The Xero Responsible Data Use Advisory Council recently held its 7th meeting, highlighting the remarkable progress in the field of generative AI and the myriad of potential applications for small and medium businesses.
The council consists of me, Laura Jackson. popcorn hut (Business Owner) Maribel Lopez Lopez Research (Technical Analysts), Wyndi and Eli Tagi WE mana (Advisor), Aaron Whitman X-bert (app developer), Anna Johnston Salinger’s Privacy (Privacy Compliance Specialist), and Felicity Pereyra of Elevate your strategy (Data Analysis Strategist).
A discussion led by Soon-Ee Cheah, General Manager of AI Products at Xero, explored both the benefits and potential pitfalls of tools like ChatGPT.
Soon-Ee started by asking us to think about what these technologies mean and what their limitations are. When using generative AI in your business, you need to be aware of how confident you need to be in its output. To take an extreme example, a self-driving car that is 99% correct is an unacceptable risk. A 1% error rate means you will eventually fall off a cliff.
On the other hand, the accuracy rate is low in May Acceptable if AI is used to create internal business reports. However, using the same output in a press release can have disastrous consequences if misleading or violating a third party’s copyright. When it comes to using these technologies for tax and financial advice, if the advice is bad, the consequences can be severe and accountability can be difficult to measure. The key is to assess the downside of “misunderstanding” in your particular context.
Limitations of Generative AI
The discussion then turned philosophical. How well can society adapt to such major technological changes? Soon-Yi suggested that historically, humans have adopted a heuristic approach (aka mental shortcuts) to assess truth based on the asymmetry of available information. For example, if 99 recipes for French salad dressing suggest using olive oil and one recipe suggests turnip juice, most people (AI included) ignore the purple variation.But in that world DisFor example, this heuristic approach may not work well if a vocal minority mistakenly believes the election was stolen. Generative AI is only as reliable as the data it feeds.
At this point, Maribel pointed out that generative AI suggests answers without allowing it to examine the underlying data. In other words, it calls for us to outsource our own critical competencies. (suggesting that we might have been better off waiting for these technologies to advance before releasing generative AI into the world).
Generative AI can present legal and privacy implications
The conversation turned to some of the legal issues of generative AI, especially IP and privacy implications. Anna questioned the assumption (perhaps underpinning generative AI) that everything on the internet is in the “public space”. For example, posting personal information or copyrighted material online does not mean it is an open season for AI model training. But she also suggested that regulators would struggle to keep up with these developments and protect personal and property rights.
We are still in the Wild West and many legal and regulatory implications have yet to be resolved. but, lawsuit With widespread copyright infringement claims underway and privacy regulators finding violations of the law in the way personal information was collected from Internet sites, business owners are concerned that the output of generated AI, including code, is I have to be careful about assuming it’s safe to use. .
Despite the risks, the group also agreed that tools like ChatGPT have significant benefits. We’re just scratching the surface of the benefits in terms of efficiency, customer experience and better decision making. Council member Aaron said his company, Xbert, has long used AI to help accounting professionals work more productively. is in the early stages of using generative AI to unlock customer benefits.
We discussed that when using generative AI, you need to be careful about unwittingly giving away valuable data and IP. A point I’ve had to make before is that the apparently “free” version of generative AI comes at the cost of handing over data, and is unlikely to remain free for long. . The old adage “If you don’t pay for the product, you are the product” is true. We all need to be consumer savvy and have a long-term commercial view before incorporating these products into our business model.
Soon-Yi concludes with a human note, offering some reassurance that people won’t be superseded by these technologies en masse. He pointed out that old-fashioned mechanical watches still sell millions, even though digital watches tell you the perfect time. They come with economic value decoupled from efficiency. In an AI-driven future, will companies stand out by offering a human element that machines cannot simulate? That thoughtful note ended a very interesting discussion.