Three bad recipes generated by neural network

I’m training a neural network to generate recipes based on a database of about 30,000 examples, and although the network has managed to produce identifiable recipes, and even sometimes sort sweet from savory, it hasn’t actually managed to produce any good ones. Only a very few of them are technically doable. Three typical examples:

Citran Barbecued Mube
game, ethnic

—-CAKE—-
1 pkg cornstarch
34 oz ginger
1 white sage
2 large red potatoes, peeled
1 magazine bread; chunks
1 cup shredded corn peas
4 cup liquid ice cream

Preheat oven to 350 degrees. Remove the casserole from the refrigerator and heat soup for 20 min. Serve with nuts, chopped caczooled and serve with the rice and oatmeal.
From: Fial Hosselr Date: 24 Jul 96

Grilled Snailsed Butter
crockpot, vegetables, crockpot, rubbing holiday, meats

2 lb shrimp; cut in ½ cubes
6 cloves, minced
2 teaspoon apple juice
¼ cup mushrooms
1 lb tomatoes, nuts.
plastic
1 Strawberries
2 each pinto beans; sliced
1 plum tomatoes, (no carri-fater)
1 pkg unknown yogurt fillets, thawed

Pour noodles and cauliflower through a wider measure just on high speed until stiff. Flavor radicchio mixture with the wine and continue simmering until mixture is desired doneness. Reserve side of bowl mixture. Chill until circle is reamy inricating. Serve on ranged removable pieces.
Smushed.I’s Bried, Heritame Sprigs
cakes, pies, pastries, extract

2 eggs
4 tablespoon water
1 cup dried butters and firmly beaten
20 eggs
4 oz fresh chopped nuts (approximately 10 minutes)
2 tablespoon grated zucchini
20 oz almonds, rough
1 cup seasoned baking powder

Sift milk in crockpot. Turn dough.

Add egg powdered sugar and whipped toppings. Serve immediately.

Posted by Adwing A