Legal AI: Design and Development (LAWS90286)

Generative AI has rapidly emerged across society, with much speculation about the opportunities and threats it poses.

As a technology that fundamentally accepts natural language instruction, has an ability to reason on that language, and provides natural language output, there are parallels with the work performed by legal practitioners— who receive client instructions, apply legal expertise with the intent to achieve some outcome, and ordinarily provide a written output in the form of a contract, an advice or a submission.

Advancements in tools that abstract away a lot of the underlying complexity mean that you no longer need to be a "software developer" to develop powerful, practical programs. Increasingly the focus is on having the domain expertise and imagination, rather than the technical skill.

To best make use of the current capabilities of this technology, and to form a view on where it might lead, there is no substitute for lifting the hood and using the core components to build practical outputs. This provides a depth of understanding that cannot be obtained by theoretical study alone.

Society will require legal practitioners who are able to understand, build, and critique the machinations of programs that use artificial intelligence (in particular, Generative AI). Accordingly, this timely and hands-on subject provides a much-needed set of tools for those engaged with law and technology: it steps through how to build Generative AI based programs for legal solutions, giving students an opportunity to put that learning into practice and become leaders in AI within their chosen career area.

Principal topics will include:

Overview of Large Language Models
o What is an LLM? How does it work?
o How can we effectively prompt engineer LLMs?

Programmatic Access to Large Language Models
o How do software programs interact with LLMs?
o How can we make LLM functionality accessible to legal users?

Information Architecture
o How do LLMs interact with data?
o What is legal data and how can it be managed?

Developing Applications
o How do users interact with LLMs – data in and out?
o How do we build applications that use LLMs?
o How can we understand and design systems that do legal work?

Measuring and Improving Performance
o What concerns should we be aware of in using LLMs?
o How can we see if an LLM is working well?
o How might LLM performance improve in the future