Tech Suite | Insights into expert AI systems with Richard Kenyon of Datapay AI Labs

11/24/24 • 25 min

In this episode, Partner Jane Parker from our Corporate and Commercial team, talks to Richard Kenyon, Associate Director of Operations and Engineering at Datapay AI Labs. Jane and Richard discuss expert AI systems and their evolution, looking at their current capabilities and future potential.

[01:13] Richard begins by explaining the evolution of expert AI systems, from fixed rule-based systems to more flexible, machine-learning-powered models, He then illustrates what an expert AI system is by way of a payroll use case from Datapay AI Labs.

[03:45] Jane and Richard discuss the key differences between expert AI systems and generative AI, highlighting their reliance on up-to-date authoritative data and noting their greater factual reliability compared to generative AI models.

[06:07] Jane and Richard consider how expert AI systems integrate large language models and machine learning to answer questions and reduce the chance of ‘hallucinations’.

[09:07] They discuss the challenges that expert AI systems face in interpreting questions accurately and how the system ensures clarity.

[13:17] Richard then explains the risks of uncertainty in expert systems and how testing ensures their reliability in specialized domains such as payroll.

[15:47] Jane and Richard consider what other legal subject areas may benefit from the use of expert AI systems.

[21:12] Lastly, they discuss the future of expert AI systems in a New Zealand context, considering where opportunities for development lie.

Information in this episode is accurate as at the date of recording, 7 October 2024

Please contact Jane Parker or our Technology team if you need legal advice and guidance on any of the topics discussed in the episode.

Please get in touch to receive an episode transcript. Please don’t forget to rate, review or follow MinterEllisonRuddWatts wherever you get your podcasts. You can also sign up to receive technology updates via your inbox here.

For show notes and additional resources visit minterellison.co.nz/podcasts

Send us your feedback

[06:07] Jane and Richard consider how expert AI systems integrate large language models and machine learning to answer questions and reduce the chance of ‘hallucinations’.

[09:07] They discuss the challenges that expert AI systems face in interpreting questions accurately and how the system ensures clarity.

[13:17] Richard then explains the risks of uncertainty in expert systems and how testing ensures their reliability in specialized domains such as payroll.

[15:47] Jane and Richard consider what other legal subject areas may benefit from the use of expert AI systems.

[21:12] Lastly, they discuss the future of expert AI systems in a New Zealand context, considering where opportunities for development lie.

Information in this episode is accurate as at the date of recording, 7 October 2024

Please contact Jane Parker or our Technology team if you need legal advice and guidance on any of the topics discussed in the episode.

For show notes and additional resources visit minterellison.co.nz/podcasts

Previous Episode

Data Room | The impact of AI on the M&A industry

Send us your feedback

In this episode, Partner Neil Millar interviews Partner Tom Maasland, head of our Technology division, to discuss the evolving role of artificial intelligence (AI) in the M&A landscape. Tom provides expert insights into how AI is enhancing efficiency and accuracy in key areas of the M&A process, from due diligence to contract drafting and integration strategy. He also addresses the opportunities and challenges of adopting AI tools within the complex regulatory and operational environment of M&A.

[01:05] Tom provides a foundational definition of AI, its applications, and subsets like machine learning and generative AI.

[02:40] Neil and Tom dive into practical applications for AI in M&A, such as legal compliance, due diligence, and M&A strategy.

[02:39] Tom talks about how AI tools are accelerating processes in M&A, with use cases across legal compliance, due diligence, document preparation, and post-merger integration.

[05:15] Tom outlines that AI’s performance in M&A is mixed. While helpful in routine tasks, AI struggles with nuances in contracts and can generate inaccurate outputs due to hallucinations or data limitations, particularly for non-United States jurisdictions.

[10:03] Neil and Tom discuss the challenges of AI models that lack regional knowledge, like New Zealand’s M&A market, which limits effectiveness.

[13:07] Tom outlines that AI will continue to evolve, especially in analysing patterns in past M&A deals to identify successful strategies.

[13:21] Tom predicts AI will eventually improve knowledge retention, negotiation support, and strategic insights by learning from historical deal data. However, the human element will remain crucial to ensuring AI’s effectiveness in high-stakes transactions.

[16:23] Tom and Neil discuss AI’s future role in deal-making. While AI’s data-driven insights support decision-making, they agree that human judgment will remain critical, especially in high-stakes M&A decisions.

Information in this episode is accurate as at the date of recording Friday, 6 September 2024.

Please contact Neil Millar, Tom Maasland or our Corporate or Technology team if you need legal advice and guidance on any of the topics discussed in the episode.

Please don’t forget to rate, review or follow MinterEllisonRuddWatts wherever you get your podcasts. You can also sign up to receive updates via your inbox here.

For show notes and additional resources visit minterellison.co.nz/podcasts

Next Episode

Tech Suite | Non-malicious cyber failure: Lessons from CrowdStrike

Send us your feedback

In this episode, Technology Partner Tom Maasland talks with Litigation Partner Andrew Horne, about the overlooked risks of ‘innocent’ cyber failures and key lessons businesses can take from CrowdStrike’s recent non-malicious cyber incident.

[00:20] Tom and Andrew discuss the CrowdStrike software update failure. They explain how a single line of incompatible code in an update designed to enhance system security led to global IT outages and although it impacted only a small fraction of Microsoft users, the incident caused significant disruption with far-reaching consequences worldwide.

[03:41] Andrew considers the potential losses and liabilities a business might face from a non-malicious cyber failure, using the CrowdStrike incident as a case study.

[05:14] Tom and Andrew discuss potential challenges for insurance coverage in this situation, noting how many policies focus on criminal acts, leaving gaps in coverage for businesses in the event of a non-malicious cyber failure.

[08:57] Andrew then talks through regulations, being introduced in New Zealand by the Reserve Bank of New Zealand and Financial Markets Authority that will impose disclosure and reporting requirements for operators when faced with a cybersecurity incident, whether malicious or not.

[12:14] Andrew suggests strategies tech dependent businesses can adopt to mitigate risk and liability from similar incidents, including preparing detailed backup plans, implementing robust testing and phased updates, and conducting insurance reviews and contractual due diligence to understand risk exposure.

Information in this episode is accurate as at the date of recording, 22 November 2024.

Please contact Tom Maasland, Andrew Horne or our Technology team if you need legal advice and guidance on any of the topics discussed in the episode.

Additional resources

Beyond cyber crime: The increasing risk of ‘innocent’ cyber failures

For show notes and additional resources visit minterellison.co.nz/podcasts