Blog Image

Publications

Episode 2: Limitations and Benefits of Generative AI

Home Insights Episode 2: Limitations and Benefits of Generative AI

Published on:

Published on: April 05, 2023

Share:

Please note, this is a script from episode two of our video series, the digital download: Generative AI, which you can access here.


Welcome to the Digital Download. In this episode, we're lucky to have Megan Tapsell, Chair of New Zealand's AI Forum join us. Megan will be sharing her insights into the opportunities arising from generative AI tools as well as some of the limitations that New Zealand businesses need to be aware of.

Benefits of Generative AI

One of the benefits for many Aotearoa organisations with generative AI is lowering the barrier of entry. To date, the large data sets required to build and maintain effective AI models is proving an expensive barrier of adoption to many organisations.

Bill Gates has said that ChatGPT is revolutionary and has predicted that whole industries will reorient themselves around it.

Models like GPT are trained on massive data sets which means that businesses can leverage this capability and shorten the time it takes to implement a useful AI model into production, and this is a game changer for our Aotearoa businesses. For example, with the use of GPT, we are hearing reports of huge increases of productivity of software developers, which in turn, will lead to faster project delivery and cost savings.

Limitations of text-generating tools

However, while generative AI tools like ChatGPT have some compelling use cases, organisations looking to use them need to be aware of their limitations. For example, ChatGPT delivers responses very confidently and plausibly, but sometimes gets things wrong and sometimes simply makes things up. Anyone looking to use ChatGPT as a research tool needs to think critically and check what it's saying is correct.

Bias is another known limitation and one that software developers have been grappling with for years. This is because AI is only as reliable as the data set that it's trained on. If that data has embedded within it bias or discrimination, this will be replicated in the outputs it produces. In an Aotearoa context, GPT can process queries into Te Reo Māori and about mātauranga Māori and produce some astonishing results, like generating a whakataukī or proverb. But tread carefully it can get things wrong. There is valid concern that Māori haven't been involved in testing the tools for bias or for false information. GPT may sound like it's fluent in Te Reo Māori, but your safest course of action is to check with a Te Reo speaking human for guidance.

It's inevitable that New Zealand organisations will soon start embracing the benefits of generative AI tools. However, organisations need to be aware of their limitations and apply common sense when using them.

Next week, we'll be discussing some of the privacy, security and other risks arising from the use of generative AI tools. We look forward to you joining us.

Talk to one of our experts:
Related Expertise