A-I-A-I–OH?

Last night, I was introduced to CLAUDE.AI. The recommendation came from an individual well-versed in artificial intelligence who explained how she uses it and why I should consider it a tool to support of my own writing. I immediately pushed back. After a quarter century of research and instruction about the creative process, I questioned AI’s ability to perform what I now know are the four fundamental skills associated with creative thinking: suspending judgment, questioning, detailed observation, and most importantly, finding connections where none exist.

The discussion echoed a conversation I would have on the first day of class in the “Imagination and Entrepreneurship” class I taught for nine years at Miami University in Oxford, Ohio. When students asked about my expectations of them, especially when it came to getting a good grade in the course, I admitted I did not have an answer. Everything up to that point in time was history. The best I could do was share my excitement when a student “took me some place I’d never been before.” If any AI engine had bothered to ask be about my expectations of its ability, I would answer the exact same way.

Last night we ended the conversation as we often tend to do. We agreed to disagree. However, I knew we would likely pick up the conversation again and again as AI intrudes deeper into our professional and personal lives. To prepare for that eventual discourse, I took her advice, loaded CLAUDE on my desktop, and took it on a test drive consisting of two conversations.

The first involved a political campaign ad that played during the USA/Paraguay World Cup match last Friday. I entered the following prompt.

Journalist Jonathan Martin, who regularly appears on Morning Joe, appears in a campaign ad talking to GA governor Brian Kemp about Rick Jackson, Republican candidate for governor in 2026. Martin smiles as he points out that Jackson was a donor to Kemp’s 2022 gubernatorial campaign, thus giving a false impression that both Martin and Kemp are supporting Jackson’s candidacy. Is this kosher?

CLAUDE’s response was more than sufficient. It demonstrated its ability to synthesize information from multiple sources. I will share the highlights from its much longer response.

Worth untangling what actually happened here, because the mechanics matter for the “kosher” question. Jonathan Martin is POLITICO’s politics bureau chief and a frequent Morning Joe contributor, and he runs an interview series called “On the Road with Jonathan Martin.” Back in April, Kemp sat down with him at the Masters for that series, a standard journalistic conversation about Georgia politics, Trump, and 2028. At some point Kemp answered a question about Rick Jackson by saying Jackson had been a good supporter who’s a donor with a great story, and that the governor’s race seemed ripe for a political outsider who could self-fund.

Jackson’s campaign then took that footage and built a 30-second ad called “Great Story” out of it, which is what’s airing now during today’s runoff against Burt Jones. [For this part of the answer, CLAUDE credited reporting by the Washington Examiner.]

CLAUDE then presented an uncredited analysis even though, as you will see, there are some assumptions based on legal precedence and ethical behavior. Again, the following are excerpts from a more detailed breakdown of the facts.

So the situation isn’t really “Martin appeared in an ad supporting Jackson.” It’s that a campaign clipped a legitimate news interview and repackaged it as campaign material, which is a different and fairly well-worn practice. Legally, it’s almost always permissible.

Ethically it’s messier, and the mess lands on a few different people. For Martin and POLITICO, the problem is that voters seeing the clip in paid political advertising have no way of knowing it originated from an independent journalistic interview rather than a friendly campaign-trail chat; it borrows the credibility of “this is real news, not spin” and repurposes it for persuasion, which is exactly the blurring that makes journalists uneasy about handing interviews over in the first place.

None of that makes Martin complicit in promoting Jackson. He didn’t make the ad, doesn’t control how his interview gets licensed or clipped, and his question prompting an honest answer from Kemp is unremarkable journalism. The “kosher” question really belongs to the campaign that repurposed it.

Useful, but not exactly what I would call creative. Any researcher could have Googled background on the specific campaign ad and the intellectual property rights and limitations associated with interviews of public figures. As I suggested above, everything about this story was “history,” except CLAUDE missed a critical chapter. It did not bother to ask whether the ad was more than unethical. Kemp may have said Jackson was a “great guy,” but does that equal an electoral endorsement. Obviously NOT! On Sunday, Kemp formally endorsed Jackson’s opponent, Lt. Governor Burt Jones.

Which brings me to my second exercise. As many of you know, in 2023, I published a fictional account of John F. Kennedy’s assassination titled, “In the National Interest.” The idea first emerged in 2004, during the session on “suspending judgment” in my imagination class. I offered the following example. Every theory of the assassination involves an individual or organization which professed a personal or policy vendetta against the president. What if the opposite was true? Even though it took 19 years to research, write and publish the story, the creative seed from which it sprang was always the same. What if CLAUDE had existed 22 years ago? How might it have made the eventual process easier and more efficient? I gave CLAUDE a chance to show me with the following prompt.

All of the John Kennedy assassination narratives involve a malevolent motive? What might be a benevolent motive for the assassination?

CLAUDE began with what I already knew.

Most of the well-known narratives, mob retaliation for RFK’s organized-crime crackdown, Cuban exile or pro-Castro revenge, KGB retaliation, an institutional CIA/military cover-up of the Bay of Pigs failure, even the fringe LBJ-ambition theories, share the same emotional core: fear of exposure, revenge, or the pursuit or protection of power. They read as malevolent because the stated aim is to harm Kennedy or to serve some narrow interest at his expense.

Then CLAUDE admits that it could be done but presents a weak scenario that still relies on opposition to Kennedy’s foreign policy toward the Soviet Union and Vietnam.

One version is the “tragic necessity” model, essentially the Brutus framing: he didn’t claim to hate Caesar, he claimed to love Rome more. Applied to 1963, you could imagine a hardline national-security faction that genuinely believed Kennedy’s quiet diplomacy with Khrushchev after the missile crisis, or his early moves toward Vietnam withdrawal, would invite Soviet adventurism or a wider communist advance, and that removing him, however terrible, was the lesser evil against a future they believed would cost far more lives. That’s still murder and still wrong, but it’s psychologically benevolent from the actor’s own vantage point, a sacrifice framed as preventing a larger catastrophe rather than personal hatred or gain.

CLAUDE finally deals with Kennedy’s deteriorating health which was central to my narrative.

A second, much more speculative version leans on Kennedy’s actual health problems, Addison’s disease, severe chronic back pain, the medication regimen he relied on just to function in public. No serious investigation has ever framed his death this way, and I’m not aware of any real theory along these lines, but as a pure thought experiment you could imagine someone close to him believing his condition was about to become public or disabling, and acting criminally to spare him that exposure or suffering.

Close, but no cigars. CLAUDE could never imagine Kennedy would order his own death. It may be able to synthesize information, but it does not have synchronistic ability, making connections where none appear to exist. It cannot imagine that the origins of Kennedy’s order has anything to do with Woodrow Wilson’s last 18 months in office. Then CLAUDE exposes its own bias.

Both versions depend on treating the actor’s claimed or imagined intention as the measure of “benevolent,” not the act or its outcome for Kennedy, which stays identical either way.

In other words, CLAUDE can only view the motive for assassinating Kennedy from the perspective of the perpetrator. Never from the “victim’s” viewpoint. But it is CLAUDE’s conclusion about a benevolent scenario which I find most damning.

Everything above is a conceptual exercise in how a benevolent motive narrative could be built, not a claim that one is borne out by the historical record.

Anyone who has read my book knows two things CLAUDE could never imagine. First, the credibility of my admittedly highly improbable account depends on the 250+ footnotes, citations from the publicly available “historical record.” Second, one reason I meticulously documented so much of the narrative was to demonstrate that the “historical record” contains so many data points, by cherry-picking, anyone could make a plausible case for almost any conspiracy.

My question is whether this is a permanent flaw in the creative capacity of AI engines or a temporal situation that over time can be addressed. The irony, of course, is I honestly believed I could teach young adults to become more creative. Why couldn’t the scions of the artificial intelligence future do the same with their machines?

For what it’s worth.
Dr. ESP