• Skip to primary navigation
  • Skip to main content
  • Skip to primary sidebar
  • Skip to footer
  • Skip to navigation
Close Ad

The Spoon

Daily news and analysis about the food tech revolution

  • Home
  • Podcasts
  • Events
  • Newsletter
  • Connect
    • Custom Events
    • Slack
    • RSS
    • Send us a Tip
  • Advertise
  • Consulting
  • About
The Spoon
  • Home
  • Podcasts
  • Newsletter
  • Events
  • Advertise
  • About

Robotics, AI & Data

July 10, 2023

MIT & NVIDIA Researchers Are Building Tech That Could Enable Better Kitchen-Robot Precision

This week, a group of researchers from MIT and NVIDIA are showing off a system that one day may be pivotal in helping our robot chef make dinner without making a mistake.

While robotic planning systems are good at developing high-level plans, they often fail when confronted with highly-complex environments. Because of this, the group wanted to create a task-planning system that performed well in complicated scenarios with many obstacles.

The project focused on developing a task and motion planning (TAMP) algorithm to help robotic systems solve mobile manipulation problems in difficult environments. The core of the algorithm is PIGINet, which the group describes as a transformer-based learning system that, for each proposed task plan, helps the system more quickly understand the success probability of a given motion trajectory.

Today’s robotic system task planners often fail when faced with the reality of highly complex and infinitely variable real-world scenarios, getting bogged down in processing how to navigate through the unique physical geometries of their environments. The seemingly infinite variety of small things in a kitchen – random items on a counter, the different locations of a pot on a cooktop, open doors and drawers – may be easy for a human to handle but can give a robot fits. With the PIGINet transformer, the system will be able to more quickly process through and understand the success probabilities of each course of action due to the specific start state and the given obstacles within.

According to the group, the PIGINet transformer-enabled task planner gives the robot a better chance of success by better understanding the various scenarios and each’s feasibility before they are executed. Their initial experiments showed that using PIGINet substantially improves planning efficiency, cutting down runtime by 80% on problems in relatively simple scenarios and up to 50% in more complex ones.

While the group’s initial effort focused on kitchen and food-planning tasks, it believes its system can be applied to other tasks within and outside the home.

While there have been a lot of venture capital dollars and product development hours spent on developing kitchen robotics, you can see by this project and those similar to it just how early we are in developing truly advanced kitchen automation. The kitchen is one of the most complex and variable work environments, and creating a robot that doesn’t simply automate a single repeatable process is extremely difficult. With projects like this one and EPIC Kitchens, we are laying the foundation for our robot chef future.

You can watch a video on their project and how it works below:

PIGINet: Sequence-Based Plan Feasibility Prediction for Efficient Task and Motion Planning

July 3, 2023

Researchers at Cal Poly Are Studying The Social Impact of AI & Robotics on the World of Food

Last fall, a group of researchers at Cal Poly was awarded a $700 thousand grant from the National Science Foundation (NSF) to study the social and ethical impacts of AI and cooking automation.

The study will last four years and explore the benefits and risks to individuals and the impact on family and communal relationships, creativity and culture, economics and society, health and well-being, and environment and safety.

The study is led by Patrick Lin, a philosophy professor and director of the Ethics + Emerging Sciences Group at Cal Poly.

“Robot or AI kitchens would automate a special place and communal activity in the home, so that immediately warrants critical attention,” Lin said in the announcement. “Outside of the home, restaurants are one of the most essential and oldest businesses, given the primacy of food. They are the bedrock for an economy, the soul of a community, and the ambassador for a culture. But the pandemic is causing a seismic shift in the restaurant industry, and robot kitchens could be a tipping point that forces many restaurants to evolve or die in the coming years.”

According to Lin, the primary work output will be a public “ethics impact report” that evaluates the societal impacts of robots and AI on this “last mile” of food automation. This will include examining everything from robots flipping burgers or making restaurant pizzas to using AI and robotics in the home to produce and create complete meals.

It’s an interesting project that came onto my radar because Lin personally invited me to participate in a workshop hosted at Cal Poly to discuss the impact of robotics and AI on the last mile. While I usually don’t participate in these types of research projects, I decided to take him up on it since this is an area that I’m pretty fixated on of late.

One potential area I am particularly interested in is how human workers will react to the addition of automation to their workplace. While I expect some workers will embrace the opportunity to use technology to make their work-life easier, others will bristle or outright resent some of their previous tasks being taken over by automation.

One operator who experienced this firsthand is Andrew Simmons. He recently saw former employees undertake a social media campaign to disparage his restaurant for using robotics in the kitchen, including reporting the restaurant to the local health department. What’s interesting about Simmons is, unlike many of the headline-grabbing robot installations at national chains like Sweetgreen, he’s a small one-restaurant operator who is reinventing his entire restaurant workflow through an automation-heavy tech stack. I imagine other smaller operators will attempt to follow the template he’s created (he says he could automate future restaurants for $70k), particularly if he shows he can be successful.

As restaurant robots become lower-cost and more accessible, there’s no doubt society at large will need to think through what the impact will be. I’m excited to participate in Lin’s workshop to help think some of these through, and I hope to share some of the insights from the workshop. I will be limited in what I can share – Lin explained that the workshop would follow the Chatham House Rule, which forbids the identification of other participants without their expressed consent – but I do plan to write about some of the key insights discussed at the workshop in the future, so stay tuned.

For those who didn’t get an invite to this workshop and want to discuss this exciting topic, I suggest coming to The Spoon’s Food AI Summit, which is taking place in the Bay area this October!

June 21, 2023

Karakuri Joins The Growing List of Food Robot Startups That Have Shut Down

More bad news on the food robot front.

Karakuri, a startup that made a robotic food kiosk that assembles various cold and hot ingredients into prepared meals, is shutting down, according to founder Barney Wragg.

In a post on Linkedin, Wragg cited the pandemic and the challenging fundraising environment as the reason for the news and included a link to a Google Sheet with Karakuri employees who Wragg said it was “incumbent” on him to assist in finding new roles.

From the post:

It’s with a very heavy heart that I have to report that our journey at Karakuri is coming to an end.

For the past five years, we’ve developed and deployed robotics for the QSR industry. We’ve survived many challenges, including the pandemic and our bank going bust us, but sadly we’ve been unable to find the funding we need to move to the next level.

Most of all I’d like to thank the incredible team we’ve built. They’ve stayed dedicated to the challenge and built incredible technologies in the face of abject uncertainty.

It’s incumbent on me to help these great people find new roles, spread their wings, and share their talents with others.

Attached is a list of the folks who are available and their preferred contact details.

Please feel free to reach out to anybody you think you need or could help find new roles.

I’m also on hand to help in any way I can.

Thanks, Barney

While it’s a bummer Karakuri couldn’t survive, it’s not surprising. Food robotic startups suffer from several disadvantages, including incredibly long development cycles and being capital-intensive.

Ex-Picnic CEO Clayton Wood summed it up well in a Linkedin post where he explained that a food automation startup’s “existential risk is being successful enough at the seed stage and building momentum (and costs) toward your scaling stage, only to find no Series A/B/C investors. Without planning and execution, you will be unable to survive. Progress means spending–and cutting spending to stay alive eliminates progress.”

Clayton says he believes newer startups will benefit from an earlier recognition that they need be frugal from the outset, unlike many of the first-generation food robot startups who launched in what was a more friendly fundraising era.

I also expect more food robot startups will start to look to commercialize a product or a subsystem more quickly in order to get to positive revenue faster. As I wrote a few months ago after our food robotics mini-summit, investors like Buck Jordan see a path to revenue through offering a portion of a founder’s big idea to the market instead of waiting years until the full vision is realized.

“I suspect that some robotics companies who are a little more responsible, or a little more revenue-oriented, are going to start paring down their objectives,” said Jordan.

Jordan pointed to Creator, a maker of fully roboticized restaurants, as an example of a company he believes has valuable technology that could be ‘parted out’ to the market and be successful.

Make that had valuable technology. Creator didn’t ever sell a portion of its systems and instead tried to make a full robotic restaurant. The company shut down in March.

June 19, 2023

Podcast: How One Operator is Reinventing His Restaurant With Technology

When Andrew Simmons decided to buy a restaurant in January 2020, little did the long-time entrepreneur know that in just a few months, he would be forced to close his doors due to COVID. 

But instead of giving up, he knew he had to get creative to survive. Survive he did, and when he reopened his doors, he kept tinkering, trying to figure out how new technology could make his restaurant more efficient. 

Andrew’s been an open book during the process, open-sourcing his learning as he navigates his journey via posts on Linkedin and a blog. He shares what works and what doesn’t, providing a potential blueprint for other operators thinking about how technology could change their business. 

During this podcast, Andrew and Mike talk about:

  • How the installation of a pizza robot from Picnic completely changed how he does business
  • How one piece of game-changing technology, like a pizza robot, forces other changes and adoption of new technology throughout the restaurant’s workflow
  • The impact of new technology on his unit price for pizzas 
  • How analytics software helped him realize his dine-in business was not profitable and how it changed his thinking about how he ran his restaurant
  • How he was forced to rethink how he used employees through the use of technology and how the employees (and former employees) have reacted
  • His pizza subscription concept and how he believes it can help him pay for opening new restaurants
  • Andrew’s plans to launch a 100-unit restaurant chain built using off-the-shelf restaurant technology

If you are considering using technology such as robotics for your restaurant, this episode is a must-listen! You can listen to the conversation on Apple Podcasts, Spotify or click play below.

June 8, 2023

Sweetgreen: We’re Going Full Robot

Sweetgreen CEO Jonathan Neman told investors this week at William Blair’s Growth Stock conference that the company plans to have its Infinite Kitchen robotic makeline installed at all new locations within a five-year timeframe. The proclamation marks the first time the fast-casual salad chain has indicated they plan to deploy automation across all of its new locations.

According to the analyst’s note about Neman’s disclosures at the conference, the early results from the Infinite Kitchen have been extremely positive:

Sweetgreen’s first Infinite Kitchen (robotic assembly makeline) has proven that the technology works, with zero downtime so far and significant benefits related to faster throughput and improved portioning and order accuracy (order portioning and accuracy have historically been the company’s weakest links in NPS). With less labor intensity and greater job satisfaction so far, we believe the Infinite Kitchen has the potential to structurally change sweetgreen’s labor model, as roughly half of variable labor stems from assembly.

Another interesting disclosure from the conference was the success of the company’s membership program. According to William Blair, Sweetgreen’s sweetpass and sweetpass+ “are yielding incremental guest frequency, with gamified and personalized challenges yielding a 15% frequency lift while sweetpass+ ($10 monthly membership offering a $3 daily discount) is exceeding expectations with a significant lift in frequency (management indicated it would be happy with an incremental two visits per member per month) and low churn.”

These results show the company has improved food production, customer experience, and loyalty by deploying new technology. The result is a stronger bottom line, as Blair predicts the company will continue its fairly impressive 20%+ EBITDA margins in established locations and new build locations in the coming years. For a company that has long stated it’s as much a technology company as a restaurant company, it looks like its approach has continued to yield dividends and keep Sweetgreen one of the most interesting chains in the restaurant business.

June 7, 2023

Robot Restaurant Concepts Have Struggled. Will Kernel Buck The Trend?

Last month, the NY Post revealed that Chipotle founder Steve Ells is plotting a comeback with an automation-heavy restaurant concept called Kernel.

According to the Post, who viewed the startup’s pitch deck, Ells plans to build a chain of restaurants that makes the food centrally in a ghost kitchen and finishes the food in small-footprint retail storefronts. The centralized production facility and the restaurants will feature a significant amount of robotics and automation to produce the food and keep the overall employee count down. According to materials reviewed by the Post, the company states that they believe “a 3-person labor model can work.”

The company has already started building its production facility in NY and plans to launch its first restaurant in NYC in the fall of 2023. Ells is funding the company for now but is looking for investors (hence the investor deck).

As I wrote yesterday, robot-restaurant concepts have often struggled. But given Ells’ experience building a hugely successful restaurant chain, it’s worth asking: Will Kernel buck the trend?

I think they just might. Here are a few reasons Kernel might have a chance at succeeding where others have failed:

Ells is a Proven Restaurant Operator

Unlike the founders of Eatsa, Spyce and Zume, Ells is a restaurant operator with a proven track record of building a restaurant brand from the ground up. During his tenure as the company’s CEO, Chipotle pioneered the fast-casual concept and had one of the most successful IPOs ever for a restaurant chain. Clearly, he knows something about building restaurant concepts.

Ells Has An Intuitive Understanding of Food Unit Economics

During the early days of Chipotle, Ells focused heavily on the unit economics of the burrito business, calculating just how many he needed to sell daily to make a profit. It’s that kind of focus on the different cost-levers that is required when forecasting the cost of building an automation-centric restaurant business that will have higher up-front capex costs but, over time, should ultimately create more efficient restaurants with lower daily operating costs, and isn’t susceptible to the high employee turnover of employees.

The Hub & Spoke Model Can Work If Done Right

Some ghost kitchens have struggled for various reasons ranging from low-quality and high costs. However, fast casual brands have shown to be a logical pairing with centralized commissary kitchens, especially in high-rent markets like NYC (where Kernel plans to open multiple locations). Chains like Fresh&Co have grown fresh-forward concepts across metro areas like NYC through centralized batch cooking of ingredients and doing final-prep in smaller footprint storefronts. If Ells can leverage automation to take on the majority of food prep and save capital to invest in a greater number of smaller stores, he might prove this model as a recipe for the future.

Don’t Build It, Rent It

While many of the early venture-funded robotic food restaurant concepts spent most of their capital building out proprietary platforms, nowadays, a restaurant builder can leverage any of the available platforms to deploy in their food production and food service workflows. One has to look no further than a small operator like Andrew Simmons to see that restaurants can be built by piecing together systems that use robotics-as-a-service payment structures, which lower the overall capex required and allows flexibility to create a workflow over time as needs change and lessons are learned. My guess is Ells is planning on leveraging systems where others have paid all the upfront cost of development, and he can be a customer who benefits from a service and maintenance agreement.

There are still many unknowns about the Kernel concept, including what automation platforms they plan to use (or create) and just what the consumer experience will be like. But if Ells shows the brand-building prowess from his Chipotle days wasn’t a fluke and can be flexible in architecting a production workflow that carries a lower upfront capex hit than early robot restaurant efforts, he may be on his way to building one of the first true robot-powered restaurant chains.

June 6, 2023

A Food Tech Case Study: Four Lessons From the Demise of Zume

Late last week, news broke that Zume, the company famous mostly for raising a whole bunch of money for its pizza robot & cook-on-the-road food trucks concept, had shut down. The company’s demise, first reported by The Information, comes after burning through $450 million and a well-documented pivot away from its pizza robot and delivery technology products to sustainable packaging in 2020.

Here at The Spoon, we’ve followed Zume since its inception and even had its CTO speak at our event just months before the company laid off 500 employees and dropped its automation business.

When Zume inevitably shows up in business school case studies in years to come, here are a few lessons we can extract from the company’s journey:

Startups Should Pick One Thing To Be Good At

You always hear startup founders talk about how important it is to focus, partly because their first product needs to be really good, but also because distractions can keep a team from executing the way they need to execute. Zume was trying to reinvent both food making and food delivery, which meant they were basically running two highly capital-intensive startups in one. They were early enough at pizza automation to have a shot at success, but then they made their job infinitely more complicated by also trying to create an entirely new kind of delivery truck complete with built-in ovens. That’s a lot of time and capital to achieve marginally fresher pizza.

Creating Custom-Designed Delivery Fleets (with Built-in Cooking!) is an Expensive Fool’s Errand

We don’t just have Zume to prove this, but also Wonder, a company that decided they would differentiate by creating an entirely custom-built delivery fleet in which the food was cooked after it left the kitchen. Both of these companies burned through hundreds of millions of dollars and, in the end, realized that they probably should not have invested all that investor capital in something other companies are pretty good at (delivery) just to make sure the food was maybe a little fresher by the time it got to the customer’s door.

Pivots Should be Somewhat Adjacent to Core IP

After spending years and hundreds of millions of dollars to build a food automation and delivery tech business, the company pivoted to a completely different business in sustainable packaging. The company, which had developed a fairly interesting realtime delivery intelligence platform in addition to a food automation platform, likely could have pivoted to a less-capital intensive business in either of these areas and continued to maintain momentum. Instead, they started over in a fairly crowded vertical and never got enough traction to make a go of it.

Tech Companies Rarely Succeed in Creating Customer-Facing Restaurant Businesses

Time and time again, we’ve watched as startups try to build the “restaurant of the future” and fail because building a consumer brand requires focus and capital, capital they don’t have because they’ve spent it all developing technology. Zume, Eatsa, Wonder, and others have shown it’s probably best to choose between being a technology company or a restaurant business, but probably not both. Some may point to Sweetgreen as something of an exception, but even they had to acquire a company in Spyce which had tried and failed to build a consumer-facing brand for their own restaurant robot technology.

May 30, 2023

Serve Robotics Strikes Deal With Uber to Scale Up to Two Thousand Sidewalk Delivery Robots

Uber Technologies Inc. is gearing up for a robotic future, bolstering its partnership with Serve Robotics Inc., a maker of sidewalk delivery robots. Following a successful trial in Los Angeles, the companies announced they would deploy up to 2,000 of Serve’s delivery robots in multiple markets across the United States. The deal marks one of the largest robotic delivery fleet deployments ever announced.

According to Serve, the expansion follows a year of strong growth, which saw the startup’s robotic deliveries grow 30% month over month since their introduction in 2022. Serve says that they currently serve over 200 Los Angeles restaurants.

“We are thrilled to be growing our partnership with Uber,” said Dr. Ali Kashani, co-founder and CEO of Serve Robotics. “This partnership is a major step towards mass commercialization of robotics for autonomous delivery, and it is a testament to the success of our partnership. We are excited to continue our work with Uber to bring this innovative technology to more cities across the country.”

For Serve, fleet expansion is made easier by the higher utilization of their robots in the field resulting from their fleets serving more than one customer in a given market. In the Los Angeles market, for example, the company’s fleet of about 100 robots delivers for both UberEats and 7-Eleven.

According to the company, they are eyeing San Jose, Dallas and Vancouver for possible expansion and have also started working with Pizza Hut in Vancouver and Walmart in Arkansas.

Long term, it will be interesting to see how cities begin to accommodate growing sidewalk robot traffic. Some cities have banned them, while others have begun to classify them as pedestrians.

May 22, 2023

Forget Getting Carded. Amazon Wants to Scan Your Palm To Make Sure You’re Of Age

Attention all baby-faced twenty-somethings who regularly get carded buying drinks: You may soon face a new high-tech twist on the age-old right of passage in the form of palm-scanning.

At least, that’s if Amazon has its way. According to the company, they’ve just added age verification to their Amazon One palm bio-authentication platform. According to a blog post about the new capability, customers enrolled in Amazon One can use age verification by uploading a photo of the front and back of their government-issued ID, such as their driver’s license, and a selfie. Once signed up, customers will no longer have to show ID when buying drinks at participating outlets.

When in a store or drinking establishment, customers can show they are of age by hovering their palm over an Amazon One device. A “21+” message and the customer-uploaded selfie will appear on the screen. From there, the bartender confirms the photo on their screen is the same as the customer’s and proceeds with the sale.

Bars and their customers might see the benefit of using Amazon’s tech-forward solution, especially if it can mean shorter lines. Bars might also save employee resources otherwise dedicated to carding customers. That said, I’m not sure your local watering hole is ready to install a palm scanner just yet.

While some may also be leery of putting too much of their personal data into the hands of big corporate tech, my guess is digital natives who are ok with sharing info online might prefer the convenience of using bio-authentication methods like palm print authentication. Amazon is also quick to point out that they don’t store users’ driver’s licenses in their system, and the IDs are verified by an ISO 27001–certified identity verification provider (ISO 27001 is an international standard for information security). However, they do store a copy of your palm information, which may still be too much for privacy-concerned customers freaked out about bio-authentication.

For those who are interested in getting palmed, you can check it out next time you catch a Rockies game, as Amazon has installed the age verification-enabled Amazon One In Coors Field at the SandLot Brewery and the Silver Bullet Bar.

May 19, 2023

Is Jeff Bezos Eyeing The Buildout of an Underground Delivery Network?

Today, Wendy’s announced they will trial an underground delivery system later this year in partnership with Pipedream Labs. According to the announcement, the system will deliver orders to customers via a carside pick-up portal using “autonomous robots” that traverse an underground pipe system.

Spoon readers might remember Pipedream Labs as the company with big plans to build an underground delivery network of pipes around cities to shuttle food or other items all the way to the home. The company is working with Wendy’s and other restaurants in the near term – you gotta pay the bills after all – but still has hopes to build the bigger vision of a citywide underground delivery network.

In fact, in a recent Twitter thread, Pipedream CTO Canon Reeves said the company is now courting master-planned community builders with a system that would deliver into the home.

According to Reeves, the Home Portal system would look something like this:

Pipedream Labs Home Portal. Photo: Canon Reeves

And the delivery robots look like this:

Pipedream Delivery Robot. Image Credit: Canon Reeves

Building these systems into new master-planned communities makes lots of sense for a couple of reasons, the first of which is retrofitting existing homes for underground to in-home delivery would be extremely hard and very expensive. Master-planned communities present greenfield build opportunities for concepts like this, where customers can be presented with the option as a feature in a new home, and the cost of the home system can be rolled into a mortgage. Home builders can also build out the delivery infrastructure as they lay down other infrastructure, either going underground or along the community right-of-way areas (as they did in Atlanta in a public right-of-way).

But even if the company just focuses on new build opportunities, the idea is still a little far-fetched, the kind of far-fetched where you almost need a utopia-curious billionaire who invests in crazy ideas to get behind something like this.

Someone like, I don’t know, Jeff Bezos:

Jeff Bezos watching a demo of Pipedream Labs Home Portal. Image Credit: Canon Reeves

According to Reeves, Bezos stopped by last month to check out the home delivery prototype. And while Reeves didn’t say anything beyond that – like Bezos is interested in investing in the system – one could speculate that the guy who founded the biggest online ordering marketplace in the US might just be curious about what a future with an underground delivery network might look like.

Could he be there on behalf of Amazon? Maybe. It’s not like Amazon doesn’t invest in delivery infrastructure, and, in fact, the company invested around $40 billion from 2014-2020 and continues to do so. And, let’s not forget, Amazon itself has explored the idea of underground delivery before and was granted a patent for the idea in 2017.

And even if this isn’t an Amazon thing, but a billionaire-investor-Jeff-Bezos-thing, Bezos has shown a penchant for investing in big ideas like space flight, and if Elon can build underground tunnels for shuttling people around in Teslas, Bezos would be entirely in his right to think sending items around underground in pipes might have a future.

May 15, 2023

Two Years After Buying Spyce, Sweetgreen Launches Infinite Kitchen Robotic Restaurant

Last week, Sweetgreen opened the company’s first robotic restaurant in Naperville, Ill, a suburb of Chicago.

The new automated restaurant, which the company calls Infinite Kitchen, comes almost two years after the company acquired Spyce Kitchen, a startup building automated robotic makelines.

The Infinite Kitchen name is not new; Spyce first used the name when it launched its second-generation robotic kitchen platform in November 2020 and, like the new Sweetgreen Infinite Kitchen, the system was visually reminiscent of the Creator burger makeline. The system’s conveyor belt runs under ingredient dispensers that drop customized mixes of fresh ingredients into bowls. You can see the Sweetgreen version of the Infinite Kitchen in action below.

In the video and the press release, Sweetgreen takes pains to make clear that while it sees automation as a way to add efficiency to operations and enhance the customer experience, they are not doing away with humans as part of the Sweetgreen experience.

“Every meal begins with human hands,” says the video’s narrator, “from our local farmers to our team members, all there to guide you through the process.”

With the Infinite Kitchen, Sweetgreen has also rethought the customer process flow, integrating digital touchpoints (including self-service kiosks similar to those from Spyce).

From the release:

When visiting the Naperville Sweetgreen restaurant, customers are greeted by the new “host” position which provides a more personalized connection between team members and guests. To order, customers can utilize self-service kiosks, place an order through the mobile app, or order directly from the restaurant’s host. The new restaurant format also brings in a new Tasting Counter, brand-storytelling digital screens and a revamped merchandising strategy for an authentic Sweetgreen experience at every touchpoint. Customers visiting the store will be able to shop exclusive merch with designs inspired by the new store joining the Naperville community.

According to the company, Sweetgreen will open its second Infinite Kitchen location later this year at an existing restaurant, where the company hopes to learn how to integrate and retrofit the new technology into an existing kitchen.

Long term, expect the company to expand the use of automation to most of its locations. Company CEO Jonathan Neman has said that about half of Sweetgreen’s labor is food assembly. “And this Infinite Kitchen takes the majority of that,” Neman said in November.

May 11, 2023

Recipe for Disaster? ChatGPT is Tasked to Create Unique, Tasty Dishes and Fails Miserably

So you think your newfound ability to prompt ChatGPT for AI-generated recipes could result in a culinary masterpiece?

Hold that thought, advises the World of Vegan, a popular wellness website focused on vegan living. The site recently undertook an intriguing experiment powered by generative AI, where they prompted ChatGPT to conjure over a hundred diverse recipes. The group prompted the AI bot to whip up new and innovative recipes for a variety of occasions ranging from date night dishes to brunch and dessert ideas. From there, the site’s chef team tested each recipe to see how they tasted.

The result? Not good.

All this prompting led to what the group described as “hilariously pitiful results.” With many of the recipes, the chef team at World of Vegan spotted ingredient formulations that “would clash right away and where the mishaps would occur.” The team also felt the recipes were largely “deceptive,” seeming ordinary at first glance but often described as “rich” and “decadent” when they were quite the contrary.

“I had a feeling ChatGPT would struggle with recipe development, since developing recipes is such a delicate mixture of fine art and science,” World of Vegan founder and chef Michelle Cehn told The Spoon. “But I was shocked by just how difficult it was to find a single spring recipe written by ChatGPT that worked with a passing grade. This is a crucial warning for both food bloggers seeking shortcuts and home cooks looking for quick recipes. You’ll save yourself a lot of trouble (and wasted time, energy, and money) by bypassing ChatGPT and opting for a trusted blogger’s highly-rated recipe instead.”

Image credit: Erin Wysocarski

One of the biggest fails cooked up by the World of Vegan team was a vegan scalloped potato dish (pictured above), which the recipe’s chef said had an ingredient list and cooking instructions that were out of order. The resulting dish had an off-putting color, a pungent sauce, and tasted bad.

According to World of Vegan, out of the 100 or so recipes the team cooked up, only one – a cauliflower taco dish – resulted in an appetizing result.

Cehn believes the resulting 1% success rate might be due to ChatGPT’s reliance on what is essentially flawed data, namely millions of subpar recipes drawn from the Internet. With this as its foundation, things are destined to go poorly once the bot is tasked to create a unique recipe.

“A human brain can’t access all that information, so people are likely independently (and unintentionally) creating duplicate recipes online. Since ChatGPT must create a truly unique recipe, it has to get a little weird to create one that’s not plagiarized.”

While one might expect a site focused on creating recipes to be skeptical about AI filling its shoes, I don’t doubt the poor results are that far off from what others may find if they conducted a similar experiment. Good recipes often result from lots of experimentation and applied knowledge, something that you don’t get when a bot freewheels up a new dish idea out of thin air.

And while a more specialized AI trained on the compatibility of various culinary ingredients – something akin to a chatbot based on Chef Watson – might yield better results, we don’t have that, at least not yet.

Bottom line: human-powered recipe creators are still necessary…for the time being.

Previous
Next

Primary Sidebar

Footer

  • About
  • Sponsor the Spoon
  • The Spoon Events
  • Spoon Plus

© 2016–2025 The Spoon. All rights reserved.

  • Facebook
  • Instagram
  • LinkedIn
  • RSS
  • Twitter
  • YouTube
 

Loading Comments...