Select Page
Gordon Smith

About Gordon Smith

Journalist by day, cunning linguist by night. A passion for politics, hypnotically involved in human rights. An Australian born with a Japanese tongue, hoping to hold the big wigs in government to account.

Approx Reading Time-10Finally, a glitch in the system. Someone let an AI organise dinner, and the results are not good.




As mankind continues its journey towards technological excellence, so too does it journey towards its robotically induced obsolescence.

While AI is slowly but surely inching closer and closer towards its inevitable domination of the workforce – including the until-now sacred duty of naming Pokémon – there is at least one area where cold metal hands will never take hold:

The kitchen.

Yes, we may never see a day where our Iron Chef is, in fact, an iron chef, nor a day where our KitchenAids no longer need our aid. At least, that’s judging by the performance of research scientist Janelle Shane’s neural network-powered program. By feeding the beast hundreds of pages of cookbooks, she has taught the emotionless being to automatically generate recipe titles. Better still, she can adjust the algorithm used from what Shane describes as “low” creativity – including such classics as “cream of sour cream cheese soup” and “chocolate chocolate chocolate chocolate chips” – to higher levels of creativity.

Such artistically inspired meals include the confusingly named “completely meat chocolate pie”, “whole chicken cookies”, or the mysterious, all-powerful “star *”.

If your socks have still not been thoroughly knocked off, the AI has also been taught the magic of recipe making.


While that may seem like the tormented cries of a being wrestling with its own conscience, it’s lightyears ahead of its earlier incarnation.

“Immediately Cares”.

As Shane notes, this is a particularly impressive improvement given the program’s limited analytical abilities, which caps out at 65 characters.

“By the time it begins the instructions, the recipe title has already passed out of its memory, and it has to guess what it’s making,” she notes in a blog post. “It knows, though, to start by browning meat, to cover with plastic wrap before chilling in the refrigerator, and to finish by serving the dish  This is from a network that’s been trained for a relatively long time – starting from a complete unawareness of whether it’s looking at prose or code, English or Spanish, etc, It’s already got a lot of the vocabulary and structure worked out.”

Even more impressive are the leaps and bounds the program has made from its very first attempts, which are more primal screams than any concerted attempt at culinary excellence.

So, with that in mind, are we being a little too critical of baby’s first computational meal? Well, maybe.

But, at the same time, considering our up and coming kitchen nightmare learns not by taste, nor any real skill, but by analysing vocabulary and sentence structure, I think it’s more than fair to say that the idea of a robot serving us our favourite meals is about as far-fetched as they come.

Cocktails, however, may be another story.



Share via