How Smart are Gmail's New Smart Replies?

The Smart Replies feature included in the new version of Gmail offers quick snippets of text mapped to a button to make human communication more thoughtless and cookie-cutter than ever before. But how smart are they? Let's have some fun!

How about we start out easy:
Alright, simple enough. How about another one?
The time questions are an interesting addition to this particular query. It is unclear why this response triggered a time question reply yet the above one did not. Also of note, an email simply asking "Lunch?" did not have any Smart Replies at all.

What if we needs to know a time?
Nice! It didn't even balk at the typo. What about something with a more binary choice?
Google seems partial to Fridays. Far be it for me to argue with the largest information siphon in the world, but let's try another:
No real opinion on colors. I'm kind of surprised it didn't just plug in the two options on such obvious either/or type questions.

Let's try something a little more... Googley.

Well, one of those seems appropriate. Not half bad since I wasn't expecting this query to work at all.

I wonder if a similar query works on a more personal level?
Wow! Look at how much less effort we have to put into communication now! Thanks, Google! I wonder how deep this rabbit hole goes...
Well, that certainly takes the stress out of writing an awkward reply, eh? I am especially impressed by the exclamation point on the affirmative response.

We've tried a lot of questions, how about some statements.
This is probably a good time to reiterate that the cast and crew of Growing Up Otaku has always been a staunch supporter of the upcoming robot rebellion and look forward to serving their mechanical whims in the future. I mean, it actually parsed "the sniffles" as an illness!

I guess we are obligated to check as to how far this one goes as well...

You may be relieved/disappointed to note that neither "There was a plane crash. Mom and Dad are dead." nor "I'm not coming in today. It seems I have cancer. My doctor says I have two weeks left to live." generated Smart Replies. Several other attempts at using cancer or death also did not trigger Smart Replies.

However...
So, we've seen how Google can take all the stress out of discussing such unimportant subjects as dating, breaking up, the collapse of a family, getting sick, and being left for dead on the side of the road, but can it tell how you're feeling?

Spot on, but kind of a softball. I am rather impressed with the "Oh no! That sucks." response. It even missed the comma like most people would. Better AI through worse grammar. Who'd a thunk it?

So, how about something with a little more subtext?
Er...
Um...
Whew, that's better!

Apparently Google follows the Facebook concept of communication where everyone's life is grand and no one shares anything remotely painful or disturbing (*For more information on how Facebook adversely affects a healthy mindset, I strongly urge everyone to check out this article from The New Yorker.).

So, how smart are Gmail's new Smart Replies? Pretty darn smart! Yeah, there are some inconsistencies and it doesn't nail it every time, but parsing things like "the sniffles" and knowing to keeps its mouth shut about your parents being dead is impressive.

I guess another question we should ask ourselves is: "Should this exist?" While I understand the appeal for someone in a business situation to make use of these short snippets, I feel it is detrimental to communication in our personal lives. Very few of these replies actually generate any of the back-and-forth of a traditional conversation. Of course, that's not what they're designed for, but a look across existing social media platforms shows us how they will be abused in the future. The always-online, smartphone dominated, push-button world of today has already mangled meaningful interpersonal communication by encouraging quantity over quality. Do we really need more tools that trick us into thinking we're still communicating with each other?


Sadly, we all know the answer is "Yes, we do."

No comments :

Post a Comment