Testing Artificial Intelligence (AI)

I was invited by Google to help with their cooking recipe AI!

Invited to participate in artificial intelligence testing, I could not wait to get involved in this particular themed cooking party project. This was exciting to me to see what such computerized intelligence could produce. Based on my professional Chef capacities, my involvement was expected to be from the beginning to the end of the process. The organizer, Misha, a Google Product Leader, expected feedback from me. A quick thank you here to Misha for including me and my husband in this fun venture.

Considering how crazy online recipes have gotten over the years, I wonder how such a system could identify well-created recipes and inspire new ones from such data. My first instinct was that the end results could turn out either ok, weird, disgusting, or a combination of all. A total of 8 people were invited to cook up a storm, enjoy the fruits of their labors, and comment on their experiences. 

The organizer used a large language model (LLM), an artificial intelligence algorithm, that uses deep learning techniques of massive data to understand, summarize, and generate new content such as a seven-course-themed cooking party. 

First, we all had to fill up a checklist providing us with choices of themes, cuisine preferences, food preferences, dislikes, food to steer clear of, dress code, and more. From our lists, the AI created a new comprehensive breakdown for decorations and a Greekbodian (Greek and Cambodian) menu with recipes. Each participant had to shop for detailed ingredients and prepare their unique recipe strictly following the recipe instructions to the letter. Once the assignments and chores were done, we would then taste the food and give our fit back on all the courses. 

The charming galactic atmosphere and table setting were created with a few budget-friendly items suggested by AI. It was just perfect with the addition of the personal touch of the host, using items already available on the spot. Everyone was eager to get going and in a joyful spirit, even before our Nebula Nectar Ouzo Sprits drink!

So, here is what AI generated. A theme party called “Galactic Garden” and the following recipes:

1. Cocktail: Nebula Nectar Ouzo Sprits

2. Appetizer: Stellar Souvlaki Tempeh Twinkles with Tzatziki Terrain

3. Soup: Galactic Avgolemono with Cambodian Constellations

4. Salad: Orion’s Odyssey with Chicken Cosmos and Crispy Rice Rings

5. Side Dish: Pleiades Pesto Grilled Veggies with Phnom Penh Flare

6. Entrée: Bison Black Hole Moussaka with Amok Aurora

7. Dessert: Milky Way Baklava with Celestial Coconut Cascade

As we started to prep and progress towards cooking, we noticed some issues with the recipes. Immediate attention was needed from me to bring appropriate corrections.

We communicated well to solve serious issues with some of the recipes with the permission of our leader. Ultimately, the goal became to make our experience pleasurable while taking notes for future corrections.   

Of the seven recipes, only one was perfect as it was generated. The dessert “Milky Way Baklava with Celestial Coconut Cascade” delighted us all. I personally added it to my favorite recipe list. Most other recipes had mainly unbalanced ingredients, too much of one ingredient overpowering the others, bland flavors, wrong flavoring mixture, lack of seasonings, and poor visual appearance. That said, most foods became eatable with a few tweaks I advised and the special attention given by each recipe preparer. One thing is for sure, rice was omnipresent in way too many recipes.

However, I must mention that one made the unanimity as the most disgusting thing we tasted. The appetizer required my expertise to make it more appealing and eatable. It called for yogurt with mashed cucumber as a base for the grilled Tempeh. Not only the flavor was bad, but the appearance was not appetizing. I simply changed the entire yogurt flavor spicing it up with some exotic spices, a touch of soy sauce (for color), and some other seasonings. To everyone's surprise, it turned out much more appealing and suddenly quite eatable. Ultimately, we had lots of fun with this AI experience which made up easily for the mixed and somewhat distasteful results we got. 

My conclusion to this experience is this. Some of the AI-generated recipes are no different than some of the crazy recipes I have seen online. It is a recipe jungle out there, and what AI produced here represented well such reality. That said, I was impressed by its meticulous organization and the potential of its creativity. I have no doubt that with appropriate corrections and the right professional guidance, a well-trained artificial intelligence can become a great recipe creator or even a good robotic Chef. That is a scary thought, considering how many jobs may be lost in such an industry in the near future. While the taste buds of a great Chef may make a difference today, who knows what tomorrow will bring once an AI is well-programmed and rounded in knowledge. The potential is there no doubt. I have already seen what robotics have created allowing us to produce hundreds of burgers in a flash, a variety of flavorful coffee cups, or refreshing cocktails.

I will finish with these comments. AI is here, and we better be conscious of whatever it may bring. No doubt there are some serious dangers of poor or purposely misleading information guiding AI, ultimately us, to the wrong path. Algorithmic bias, social manipulation, social economy inequality, discrimination, malware, hackers, weapons automatization, poor ethics, surveillance and privacy violations, and uncontrollable self-aware AI are issues that demand serious consideration and are already a concern. How we and AI navigate life with such flaws is probably one of the biggest challenges we all face now. Its proper growth will depend on the development of a culture that emphasizes good ethical data roots, worldwide agreeable standards, and strict legal regulations all monitored on a regular basis by caring professionals. Not a simple task for sure. 

We know that AI learns and is capable of creating new data from large data sources. The recipes showed us such incredible power. I am now wondering if AI will be able to learn to create on its own what is ultimately true, just, right, and safe? Will humans and AI find ways to bring a reliable and secure intelligent system? Will we be able to rely on good engineering practices and alert systems? Will many different AI systems bring more confusion or work well together? To be honest, I question whether it will ever be a perfect world. Why? Because there is so much misinformation out there and evil mingling in our affairs. It would take a super-intelligent AI to figure things out for sure, but who knows at this point. That is the scary part for sure. No doubt, we all should keep a close eye on such technology and not hesitate to point out shortcomings, so proper corrections can be quickly made. If we do, we could see incredible progress and transformation in many industries. Adjustments will be needed in all aspects of our lives and we must be willing to change as technology progresses. With AI, we can bring incredible potential to people and change lives for the better. But we also can bring misery to us all, even destroy humanity, if we allow the growth of misinformed or mischievous AI. Once again the choice is ours, at least for now.

Categories: : Expertise