Join us at TNW Conference 2022 for insights into the future of tech →

Your sardonic source for consumer tech stories

It’s 2021 and Twitter still can’t tell the difference between cats and dogs

Huh?

It’s 2021 and Twitter still can’t tell the difference between cats and dogs
Ivan Mehta
Story by

Ivan Mehta

Ivan covers Big Tech, India, policy, AI, security, platforms, and apps for TNW. That's one heck of a mixed bag. He likes to say "Bleh." Ivan covers Big Tech, India, policy, AI, security, platforms, and apps for TNW. That's one heck of a mixed bag. He likes to say "Bleh."

Twitter launched its topics feature almost two years ago to group tweets under different interest categories. It’s nice to see tweets or news on topics that I care about on my timeline.

However, sometimes the social network’s tagging goes horribly wrong. The funniest example I’ve seen is of the algorithm tagging cat pictures under dog topic (and vice versa). Here are some examples:

I asked the company about this mismatch. And in a statement, Twitter admitted that its algorithm makes blunders sometimes:

Topics help customers discover their interests and find great conversations about the things they care about. We use a variety of signals to classify Tweets into Topics, and we know that sometimes we get it wrong. We’re continuously improving our systems and appreciate feedback like this to help us get better.

Sure, I get it. AI is not perfect, and it might have some glitches in identifying objects. But haven’t researchers been working on identifying cats and dogs for years now?

The algorithm might slip in categorizing text-based tweets as it might fail to contextualize tweets. But differentiating cats and dogs in a picture should be more straightforward.

If you look at the tweets above, Twitter has been confused between cats and dogs for quite a few months now. So it’s a bit baffling that it hasn’t tweaked the algorithm.

It’s all fun and games till the company is just mixing up cats and dogs, but given its algorithms have been accused of white bias in past, it should be tweaked before it wrongfully tags more serious topics.

Get the Plugged newsletter

Subscribe to our snarky newsletter all about consumer tech.