Apparently Watson, the IBM AI that won Jeopardy, is actually pretty good at making recipes. That said, this is because it analyzes chemical compositions of known good recipes to find the compounds that make us like them and finds things that can produce similar profiles, rather than just sticking strings of text together in new ways.
"Pretty good" isn't really enough when it comes to food or drinks. Those pictures are still giving people more than five fingers on each hand. Extra legs. All kinds of things like that. Why would it do recipes any better?
"Pretty good" isn't really enough when it comes to food or drinks.
It is as long as the results are curated. I'm not proposing to use AI to generate new recipes for every bottle and to just sell them as-is.
Those pictures are still giving people more than five fingers on each hand. Extra legs. All kinds of things like that. Why would it do recipes any better?
But not in all pictures, and there are techniques to reduce these issues. And again, I'm not saying you connect the AI to the production machine and let it run wild. There are fully correct pictures. Why would you not be able to curate generated recipes the same way I can curate generated pictures already?