The neural network is more of a foodie than you are.

I’m training a neural network to generate recipes based on a database of about 30,000 examples, using the open-source char-rnn framework. It sometimes hallucinates ingredients that it’s quite sure definitely exist and you desperately need them for your recipe.

Try asking your foodie friends if they know where to find the following:

2 oz fresh coriander liqueur
1 cup perchymarrange
8 oz chocolate mustard
4 oz picking oil
1 cup conding beans
2 tablespoon nomained beans
¼ lb lime jean meat
2 cup pickling kee thyme
1 cup water (heeped or fipetsscarn/sneas)
2 oz pkg ground pumpkin sprigs
2 cup stem bread
2 cup walnut tomatoes
1 sand ginger
1 can cracked bread strips
½ cup wripping oil
1 lecture leaves thawed
6 squares french brownings cream
1 cup italian whole crambatch