The first time I saw Sleepless in Seattle growing up, I never thought I’d end up living in Seattle… I even remember where I first had lunch when I moved there (spoiler: a place that appears in the movie). Since then, I’ve seen it a couple of times, but it’s been over 10 years since the last time I saw it. It showed up in Netflix, so I put it on in the background while I was doing other stuff.
I love seeing how movies depict human-computer interactions or user experiences (UX) — naturally, computers showed up in this movie: there’s a scene where Meg Ryan’s character (Annie) is investigating Tom Hanks’ character (Sam). She works in a newspaper, so she has access to a news archive of sorts. Let’s see some screens.
To set the context, here's Annie (totally not creeping err… stalking err…) investigating Sam:
Now let's see some screens!
Screen 1: Kind of reminds me of a *nix system, where the commands (requests?) are well defined (I’d add a “type ? for help” since there could be users who are not familiar with this system… although you could have needed formal training for some systems… anyway). Annie capitalizes FIND so maybe there are some well-defined commands in this news-searching system.
In this case, would this be a wildcard search? Entering the name Samuel Baldwin could pass off that query to try and match everything — titles, sources, content, known individuals, etc.
Tangential note: In 2023 the tech industry is still trying to deliver on this — ChatGPT could do a good job at identifying what a user’s intent may be behind a request, though (almost there!). We'll get there later.
Screen 2: Looks like the system immediately picked up the parameters (Sam’s name) and it associated it with a person or contact (maybe that’s why it adds an "s" to the end indicating plural). At this point, I assume this UX is made-up. If it was a wildcard search, there could have been many more results likely matching Samuel, or Baldwin, or Samuel Baldwin.
I think the wording is interesting too — instead of saying “PRESS Y”, it uses the verb “STRIKE” this feels old school. Also… “enter factors”… is this like… factors of the person? Should it be filters, attributes, characteristics? So many possibilities…
Image 3: Since there are too many results, Annie does a sort of SQL INTERSECT with Sam’s name and Sam’s son’s name (Jonah) — hoping to find results that have them both.
Screen 4: Success! However this seems strangely inconsistent — we saw a number of results before, but this time, there is no number.
Screen 5: Apparently, there’s only 1 result. A sad result.
Screen 6: Annie knows that Sam lived in Chicago, so maybe focusing on that fact will yield useful results?
Note how at this point it’s a free-for-all… a search anarchy… searching feels like the wild west. What if Sam’s last name was similar to a city name? What if one used city name, then person name? Anarchy, I tell you.
Screen 7: A manageable set of results! After Annie struck Y, the system printed out not only the names, but a surprisingly relevant key attribute of each person. Yep. Totes made up (I think).
Annie decided to select only one of the list:
Screen 8: The selection mechanism seems to work by underlining character by character… maybe the item selected using the cursor becomes the parameter(s) to select the specific item (but what if you kept underlining past the characters? Or the row?).
I’d likely have used numbers, and then paginate, then ask the user to enter the item number from the current page. In any case, I’M SO GLAD WE HAVE GUIs!
Screen 9: Immediately after selecting this, Annie launches… THE FAX INTERFACE. I wonder what the other default actions were. Maybe fax isn’t too bad, but I’d assume that you could at least print (dot-matrix style, naturally), or maybe email (assuming you at least have PINE)?
… I assume the To: is for the name/contact, especially if faxing this to a large business. Also, you can save this to a file! Nice!
That whole process took a couple of minutes, which is not bad.
What if Annie was trying to investigate Sam in 2023?
Annie could use Google, Facebook, Instagram, etc. It would definitely take longer, but there'd also be troves of data to sift through. Besides all those sources, here’s what would she could see in ChatGPT and Bing Chat.
What if we asked OpenAI's ChatGPT who is Sam Baldwin?
ChatGPT tries to play it safe by not immediately giving out details or hints whether it knows how many Sam Baldwins it is aware of. After poking a bit more, there is some historical result, which is not bad.
It'd be awesome if it had identified that it is a fictional movie character… maybe with a little nudge:
What if we asked Bing Chat who is Sam Baldwin?
Pretty similar, but with fresher results and handy links to check sources. Still — same issue as with ChatGPT, however, with a little nudge:
Investigating someone still requires a lot of manual work — to parse a bunch of potential data stores/services, which these modern tools might not have access to because they're walled gardens.
Bonus Screencap
A fun reminder that nothing has changed since 1993: