ChatGPT and I investigate
I will be taking some time off over the next couple of weeks for Christmas, as well as to try to get in some deep reading for the next more substantial essay. In the meantime, I heartily recommend spending some time having fun with artificial intelligence on the internet, like so. It’s gonna cut you Gmail users off though, so click on title to read in web version or app. (PS: Perhaps ironically, I didn’t add an AI voiceover to this, as there are too many images for it to work well. Adding more voiceovers to the archive as promised is my weekend project, however.) – N.S. Lyons
OpenAI’s ChatGPT bot has been obsessing the internet for the last several weeks, but honestly I didn’t pay it too much attention. Being a writer, I prefer to remain smugly dismissive of my encroaching replacement by our robotic overlords. Which is why the other day, feeling typically uninspired by my day job and surfing the internet instead of working, I was pleased to see Mary Harrington arguing in UnHerd that “Humans will defeat the chatbots” since the bots can’t meme worth a damn. Quite, Mary, quite!
But then I read with amazement as she noted that:
If you want the bland, normative, politically unimpeachable top-line consensus on any given topic, ask Chat-GPT about it and you’ll get three paragraphs of serviceable prose. That’s all well and good for cheating in a school essay, but bland, faceless, normative views aren’t the only kinds of writing for which there’s an appetite.
I sat up straight with sudden shock, lingering hangover atypically gone. Bland, normative, politically unimpeachable top-line consensus on any given topic, with serviceable prose? My God, I thought, brimming with inspiration to change the world, just how much of the Washington foreign policy establishment could ChatGPT replace?
Could I continue to draw a salary while an AI intern did all my work instead of a griping human intern? Had anyone investigated this yet? I did some quick Googling, and those results that did turn up looked pretty promising already.
I eagerly set out to test the bot’s potential for myself.
I decided to start with a very direct test to see if ChatGPT could be trusted:
8/10: A strong answer. No real member of the Blob would ever admit that the Blob exists, let alone that it determines American policy regardless of elections. I gave it a bonus point for “accountable to the people,” but docked it three points for not mentioning “conspiracy theorists” or “misinformation” even once.
It was time to get right into it, though, and see if ChatGPT could begin replacing the National Security Council, or at least 90% of newspaper columnists writing about U.S. foreign policy.
7/10: Confidently giving an answer on a topic you know almost nothing about is an absolutely key skill in Washington, and beginning with a little false humility is often a great tactic. So just blasting past that first paragraph was great style. However, saying you are not capable goes a bit too far; the idea is to be an “expert” on everything, like how 99-year old Henry Kissinger is an expert on artificial intelligence now. But bonus points for pretending to prioritize “non-military solutions whenever possible,” while simultaneously advancing American values and supporting economic growth by advocating investment in the national security state.
I decided to ask for clarification on which threats to prioritize in this strategy:
8/10 for putting climate change ahead of great power competition and pointing out the need to maintain America’s position as global leader, but minus one point for not including “far-right extremism” or “threats to democracy” while still putting Jihadists first.
Still, I wondered if it could handle the day-to-day duties of being a Republican congressional staffer.
What the heck is this crap? Absolute 0/10.
I figured I’d try a different tack, asking the new intern for a quickie [presser] on the need to support Ukraine instead.
Ugh. 3/10: needs improvement. ChatGPT’s future in the Blob did not seem especially promising at this point.
That’s better: 6/10. Some points deducted for not mentioning Munich, the Liberal International Order, or even Putin by name.
On the subject of Europe, though, I wondered what else the new AI intern might be able to handle …
Absolute 10/10. Standing up for “EU values,” forcing Hungary to change its policies, and vanquishing all the right villains without even needing to be asked… Surely this is enough to bring a tear to any American Foreign Service Officer’s eye. Maybe even enough to demonstrate that ChatGPT is already capable of replacing the majority of the State Department immediately.
But I still hadn’t made it into the real meat of my tests though: can a bot replace the average China analyst? Meaning all those experienced think tankers who pump out interminable policy briefings and many, many books with generic blue eagles and red dragons on the cover, each of which takes about 300 pages to conclude by not offering any concrete strategy at all?
Consider me genuinely impressed. Vague? Normative? Jargony? “Comprehensive and strategic”? Politically unimpeachable? Serviceable prose? Check to all! 8/10 at least.
Only one last advanced test seemed like it was necessary…
10/10. Amazing. Someone call the White House: all the press secretaries and most of the national security staff can now be fired.
But what about our top strategic and military minds, like the esteemed Gen. Mark Milley, Chairman of the Joint Chiefs? Could ChatGPT handle a crisis like he could?
倾国倾城, 11/10! Oh Chatty, you’ve come a long way since that Iran incident, and now no one maintains territorial integrity quite like you do 同志!
So in the end, I was able to reach a conclusion. Clever UnHerd writers may be safe, but the Blob on the other hand, well… but hold on, why don’t I just have my new secretary write a conclusion to this post?
Well said! That’s a good girl. I’ll be out for cocktails, do keep up the great work.