ChatGPT Thinks There Are Two R’s in the Word Strawberry, Proving AI Can Be as Dumb as People

There are tons of hopes that people had for the onset of artificial intelligence. Using popular programs like ChatGPT to help us with some of the more mundane aspects of our jobs was probably at the top of the list. You know the type of "busy" work like slogging through spreadsheets and correcting light spelling/grammatical errors.

Or digital file management — sequentially numbering a bunch of files with a similar name but a different number appended to the end of it and breaking up these files into different subfolders under a larger folder could be a great job for AI.

And then there are even more loftier usages for AI — using patterns recognized in human biology to predict and pre-empt cancer diagnoses is another great usage for this wondrous technology.

Instead, many people are using it for creative endeavors like coming up with music, artwork, and completing their homework assignments. Which is a bit of a drag.

But maybe some AI programs aren't so adept at even doing that?

Depending on your experience with using AI software, you might be questioning its efficacy in being able to help you with your daily routines.

It seems like TikToker @yallwatchthis certainly isn't instilled with confidence after this interaction he had with ChatGPT. He asked the program a simple question: How many R's are in the word "Strawberry"?

ChatGPT struggled with the question.

The TikToker informs ChaptGPT that he "counted three," to which the software responded that it separated the R's in the words "berry" and "straw" separately.

He then asks ChatGPT to break down its logic, to which the application happily obliges. It then goes on to highlight only two of the r's in the word, pointing the man's attention there.

Even though he says that he can "physically see 3 R's" ChatGPT goes on to tell him that there are "indeed only two R's."

He then asks ChatGPT if he's "crazy" but the application assures him it's just an error on his part where he accidentally miscounted the R's after taking a cursory glance at the word.

After attempting several times to get the application to count out the number of R's as it spells out the word, ChatGPT keeps replying that there are only two R's in the word, offering up the same vague breakdown over and over again.

At this point it seems like a "who's on first" type of situation with the ChatGPT bot. The TikToker successfully gets the application to spell out the word and then manually map out the correct way to spell the word, listing the position of each of the world's letters with a numerical value appended to each.

After delineating that the word "strawberry" indeed has three r's, when he finally asks the program to tell him how many r's are in the word strawberry again, ChatGPT, again, informs him that the world only has three r's, completely ignoring the "r" in the third position.

However, after he asks ChatGPT to spell Strawberry after removing all of the r's, and then adding them back in and counting them as they put them in again, the application doesn't ignore the first "r" in the word, but rather the one that comes after the second "r" in "berry."

He then asks the app to capitalize all of the R's in Strawberry, yielding the word to look like "stRawbeRRy."

Next, he requests ChatGPT remove all of the letters that aren't R. Next, he asks the app to count them out.

ChatGPT only writes two R's as a response.

It's only until the TikToker informs ChatGPT that it's "missing one" of the R's that the software finally concedes that it's been incorrectly counting the number of R's in the word "strawberry" throughout the duration of their conversation.

TikTokers who saw the video seemed just as frustrated, if not amused, as he was by the back-and-forth with ChatGPT as they debated the number of R's that are in the word "strawberry."

"I cannot believe I watched this entire video and was getting heated," one person said.

Another penned: "Not ChatGPT gaslighting you."

Someone else replied: "This feels like talking to my ex."

But there was one silver lining to the patience demonstrated by @yallwatchthis, and it's that he was able to reprogram ChatGPT to correctly understand how strawberry was spelled. That is, according to this user: "This feels like talking to my ex."

Maybe this is why Dall-E and AI generated art is always miscounting appendages and adding a fifth member of the Beatles to artwork.