Can we let Watson have a go at a DIY recipe?


There has been a lot of talk about Watson, IBM’s supercomputer, creating pretty weird but very tasty and healthy recipes.

Meanwhile someone on reddit posted the way Watson works.

Naturally, this got me thinking. What if we let Watson run free through the official and DIY recipes. @rob @JulioMiles, what thinketh ye?

I could totally see it coming up with something that none of us thought about and seriously surprise us for the positive. Although Soylent is more chemistry than cooking.


That’d be awesome - Watson appears to have some “understanding” of taste, so could conceivably produce complete nutrition recipes with specific taste parameters.


Why not just setup a genetic or evolutionary algorithm to do this. It is probably what Watson is actually using, then you could just let it crunch on a regular desktop for a couple of days or weeks until it spits something out.


I think someone already did something like this (the genetic algo). But I don’t think it incorporated any reasoning and scoring regarding taste (or texture).


@HarveyDesu , because Watson already has all the little bits and pieces to produce taste and texture variations . It’s incredibly hard to produce ratings of sensory variations , because you have to parse and rate texts (or manually create a list.) What’s neat is that Watson is able to infer, correlate, and predict (mostly correctly) sensory variations by processing what’s already in its database. While I wouldn’t call it “intelligent,” I would say that it’s able to produce logical theories based on its knowledge, and is going to be just as accurate as any dedicated recipe production algorithm.

Wheel reinvention is well and good if you can do it better, but I don’t think it’s feasible in this case. Unless there’s an open flavor and combination database that could be easily appended to the existing recipe optimizer…


Yes please yes please yes please


Who said anything about flavour? How about letting Watson loose creating a recipe that’s optimised for students? Or a recipe that’s optimised for ultra marathoners?

If Watson can do this, what would happen to Rosa Labs? Suddenly there’s a machine cranking out the same product they have but at a fraction of the time.


The problem isn’t recipes, it’s marketing, production, and distribution. There doesn’t seem to be a big enough market for the food industry to sit up and take notice (yet.)

Give it a year and we’re going to see either Rosa Labs as the single player, with small growth, or dozens of spinoffs and major competition from Pepsi, Kraft, Tyson, and so on - Walmart, Costco, and Target will have the knockoff brands.

It’s potentially very disruptive to cheap crappy foods, and an ignorant consumer means more money than an enlightened one unless they have a product at the appropriate pricepoint.


Most competition will be cheap knock-offs that offer all the visual appeal but none of the benefits of Soylent. Big companies simply cannot produce something like this, it’s almost against what they are.


I think there is room for competition for things like flavours, different blends for athletes, weight loss etc. I’d expect some of the sports nutrition companies to release competitors to soylent as they already have meal-replacement drinks already so not much of a new area for them and I get most of my diy ingredients from a couple of these places.


I’d LOVE to have Soylent sold at Costco! Just go in and get a 3-month box every once in a while.


Heck yes… Costco is like 1/4 mile from our house, sign me up for that plan!