Cultivating an Efficient AI Machine: Using AI Today and Tomorrow

Goal: My primary objective is to lay a foundation for understanding AI tools - their capabilities, limitations, and potential impacts on the legal industry. This article will serve as a guide to help legal professionals navigate the intricate and ever-changing AI landscape.

Cultivating an Efficient AI Machine: Using AI Today and Tomorrow
Photo by Thomas Kelley / Unsplash

Introduction

Overview: In response to numerous requests, I'm here to demystify the workings of ChatGPT and similar AI tools. This presentation, initially given to the State Bar on November 2nd, aims to open up the 'black box' of AI, offering insights into its mechanisms and applications, especially in the legal domain.

Context: The realm of AI is evolving at a breakneck pace. Every time I take time to write an article on AI, I fall behind just as fast. This rapid development cycle underscores the importance of staying current with AI advancements, especially for professionals in the legal sector.

Goal: My primary objective is to lay a foundation for understanding AI tools - their capabilities, limitations, and potential impacts on the legal industry. This article will serve as a guide to help legal professionals navigate the intricate and ever-changing AI landscape. It's not just about grasping the current state of technology but also about being prepared for its future trajectories.

Personal Insight: As the managing partner of Promise Legal and someone deeply embedded in the intersection of technology and law, I bring a unique perspective to this topic. My dual background in law and computer science allows me to dissect these tools not just from a user's standpoint but also from a technical and legal perspective. This article blends my professional experiences, aiming to bridge the gap between legal expertise and technological fluency.

Background

Access to Slides: For those interested in delving deeper, the slides from the State Bar presentation are available at txhq.org/hyfnai. This resource is more than just a slide deck; it's a repository of additional materials that can be immensely beneficial. You'll find examples, practical links for implementing AI in your practice, and other insightful resources that complement this article.

About Me: I am Alex Shahrestani, the managing partner of Promise Legal. Our firm is a one-stop shop for tech startups, headquartered in Austin, Texas, with practice extending to Houston and San Antonio. My journey in the legal tech space is marked by founding the Journal of Law and Technology at the University of Texas, which stands as a testament to my commitment to merging legal practice with technology.

In addition to my role at Promise Legal, I'm actively involved in the tech community through positions like the vice president of EFF Austin and vice chair of the open source software committee for the ABA Science and Technology Section. My past experiences include serving as a council member for the Texas Bar's Computer and Technology Section and being recognized as an ABA On the Rise Top 40 Young Lawyer.

My educational background is a blend of law, philosophy, and computer science, with a JD from the University of Texas. This combination fuels my approach to tech law, providing me with an insider's perspective on technology, which is crucial for navigating and understanding the complexities of AI in the legal field.

Privacy Concerns: In the legal world, the topic of privacy in relation to AI is a pressing concern. However, the apprehension surrounding AI and privacy should not be any different from the caution needed when handling any digital tool. Whether it's an email or an AI application, the key lies in understanding the tool and using it wisely. The risk of disclosing sensitive information inadvertently is as real with AI as with any other digital medium. Therefore, the focus should be on understanding AI's capabilities and limitations to prevent accidental disclosures. This applies universally, not just to solo practitioners and small firms but also to in-house counsel and larger law firms.

Understanding AI: At its core, AI might seem shrouded in mysticism, almost like magic. It's easy to view AI as a black box performing astonishing tasks, akin to the amazement surrounding early perceptions of Google Searches or Excel spreadsheets. However, AI is fundamentally about algorithms and mathematics applied to vast amounts of text. Comprehending this is crucial in demystifying AI and leveraging it effectively. The key is to recognize that AI, much like search engines, operates on the principle of matching inputs to a large database. However, AI differs in how it interprets and predicts responses based on these inputs, which is more nuanced and complex than the relatively straightforward task of a search engine linking user queries with relevant web pages.

AI Basics

AI vs. Search Engines: The common ground between AI and search engines lies in their fundamental operation of matching user inputs with a database of information. However, AI differs significantly in its approach. Where search engines are designed to link a query to the most relevant web content, AI goes a step further. It not only matches inputs to a database but also interprets and predicts based on these inputs. This is a key distinction that sets AI apart, making it a more complex and nuanced tool compared to the direct matching process of search engines.

Mechanics of AI: Understanding how AI works demystifies its seemingly magical nature. Imagine AI's database like a stack of transparent sheets, each filled with data. When viewed from the top, this stack appears as a jumble, yet certain patterns emerge – patterns that AI is designed to recognize and interpret. For instance, the frequency of the word "the" at the beginning of sentences forms a noticeable pattern in the data. When you input a query, AI looks for similar patterns and predicts the next sequence based on the data surrounding these patterns. This method explains why AI can sometimes provide different outputs for the same query, as it picks up on the most dominant patterns in its vast database.

Pattern Matching in AI: In the context of legal practice, effectively using AI hinges on understanding and leveraging pattern matching. When formulating queries for AI, it's crucial to align them with the patterns present in the AI's training data. This ensures more accurate and relevant responses. For instance, asking broad, generic questions may lead to AI drawing from a wider range of sources, potentially reducing the precision of the answer. In contrast, structuring queries to echo the patterns found in legal texts or professional discourse can yield more targeted and useful results.

Fine-Tuning AI: Fine-tuning AI is akin to teaching it the nuances of legal language and the specific communication style of your firm. This process involves training the AI to respond in a manner that aligns with your firm's voice and the specific needs of legal practice. For example, you might fine-tune the AI to respond to date-related queries in a particular format consistent with legal documentation standards. This custom response style is essential for integrating AI seamlessly into legal workflows, ensuring consistency and reliability in its output.

Embeddings in AI: Embeddings allow for the incorporation of specific, detailed legal knowledge into the AI's framework. This feature is particularly useful for drafting legal documents, conducting research, and responding to FAQs. By embedding your own documents, case law, or other relevant materials, you can tailor the AI's responses to reflect the specific legal knowledge and context of your practice.

Vector Stores: Vector stores function as organizational tools for managing embeddings. They are crucial for storing and categorizing the specific legal knowledge you embed into the AI. This not only enhances the AI’s ability to provide contextually relevant responses but also facilitates the sharing of these tailored resources among team members, ensuring consistency across your practice.

Function Calling with AI: Advanced function calling in AI opens up possibilities for automation and task management in legal practice. This feature allows AI to understand the context of a task and select the appropriate function to execute it, similar to how workflow automation tools like Zapier operate. This capability can streamline various administrative and legal processes, making AI a powerful tool for enhancing efficiency in legal practice.

Chaining in AI: Chaining involves linking multiple AI tools to handle complex legal tasks. For example, one AI could generate an outline for a legal document, while another could develop topic sentences, and yet another could expand these into full paragraphs. This process allows for the creation of comprehensive legal documents with minimal human input, demonstrating the powerful potential of AI in streamlining legal workflows.

Implementing AI in Your Practice

Starting with AI: For those new to AI in the legal field, beginning with accessible, free tools like ChatGPT or Google Bard is recommended. These platforms provide a basic understanding of AI capabilities and limitations. Initially, it's advisable to use them for non-client related tasks until you become familiar with their functionalities and data privacy aspects. This step-by-step approach helps in building confidence and understanding before transitioning to more advanced uses.

Paid vs. Free AI Tools: As your comfort with AI grows, considering paid versions like ChatGPT Plus can offer significant advantages. These paid services typically provide enhanced reliability, broader features, and more robust data security measures. They also offer additional functionalities like embeddings and fine-tuning, which can be critical for more specialized legal applications. Despite occasional server issues due to high demand, these tools are generally more suited for professional use where accuracy and reliability are paramount.

Bespoke AI Solutions: For legal practices ready to fully embrace AI, exploring bespoke AI solutions is the next frontier. These tailored systems can be designed to meet the specific needs of your practice, offering a level of customization that generic AI tools cannot. While such solutions can be more costly, they offer unparalleled integration into your firm’s workflow. The investment in a bespoke solution should be weighed against the potential efficiency gains and the unique value it can bring to your legal practice.

Cost Considerations

Pricing Models: The cost of integrating AI into legal practice varies widely based on the chosen tools and level of sophistication required. Free versions of AI tools, like ChatGPT or Google Bard, provide a basic introduction without any financial commitment. However, for more advanced features and reliability, paid versions like ChatGPT Plus are available for around $20 per month, offering better performance and additional capabilities.

Efficiency vs. Cost: While the initial cost of bespoke AI solutions might seem high, ranging from a few hundred to a thousand dollars per attorney, they can be a worthwhile investment for larger firms or specialized practices. These costs need to be evaluated against the efficiency gains and the potential competitive advantage these tools offer. It's important to balance the cost with the specific needs and resources of your firm.

Self-Hosted Solutions: For heavy AI users or larger firms, considering self-hosted solutions can be cost-effective. While my most expensive month as a heavy user was under a hundred dollars, a firm of 10 could potentially operate effectively for under 500 dollars a month. This approach offers more control and customization, aligning closely with the firm's specific requirements.

The Future of AI Pricing: The current pricing models of bespoke AI solutions may evolve as the technology becomes more ubiquitous and competitive. I foresee a shift towards a usage-based model, much like OpenAI's approach, where costs are tied directly to the amount of data processed. This could make AI more accessible and scalable according to the firm's size and usage levels.

Conclusion

The Future of AI in Law: The legal industry stands at the cusp of a significant transformation driven by AI. In the next five years, I anticipate that AI will start handling tasks typically reserved for junior attorneys, thanks to its growing efficiency and the increasing volume of data it can process. This evolution underscores the importance for legal professionals to engage with AI now - to understand its capabilities, harness its potential, and prepare for its increasing role in the legal landscape.

Contact and Engagement: I encourage everyone to engage with this evolving field actively. If you have questions, thoughts, or feedback on this article, please feel free to reach out via email (a at rise dot law) or engage in the discussion via the comments section. Your insights and experiences are valuable as we collectively navigate this exciting juncture in legal technology.

Upcoming Opportunities: Stay tuned for upcoming workshops and potential consulting services, which are likely coming soon. These would be an opportunity for more hands-on, interactive learning and practical implementation of AI in legal practice. Subscribing to my blog will ensure you receive notifications about these opportunities as soon as they are available.

A Parting Thought: AI in the legal field is not just about technology; it's about shaping the future of legal practice. By understanding and integrating AI, we can not only enhance our efficiency and service quality but also stay ahead in an increasingly tech-driven legal environment.


For a more thorough, nuanced discussion, see the full talk here.