I am what’s known as a “late adopter” of technology. I have no idea how this happened; perhaps I am curmudgeonly (“Get off my lawn!”), or maybe I’ve gotten used to doing things a certain way that did not involve the latest and greatest gadget.
For instance, my smartphone is an iPhone 12 Mini that will only be pried from my cold, dead fingers once Apple stops supporting the software. I do not need to carry a big screen TV in my pocket.
But the real reason is probably my philosophy that you should never automate something that you cannot expertly do manually. Here’s an example: Back before there were companies selling lawn care software, I developed my own recordkeeping database using off-the-shelf software. It was nothing fancy, but it worked. However, before I designed the computer database, I designed the manual version — pencil and paper — so I knew what I wanted to do and, more importantly, what I wanted to avoid.
Such is the case with the current craze over artificial intelligence (an oxymoron if there ever was one). I experienced two instances recently in which AI proved to be not ready for prime time. The first was last spring’s launch of the MAHA Report, the brainchild of U.S. Secretary of Health and Human Services Robert F. Kennedy Jr.
Now, I think we can all agree that there are issues pertaining to public health that need to be improved, and it falls upon leaders in public policy to articulate a convincing argument to motivate the public to action.
Unfortunately, the MAHA Report was rife with the kind of errors that would have your high school science teacher excoriating you in front of the entire class if you had tried getting away with them. Here’s a pro tip: If you cite a research paper to make your case, someone might check out that reference. But if you rely too heavily on AI for your report, you might find that research paper does not actually exist or does not support the point being argued
The upshot of all this is twofold. First, AI isn’t always particularly intelligent. Second, a great deal of goodwill and credibility can be lost by allowing such a slipshod report to see the light of day.
The second example arose while preparing to meet with a member of Congress recently. We have been targeting members of the House and Senate Agriculture Committees to argue for passage of provisions in the Farm Bill. Like any good retail political effort, we tap association members — in this case, a good friend of mine — living in that congressional district to be part of the meetings. We wanted to discuss a point of pesticide policy that was really in the weeds, so my friend wanted to study some background information on those issues, which I promptly sent over.
What happened next surprised even cynical old me. He fed the background information into one of the AI engines with a prompt to summarize it. The engine dutifully spit out its result, which was exactly the opposite argument that we intended to make. If we had provided the member of Congress with that AI-generated material, it would have completely undermined our effort. It was a good thing that another friend attending the meeting read the document carefully and identified the problem.
Which leads me to my point: While AI can be a valuable tool, it can also be used improperly, just like any other tool. A hammer can be used to build a beautiful home, but it can also be used to flatten your thumb if you don’t have good aim (do not ask me how I know this). AI has the potential to enhance a person’s underlying expertise, but not to replace years of education and experience.
No matter how powerful the computer is, a human being still needs to evaluate the finished product to ensure that it makes sense.
