What is Gen AI and Why Should I Care? | TCDI Talks: Episode 1

TCDI Talks with Altumatim | Episode 1: What is Gen AI and Why Should I Care?

About TCDI Talks with Altumatim: Episode 1

Welcome to the first episode of TCDI Talks, where experts from TCDI and Altumatim discuss Generative AI (Gen AI) and its impact on the legal industry. In this 13-minute discussion, David Gaskey, Caragh Landry, and Vasu Mahavishnu break down the basics of Gen AI, explaining what it is and how it differs from traditional machine learning. They also tackle the important question: why should legal professionals care about this technology? Whether you’re fully invested or just curious about AI, this episode provides a solid foundation for understanding Gen AI’s potential in the legal profession.

Episode 1 Transcript

0:16 – David Gaskey

1:22 – David Gaskey
 
So, Generative AI is a form of machine learning. I mean, if you look at artificial intelligence as a field of endeavor, more so than a specific technology, that field has a lot of subcategories within it. And one of the broader fields is machine learning.
 
Machine learning is for things like pattern recognition and predictions that computers can make. For example, if you are a subscriber to Netflix or another streaming service, the suggestions that you see on the screen are based on machine learning algorithms. It looks at what you have watched, if you give it a thumbs up or thumbs down, takes those things into account, finds patterns, and then makes suggestions for what you might want to watch next. If you go beyond machine learning a little deeper, there’s what’s called deep learning.
 
Deep learning has, you can think of it as a network. They technically call them neural networks, because it’s modeled after the human brain. It’s a network of decision layers, and what deep learning does beyond just machine learning, is it can look at things in hierarchical relationships. So, if you’ve ever interacted with Siri or Alexa, you’ve actually interacted with a deep learning system, because deep learning is what’s used there.
 
Generative AI is a specific type of deep learning. A big difference between Generative AI that we’re all accustomed to now, compared to other machine learning, is you don’t have to do the training.  Machine learning, you typically train the computer to recognize the patterns you want it to know, and then you use it.
 
With Generative AI, all that training has been done for us. Large Language Models have been pre-trained on billions of data points, they call them parameters, and we are able to then just immediately interact with the model – like ask it a question and get an answer based on the information that it’s already had. 
 
So, there’s a lot of uses that come out of Generative AI, and some of those have definitely popped up in the legal profession.
 
3:43 – Caragh Landry
 
And David, exactly what you just covered is what I hoped that we, I hope that would be the message we would get across in this first session…is machine learning. We’ve used that Netflix example in legal for about 10 years now, trying to explain to everybody how Technology Assisted Review works with predictive coding and active learning. And that makes sense.
 
Relating it to Siri or Alexa, or explaining, you know, the difference between TAR and Gen AI, that’s exactly where everybody is struggling right now. So, the simple explanation you just gave, I think is really going to help all of us, especially those still struggling to understand what Generative AI is and how you use it.

4:31 – Vasudeva Mahavishnu

Just to expand on what David said, these models or these algorithms are only as good as what data is trained on. So, if you look at machine learning algorithms, prior to BERT and GPT, is that the training data set. And so, you had good success, you had bad success on that, depending on how you trained it. And people had trouble implementing that.
 
But with Generative AI, being able to train on the world’s data set, the data set from the internet, right. Millions and trillions of data points has uncovered something that I think initially people did not expect, which is that for some reason, these models have the ability to reason.
 
When you typically think of machine learning algorithms, they follow a pattern in what they’ve been pre-trained and apply those patterns. But to apply reasoning, and what we’ve discovered over time is reasoning chains, applying those reasoning chains from the pre-trained dataset and applying that to solving the problem that you presented is what I think is really exciting. And that’s the part that at least we here in Altumatim are very, very interested in applying.
 
6:05 – Caragh Landry
 
I think maybe let’s turn to the “Why do I Care” part, right?
 
I think, David, you especially at the very beginning gave a great explanation of what is Gen AI. And then also you just naturally went into how is it different from, you know, traditional machine learning that on the legal side we’re used to – that Netflix example you gave, suggestions, similarity, pattern recognition.
 
But why do you guys think people should care about Gen AI right now?
 
6:39 – David Gaskey
 
At a philosophical level, I think people should be aware of it and care about it, because as an attorney, you really have a duty to represent your client to the best of your ability. The model rules, the ABA, there’s a model rule that you have to have some technical competence as a lawyer.
 
So, at an ethical, philosophical level, I think people should care because they should do what they can to get the best possible result for their client, and Gen AI can help them get there if it’s applied right. I mean, we just talked about some of the challenges and different things that come along with using it. But if you’re using it effectively, you can better represent your client. So that’s one reason why they should care.
 
The other is, it’s not going away. I mean, there’s been a lot of hype about it, and I think a lot of people jumped on a bandwagon, which I look at as kind of a chat bot bandwagon:  everybody, “Oh, Gen AI, I have a chat bot, I have a chat bot.”
 
But there are higher level capabilities that are going, they’re already being used, and the use cases and the abilities are going to continue to increase. That it is going to transform how a lot of things are done. And being at the front end of that is way better than trying to catch up with the train after it’s left the station. So, I think people should care from that standpoint on – so what do you think?
 
8:16 – Vasudeva Mahavishnu
 
I think – just to expand on the jumping on the bandwagon – a lot of companies, a lot of people, jumped on it very quickly, because it seemed like a very easy thing to implement. Send a prompt, wrap it around an API and you get results.
 
But I think very quickly, many people in the industry figured out that the results were not so good. It’s not because the Large Language Models are not good, they were just not applied correctly. They were not engineered correctly around it. So, there’s a lot more work that needs to happen to implement Gen AI into your product.
 
So, these are ongoing challenges as the models keep getting better. But every iteration of the model getting better, you have to re-engineer your product to take in all the latest features or account for the failures so that it is more complex than initially thought when implementing Gen AI into your product.
9:23 – Caragh Landry
 
So, for me, it all comes down to volume, cost, and deadlines. Right? Like there’s way too any large effort and repetitive tasks that humans are doing right now that we just don’t have time to do, and they cost way, way too much because of the human effort that needs to be done. And we’re seeing that these tools are doing it better.
 
Not better necessarily than humans, but I think there’s an element of that. It’s definitely more consistent, and consistency can be an improvement, but better than the way that we have been doing it.  So, things like redaction, redaction for PII, redaction for other products or third parties. Those are things that can be easily automated, easily tracked and QC’d rather than fully done by a human.
 
First pass review, the initial categorization of what’s responsive and not responsive. That can take months to, you know, definitely weeks to months to, you know, lots of effort, to find what’s responsive, but more importantly, what’s not responsive to get out the result of bad search terms or an over-culling or an over-collection.
 
So, document summarization. It takes us humans forever to summarize not only a document, but a set of documents. So, if you think about witness prep and depo prep, like taking 300 responsive documents and distilling it down to a summarization of what you might want to talk to somebody about, takes us weeks, because we have to read the documents, we have to understand the documents, and then we have to synthesize the documents into something that’s concise, which humans are not good at.
 
I mean, just think about all the words I used to explain.
 
This new Gen AI can do it so quickly. So quickly for a lot less cost as well. So, we’re focused on using, we care about these tools, because our clients expect us, as David as you said, to deliver the best service, the best product, because we fall into that ABA regulation. We also need to be delivering. We need to be keeping abreast of new technology, and we need to be delivering the best to our clients that we can. And right now, there are certain tasks that have been identified for months to over a year now, that these tools are better suited for than humans.
11:57 – David Gaskey
 
That brought up a thought, too, is when you’re talking about these repetitive tasks that takes so long. Another reason to be concerned about it is work satisfaction.
 
I mean, at our law firm, that’s one of our core beliefs is that we should have a high level of work satisfaction. So, I don’t know too many people who went to law school to sit and slog through thousands of documents for weeks at a time. I mean, you go to law school because you want represent clients. You want to argue motions. You want to strategize and take that deposition, not try to summarize 400,000 documents to get to the deposition.
 
So, really, even just from a work satisfaction or improving people’s lives, I think law firms who are thinking strategically are thinking: how do I attract good talent, and how do I keep those people happy? You know, that keep the talented people in the firm so they grow into the leaders of the firm as opposed to burning out because they got stuck in doc review for six months.
13:07 – Caragh Landry
 
I think this was great. I think this was exactly what we were hoping to do for today.
 
So, first of many. Series to come. This was foundational, you know, what it is, why it’s important. Going forward, I think we’re gonna explore some of the deeper themes that David and Vasu that you guys touched on.
 
So, hopefully people will join in for those ones too. But I think we should end this one here.
 
13:31 – David Gaskey
 
I agree.

13:33 – Vasudeva Mahavishnu

All right, it’s been fun.

13:34 – Caragh Landry

Yeah, all right. Well, thanks everyone for listening. Tune in next time.

Meet the Experts

Caragh Landry

Caragh Landry | Chief Legal Process Officer | TCDI

With over 25 years of experience in the legal services field, Caragh Landry serves as the Chief Legal Process Officer at TCDI. She is an expert in workflow design and continuous improvement programs, focusing on integrating technology and engineering processes for legal operations. Caragh is a frequent industry speaker and thought leader, frequently presenting on Technology Assisted Review (TAR), Gen AI, data privacy, and innovative lean process workflows.

In her role at TCDI, Caragh oversees workflow creation, service delivery, and development strategy for the managed document review team and other service offerings. She brings extensive expertise in building new platforms, implementing emerging technologies to enhance efficiency, and designing processes with an innovative, hands-on approach.

David Gaskey | CEO and Co-Founder | Altumatim

David has been at the interface between law and technology for more than three decades. Specializing in intellectual property law, he has represented clients from all over the United States, Europe and Asia, including Fortune 50 companies, whose businesses involve a broad spectrum of technologies.

David has extensive experience litigating patent disputes at the trial and appellate court levels including the Arthrex v. Smith & Nephew case that received an “Impact Case of the Year” award in 2020 from IP Management. His litigation experience was a primary influence on how Altumatim naturally fits into the process of developing a case and why the platform is uniquely designed to help you win by finding the most important evidence to tell a compelling story.

Vasu Mahavishnu

Vasudeva Mahavishnu | CTO and Co-Founder | Altumatim

Vasu brings his natural curiosity and passion for using technology to improve access to justice and our quality of life to the Altumatim team as he architects and builds out the future of discovery. Vasu blends computer science and data science expertise from computational genomics with published work ranging from gene mapping to developing probabilistic models for protein interactions in humans.

As a result, he understands the importance of quality data modeling. His extensive experience with business modeling, code construction for front-end and back-end systems, and graphic presentation influenced the architecture of Altumatim. His creativity and commitment to excellence shine through the user experience that Altumatim’s customers enjoy.

Ready for More?

TCDI Talks - Checking In: A Unique Approach to Gen AI Video Series Thumbnail
TCDI Talks, Episode 3: What is a Specialized LLM?

If you enjoyed this video, feel free to check out some of our other great content!