THIS IS AN OPINION
We'd also like to hear yours.
Tweet us @ArkBusiness or email us
Artificial intelligence and large language models, like Chat GPT, look like they’re going to be the innovations of our lifetime, but there are going to be some bumps along the way.
Most recently, look at Mark Friedman’s reporting on the northwest Arkansas lawyer who trusted Chat GPT to draft his legal briefs for him.
It turns out that Chat GPT — as it is wont to do — decided it would be no big deal if it made up some cases and a few quoted passages out of thin air to bolster its arguments.
And the lawyer, Tony Pirani, didn’t double-check Chat GPT’s work.
“It just did not occur to me that it would completely make things up,” Pirani told Friedman in an interview.
Well, I’m here to tell you, Chat GPT, Claude, Grok — all of them most definitely will make things up for no apparent reason.
We have not used LLMs directly for content in Arkansas Business, but we do use them to analyze data, documents, survey responses and the like. We meticulously double-check the work — often to the point of losing much of the efficiency we sought to gain by resorting to AI — and find made-up information inserted more often than we should.
In one strange case, we had Claude, Anthropic’s LLM, helping us compile and format a series of survey responses. It was only supposed to compile provided questions and answers into a particular format, but we found it randomly added answers that weren’t in the survey.
There are some tricks to this, like sharpening your prompts to instruct the LLM to only use information you’ve provided and to attribute its work to verifiable sources. But these aren’t foolproof, as our case demonstrated.
And yet, we can’t let the fear of an incident like Pirani’s case hold us back from experimenting and looking for more ways to incorporate AI into our businesses. There are too many efficiencies to be gained, too many innovations to unlock and too much risk of being left behind.
What we must be is careful and curious.
We hope to learn how to do that better on Aug. 15 at the first Arkansas AI Business Conference, which Arkansas Business is co-hosting with the Arkansas chapter of the Public Relations Society of America.
The all-day event at the Clinton Presidential Center in Little Rock will feature a number of experts on AI implementation, ethics, law and more. Speakers and panelists include Elizabeth Edwards, a national expert and speaker on ethical technology adoption, and Allyson Lewis, founder and CEO of 7 Minute AI and a member of Gov. Sarah Huckabee Sanders’ AI task force.
How can we get our employees to test the AI waters? What tailored training programs are out there? How are the best-positioned companies using AI? What are the greatest legal risks?
A full agenda and registration can be found at arkprsa.org/meetings-and-events/aiconference. Registration closes today.
Sign up for Daily E-news
