Microsoft's CaptionBot doesn't understand video games

This is Microsoft CaptionBot. It's a prototype tool (which ANYONE can visit, so go and give it a try) for labelling images. Apparently, CaptionBot can recognise any image and describe it "as well as any human". It's a truly astonishing piece of tech, and represents 1000s of hours of work by some of the smartest, most dedicated people on the planet. Naturally, then, the GamesRadar+ team immediately tried to use it for mischief...

As a team, we fed it a bunch of video game images to see what CaptionBot made of them. The results are shown over the next few pages, and they're all hilarious. We kinda feel bad laughing at such a smart, innovative piece of technology, but we can't help it. And the more we posted, the more under-confident CaptionBot seemed, making us feel rather like the playground bully, flushing the nerdy kid's head down the toilet and... oh my god, we're monsters. Ho hum. Take a look at what CaptionBot things about video games...

CaptionBot says: "I'm not really confident, but I think it's an orange vase"

We say: "See what you're trying to do there, CB, but no"

CaptionBot says: "I'm not really confident, but I think it's a glass display case"

We say: "Go home, CaptionBot, you're drunk"

CaptionBot says: "I'm not really confident, but I think it's a person wearing a hat"

We say: "Sure, close enough mate"

CaptionBot says: "I'm not really confident, but I think it's a close-up of an umbrella"

We say: "That's... not even close"

CaptionBot says: "I'm not really confident, but I think it's a man standing next to a cow and he seems :-|"

We say: "That's actually pretty good, CB. Loving the fact you recognised how sad he looks"

CaptionBot says: "I'm not really confident, but I think it's a woman getting ready to hit the ball"

We say: "Sure, Luigi has a feminine figure, but not sure what you're talking about when you say 'the ball'. Are you feeling ok?"

CaptionBot says: "I can't really describe the picture"

We say: "Yeah, fair play, it's pretty troubling. Hey, why are you curled up in a ball and rocking back and forth?"

CaptionBot says: "I'm not really confident, but I think that's a close up of two giraffes near a tree"

We say: "It... it is not"

CaptionBot says: "I'm not really confident, but I think it's a birthday cake"

We say: "If this was a birthday cake... it'd be incredible. Best birthday ever!"

CaptionBot says: "I'm not really confident, but I think it's a close-up of a toothbrush"

We say: "That would be the most terrifying toothbrush ever. And everyone in the world would probably agree that tooth decay is better than putting this near your mouth"

CaptionBot says: "I'm not really confident, but I think it's a person on a surfboard in a skate park"

We say: "Um..."

CaptionBot says: "I'm not really confident, but I think it's a person holding a hotdog"

We say: "Wildly inaccurate, although - to be fair - there is a hotdog stand in the background"

CaptionBot says: "I'm not really confident, but I think it's a man standing next to a fire and he seems :-|"

We say: "Correct. Also, you would be too, if you'd just escaped a plane crash"