Microsoft recently announced a new caption image bot. This may sound familiar, as they were the ones who released the hilarious how-old.net website. This bot is no different from being hilariously wrong sometimes. What this bot does is analyse an image and tries to describe, in a natural way, what’s in the picture.
The bot was first unveiled at Microsoft’s own developer conference called BUILD. It was created to show off the company’s new AI programs that will be free for developers to use. The idea behind doing this is to try to make their apps smarter by exhibiting more human-like intelligence.
We decided to test it out by using some pictures of Exeter and put the bot through it’s paces.
Starting off with the most known, Exeter Cathedral seems to do well with the bot. I know it’s a cathedral but it can be hard to tell the difference between a church and a cathedral, so I’ll give it the benefit of the doubt here.
The High Street
Ok 2 out of 2 so far. The bot correctly identifies the pavement and the fact there are people moving. This bot is doing surprisingly well…
If you thought the last image was trying to lead you into a big fail I’m afraid not. The quay best known for huge pizzas gets this one a little bit wrong. No boats here.
Toby’s Team Celebrating Their Win
So onto uni events. The caption bot is doing remarkably well so far and that’s the case here. There were so many emotions on the night I don’t blame the bot the emojis used here.
Pretty much sums up any fair on campus.
Rugby Derby 24/02/2016
And here’s where it falls apart, unless my eyes are deceiving me. I don’t think I need to tell you why this is bad. On the plus side at least it got her gender. Try the bot out, you might be pleasantly surprised what it will get right or wrong.