multimodality and adding images, I think they will be better at understanding the whole world and interacting with the whole world. So if there’s a…
Maybe something just got filed to some governmental institutional platform.
These models can now by themselves do some Google searches, go out to the internet, they can click around a certain website, and they can retrieve maybe the regulations, the updated guidebook or something, and they can compare that to the version that is in your research platform and notice that it’s maybe not quite up to date anymore. The models will also just take way longer sequences of actions in the real world on your behalf, whether you’re doing research or something else. so yeah, at that point it becomes really important
to have them be multimodal. And that’s happening already.
Marlene Gebauer (32:01)
Well, Otto von Zastrow, founder and CEO of MidPage AI, thank you so much for joining us today on the Geek in Review. This has been a really insightful conversation.
Otto (32:12)
Thanks a lot. ⁓
Greg Lambert (32:13)
⁓
So Otto, we’ll make sure that we’ll put links on the show notes for listeners to find out more. if they want to learn more about you or about MidPage, where’s the best place for them to…
Otto (32:25)
Yeah, so we have a website. We upload lots of short videos about things you can do on our platform Also users always write us these messages and we have this chat bot on within our app or on our website where they can give us feedback or ask us questions and we respond like within minutes and Or just message me on LinkedIn
Marlene Gebauer (32:44)
Thank you again, Otto, and thanks to all of you, our listeners, for taking the time to listen to the show. If you enjoy what you hear, please share it with a colleague. We love to hear from you on LinkedIn or Blue Sky. And as always, the music you hear is from Jerry David DeSica. So thank you, Jerry. Goodbye.
Greg Lambert (33:00)
All right. Bye everyone.
Otto (33:03)
Bye.