Summarize this content to 2000 words in 6 paragraphs
We used this vintage poster to put three AI chatbots to the test. (GeekWire Photo / Todd Bishop)
On one of my recent trips to Spokane for our “GeekWire on the Road” series, I ran across a vintage poster in an antiques shop, and wondered what year it was from. The poster mentions no year but contains plenty of clues.
As is my tendency these days, I wondered if AI could help.
Turns out the answer is yes, with some caveats. The brief exercise ended up being a good illustration of the current capabilities and limitations of various AI chatbots, and their need for human collaboration and guidance.
Follow along below. Click any screenshot to enlarge, or pinch to zoom on mobile.
Here’s what happened when I tried Google Gemini.
Womp, womp.
I could have kept going, but hey Gemini, if you’re not going to try, why should I?
Anthropic’s Claude 3.5 Sonnet was much more willing to play along.
Nice effort! I pointed out a detail that it overlooked.
I asked Claude if it noticed the date and day of the week, and it said it did.
Some nice sleuthing, with one major problem: Oct. 7, 1965, was actually a Thursday, which I discovered when I double-checked the result with a traditional web search. (Google redeeming itself there.) So it couldn’t have been 1965.
When I pointed this out, to its credit, Claude acknowledged the mistake.
So now we’ve got a year: 1966.
But could ChatGPT get there more smoothly? Let’s give it a shot.
So I had to coax it along.
OK, not quite there. How about this?
And finally, three more questions to get to the answer.
It took some work, but it was reliable, and probably faster than manual research.
For extra confirmation of the year, and a reward for following this post all the way to the end, enjoy this performance by Rusty Draper and friends on the short-lived “Swingin’ Country” TV show, from 1966 on NBC — just like ChatGPT told us.
