The Xero Responsible Data Use Advisory Board recently held its seventh meeting, focusing on extraordinary developments in the field of generative artificial intelligence and the many potential applications for small businesses.
The board consists of me, Laura Jackson Popcorn Shed (Business Owner), Maribel Lopez Lopez’s study (Technology Analyst), Wyndi and Eli Tagi of We are (Counselors), Aaron Whitman XBert (app developer), Anna Johnston Salinger’s privacy (Privacy Specialist) and Felicity Pereira Upliftment strategies (Data Analytics Strategist).
In a discussion moderated by Soon-Ee Cheah, GM of AI Products at Xero, we explored both the benefits and potential pitfalls of tools like ChatGPT.
Soon-Ee began by asking us to think about what these technologies mean and their limits. When we use generative AI in business, we need to have a sense of how confident we can be in its results. To take an extreme example, a self-driving car that is 99% correct is an unacceptable risk because a 1% error rate means it will eventually go off a cliff.
On the other hand, a low rate of accuracy May Be acceptable if we want to use AI to help with internal business reporting. However, using the same results in a press release can be disastrous if it is misleading or violates the copyright of a third party. Regarding the use of these technologies in tax and financial advice: if the advice is bad, the consequences can be severe and accountability measures dangerous. The important thing is to appreciate the negative side of “getting it wrong” in the given context.
Limitations of generative AI
Then the discussion turned philosophical. How well placed is society to adapt to such a major technological change, and what assumptions will we have to challenge? Soon-Ee said that humans have historically used heuristic approaches (aka mental shortcuts) to estimate truth based on the asymmetry of available information. For example, if 99 French salad dressing recipes suggest using olive oil and one suggests turnip juice, most of us (intellectuals included) will ignore the purple option. But in the world sister’sinformation – where, for example, a vocal minority wrongly believes that an election has been stolen – this heuristic approach may not work. Generative AI is only as reliable as the data it’s powered by.
At this point, Maribel pointed out, generative AI offers answers and doesn’t allow us to question the underlying data. In other words, it asks us to outsource our own critical faculties. Male-e agreed, highlighting ongoing research aimed at showing the user the factual basis of AI output (arguably it would be better to wait for these technologies to develop before unleashing generative AI on the world).
Generative AI can present legal and privacy implications
The talk touched on some of the legal challenges of generative artificial intelligence, specifically IP and privacy implications. Anna questioned the assumption (which presumably underpins generative AI) that everything on the Internet is “in the community.” For example, posting personal information or copyrighted material online shouldn’t mean it’s open season to train AI models. However, he also said it would be difficult for regulators to track these developments and protect individual and property rights.
We are still in the Wild West, with many legal and regulatory implications yet to be worked out. but with Court proceedings With widespread copyright infringements underway and privacy regulators finding violations of the law when personal information has been extracted from Internet sites, business owners should be wary of assuming that generative AI results (including code) are safe to use. .
While considering the risks, the group agreed that there are also significant upsides to tools like ChatGPT. We only reap benefits in terms of efficiency, customer experience and better decision making. Board member Aaron, whose company Xbert has long used AI to help accounting professionals work more productively, is now in the early stages of using generative artificial intelligence to unlock benefits for its customers.
When using generative artificial intelligence, we’ve discussed the need to be mindful of inadvertently giving away valuable data and IP. The point I’ve been forced to make so far is that apparently “free” versions of generative AI come at the cost of data transfer, and are unlikely to remain free for long. The old adage “if you’re not paying for the product, you are the product” rings true. We all need to be savvy consumers and take a long-term commercial view before incorporating these products into our business models.
Male-E ended on a human note, with some assurance that humans are not going to be replaced en masse by these technologies. He noted that while digital watches tell perfect time, old-fashioned mechanical watches still sell for millions. They have an economic value that is separate from efficiency. In an AI-driven future, will businesses stand out by offering a human element that cannot be simulated by a machine? On this thoughtful note, a very interesting discussion ended.