By Dr. Ken Broda-Bahm:
Right now, we are in a liminal moment with technology. No, I’m not talking about the new Apple goggles, necessarily, but about seemingly sudden emergence of engines like ChatGPT, which have opened a lot of possibilities on where it is likely to go. The proverbial jury is out about how and when generative artificial intelligence will be changing our lives. Many seem to prosthelytize a bright future with many human tasks being eased, while many others are issuing dire warnings about reduced human creativity and a slipping grip on reality. Amid the intellectual turmoil, one truly fascinating (to me, at least) fact has emerged: AI lies. And, by all appearances, it lies easily and often.
For lawyers, here is a close-to-home example. CNN reports on a New York lawyer now facing sanctions on top of the ire of the court for filing a brief created with the help of ChatGPT that includes multiple legal citations that are simply made up. The cases found by the AI helper are realistically described and cited according to the standard forms, but they do not exist. Others in the legal world have since stepped forward to point out that this doesn’t seem to be a fluke. I learned of one colleague here in Colorado who followed up by asking ChatGPT a basic question about Colorado law, and the engine created a cogent analytical summary on the issue that sounded authoritative. But it also invented the statute it quoted, and invented several cases interpreting that non-existent law. As in the New York example, when called on the falsehood, ChatGPT doubled down and insisted that the fake laws and cases were actually real. As the Colorado colleague put it, ChatGPT “is a good and confident liar.”
The Risks of AI in a Field Based on Truth
Using AI to help with briefs and summaries of the law is just one example of a proposed use that could change the future legal landscape. I have also seen recent discussions of using AI to search background summaries on parties, witnesses, or jurors. It seems like it clearly could do that, but would the information be accurate, or would the technology simply aim for the appearance of accuracy? One would think that the bot could be given the firm instruction to fact-check its own work, and one hopes that as legal research companies like Lexis-Nexis move into the space, that will be included among the “prime directives” for AI (along with “don’t enslave the human race”). But the recent controversies open the question of whether AI will actually reduce work if all the work it produces then needs to be fact-checked. And there is also the dark possibility that as more and more online content becomes AI-generated, at some point the task of fact-checking itself will become more difficult, or even impossible. Interesting times, indeed.
The Precautions
Obviously, Fact-Check It
For now, at least, generative AI is not ready for your reliance. Whenever you intend for someone to trust your factual representations — and briefing as well as legal correspondence would certainly fit that category — you need to fact-check before it goes out. That applies whether the work was created by you, a clerk, an associate, or a semi-sentient computer network.
Don’t Displace Your Creative Voice
It is clear that ChatGPT and similar engines can do a very passable job at creating natural prose. That ability is likely to improve exponentially. But at present, the creative capacities are still pretty basic. Potentially, it could have written this blog post — but it didn’t! I put it in the same category as other sources of help. I work in a large firm so marketing people are available, and they’re great writers. Still, in all these years I have never used anyone to ghostwrite any of the content. Why? Because for me, the writing is the point. That’s where I learn, and that’s where I express. I think the same applies to the practice of law. It’s a profession, sure, but in many ways it is still an art. And part of what it is to be human is to pursue art wherever we can still find it.
____________________
Other Posts on Artificial Intelligence:
- Expect AI to Change the Nature of Legal Work…and Soon
- Consider the AI Influence on Medical Liability
- Learn from AI in Settlement Negotiations
____________________