Close Ad

The Spoon

Daily news and analysis about the food tech revolution

News

From Data-Scraping to Discernment Layer: How NotCo’s Giuseppe AI Has Evolved Over the Past Decade

Almost a decade ago, while others experimenting with AI focused on algorithms for trading, diagnostics, or digital advertising, a company called NotCo was experimenting with AI by the name of Giuseppe to create plant-based foods that could match the taste and texture of their animal-based counterparts.

According to Aadit Patel, SVP of AI Product and Engineering at NotCo, the company’s founders (Patel would join a couple of years after the company was founded in 2015) realized early on that, in order to build an AI model that could help create plant-based products mimicking the taste, texture, and functionality of their animal-based counterparts, they would need a whole lot of data.

The problem was, as a startup, they didn’t have any.

When I asked Patel in a recent interview how the company overcame the infamous “cold start” problem—the challenge many embryonic AI models face before they have built large datasets on which to train—he told me they found the solution in a very public place: the U.S. government’s website.

“In the early days, when we had no money, we literally scraped the USDA website,” said Patel. “If you go to the USDA website, there’s a bunch of free data materials for you to use. And I guess no one had actually joined it together to create a comprehensive dataset… So the first versions of Giuseppe were built on that.”

This cobbled-together dataset formed the foundation for Giuseppe’s recommendations, leading to the creation of products like NotMilk, which uses unexpected combinations like pineapple and cabbage to replicate the taste of dairy milk.

As NotCo grew, so did Giuseppe’s capabilities. New analytical labs in San Francisco and Santiago, Chile, gave the company a wealth of new data on which to train its AI. Over time, the model’s ability to create innovative food products also improved.

One of the biggest hurdles in food development is the fragmented nature of the supply chain. Data is scattered across various entities—ingredient suppliers, flavor houses, manufacturers, and research institutions—each holding critical information that contributes to the success of a product. Over time, the company realized that to create an AI capable of building innovative products, it couldn’t rely solely on NotCo’s datasets. Instead, Giuseppe would need to integrate and analyze data from across this complex web of partners.

“What we’ve done with Giuseppe is figure out a way to incentivize this very fragmented ecosystem,” Patel said.

According to Patel, pulling together these disparate datasets from across the product development and supply chain would result in a more holistic understanding of what is needed for a successful product that is better aligned with market realities.

“We realized that if we just made an AI system that’s specific to CPG, we’d be losing out,” said Patel.

Generative AI and Flavor and Fragrance Development

One recent expansion of Giuseppe’s capabilities has been the exploration of new flavors and fragrances using generative AI. While GenAI models like ChatGPT have become infamous for creating sometimes strange and off-putting combinations when designing recipes and new food product formulations, Patel explained that the company has been able to overcome issues with general LLMs by creating what he calls a discernment layer. This layer filters and evaluates the multitude of generated possibilities, narrowing them down to the most promising candidates.

“Discernment is key because it’s not just about generating ideas; it’s about identifying the ones that are likely to succeed in the real world,” Patel said. “With generative AI, you can prompt it however you want and get an infinite amount of answers. The question is, how do we discern which of these 10,000 ideas are the ones most likely to work in a lab setting, a pilot setting, or beyond?”

The discernment layer works by incorporating additional data points and contextual knowledge into the model. For instance, it might consider a formulation’s scalability, cost-effectiveness, or alignment with consumer preferences. This layer also allows human experts to provide feedback and fine-tune the AI’s outputs, creating a process that combines AI’s creativity with the expertise of flavor and fragrance professionals.

Early tests have shown positive results. When tasked with creating a new flavor, both the AI and the human perfumers receive the same brief. When the results are compared in A/B tests, Patel says the outputs of Giuseppe’s generative AI were indistinguishable from those created by human experts.

“What we’ve built is a system where AI and human expertise complement each other,” said Patel. “This gives us the flexibility to create products that are not just theoretically possible but also market-ready.”

CPG Brands Still Have a Long Way to Go With AI-Enhanced Food Creation

Nearly a decade after building an AI model with scraped data from the USDA website, NotCo has evolved its AI to create new products through a collaborative approach that results in a modern generative AI model incorporating inputs from its partners up and down the food value chain. This collaborative approach is being used for internal product development and third-party CPG partners, many of whom Patel said approached the company after they announced their joint venture with Kraft Heinz.

“Ever since our announcement with Kraft Heinz and signing a joint venture, there’s been a lot of inbound interest from a lot of other large CPGs asking ‘What can you do for us?’ and ‘What is Giuseppe?’ They want to see it.”

When I told Patel I thought that big CPG brands have come a long way over the past twelve months in their embrace and planning for using AI, he slightly disagreed. He said that while there’s a lot of interest, most big brands haven’t actually transformed their business to fully create products with the help of AI.

“I would say there’s strong intent to adopt it, but I think there hasn’t been put forth like a concrete action plan to actually develop the first AI-enabled R&D workforce,” said Patel. “There is room, I think, for new AI tech for formulators, and room for best practices and lessons learned of adopting AI.”

You can watch my full interview with Aadit below.

The NotCo team will be at the Food AI Summit talking about their new efforts using generative AI to develop flavor and fragrance, so make sure to get your tickets here.

NotCo's Aadit Patel Talks About the Evolution The Company's Food AI Giuseppe

Jason Cohen Believes Generative AI-Powered Synthetic Data Will Transform CPG Development

Back in 2007, Jason Cohen was an aspiring political scientist studying in China. As it turned out, locals—and the Chinese government—weren’t too enthusiastic about political science students from America asking lots of questions.

Luckily for Cohen, that initial pushback from Chinese officials was the beginning of a circuitous path that would eventually lead him to tea and, surprisingly, to developing AI tools that help food brands accelerate their path to market. The Spoon recently caught up with Cohen to hear about his journey from the tea markets of Yunnan province to his current role at Simulacra Data.

A Serendipitous Start in the Tea Markets

Shortly after Cohen arrived in China as a young prodigy who had graduated high school early and was sent to study politics, things quickly unraveled.

“Turns out, blonde hair, blue eyes, and bad Chinese don’t really endear you to asking about the government in rural southwestern China,” Cohen said. With his political studies cut short, Cohen was drawn to the local tea markets, where he encountered Ji Hai, a fermentation master at the Communist-era tea conglomerate CNNP. It was here that Cohen’s fascination with tea took root.

“I started hanging out in the tea market, originally out of a mix of interest in practicing Chinese,” he said. “But pretty quickly, I realized there was something more going on here.” This unexpected immersion in tea tasting honed Cohen’s palate and laid the foundation for his future endeavors in understanding consumer preferences.

From there, Cohen went to live at the Makaibari Tea Plantation in India, where he continued to study tea. He then embarked on a long journey from Guangzhou, China, through Tibet and Nepal into India, visiting tea places and picking up odd jobs along the way.

Eventually, Cohen returned to the United States, where he attended Penn State on a political science fellowship. However, as in China, his interest in politics was pushed aside by his passion for tea. “Like everything I touch, it kind of spiraled out of control,” Cohen says, describing how a small research group he started evolved into a full-fledged tea research institute, where he did his studies in sensory science and artificial intelligence. Cohen’s research at the Tea Institute eventually became the basis for his first company, Gastrograph AI.

Gastrograph AI: A Pioneering Venture in Flavor Prediction

In 2011, Cohen took the learnings from the tea institute and used them to found Gastrograph AI. At the time, he thought he could build an AI model to predict consumer preferences based on flavor. Over time, Gastrograph built a proprietary dataset of over 100,000 product evaluations from 35 countries, which Cohen claims allowed the company to accurately forecast which flavors would appeal to specific consumer segments.

“We were building a foundation model for flavor,” Cohen explained.

As CEO, Cohen helped Gastrograph AI secure large CPG brands as customers, where the company’s model helped fine-tune their products to meet the tastes of different demographics. Around this time, Cohen observed that AI researchers began to build large language models using neural networks and deep learning, but he wasn’t yet convinced of the power of generative AI for CPG research.

“I had always been a skeptic of the use of traditional neural networks and deep learning models,” he said. “In consumer research, you deal with small, expensive, and difficult-to-collect data sets. You can’t just throw a deep learning model at it and expect good results.”

The Turning Point

Cohen’s skepticism about generative AI shifted as he observed the rapid advancements in new tools based on LLMs over the past couple of years. One particular tool that caught his eye was Midjourney, the generative AI tool that creates lifelike images with simple prompts.

“The moment that the switch flipped was with the release of MidJourney,” Cohen said. “If you can generate images based on a text prompt, you should be able to do that with tabular business data.”

Once Midjourney led Cohen to reconsider the potential of AI in consumer research, he began to think about how generative AI could enable companies to generate synthetic data for scenarios that would otherwise be too costly or time-consuming to study. “It became very, very clear to me in 2022 that generative AI was going to change what’s possible to achieve in consumer research,” Cohen said.

It wasn’t long after this realization that Cohen stepped back from his role at Gastrograph and founded Simulacra Synthetic Data Studio.

Simulacra: Redefining Consumer Research with Generative AI

According to Cohen, Simulacra uses AI in a significantly different way than what he and his team pioneered at Gastrograph; instead of relying on proprietary data, Simulacra uses a “bring your own data” model. This allows companies to input their existing consumer data into the company’s model, which then uses generative AI to create synthetic data for a wide range of scenarios.

“We built an AI that learns to build a synthetic data generation model on whatever data is uploaded,” Cohen said. He explained that this allows companies to simulate outcomes—from market reactions to new products to optimizing pricing strategies—without extensive market research. “It’s much more mathematically accurate. It’s much more correct for drawing direct statistical inference,” he said.

At the core of Simulacra’s technology is diffusion modeling, which Cohen describes as challenging conventional thinking about AI models. “Synthetic data generation turns a lot of what we think about models on its head,” he said. By treating all variables as both dependent and independent, Simulacra’s AI can create a more holistic and accurate model of consumer behavior.

The Impact of Generative AI on the Food Industry

Cohen believes that generative AI will have a profound impact on the food and consumer goods industries.

“We’ve seen the market fracture, and we’ve seen a greater number of consumer cohorts than there had previously been.”

Cohen believes that in a fast-changing market, traditional market research is often too slow and expensive to keep up with changing consumer preferences. Because of the rising cost of traditional research, companies are forced to rely on smaller studies with less statistical power, making decisions based on incomplete data or gut instinct. Simulacra, Cohen explains, offers companies a way to make data-driven decisions that are both accurate and affordable.

“That’s where Simulacra is really going to make an impact.”

Beyond Digital Twins

According to Cohen, there is a big difference between Simulacra’s approach and traditional digital twin technology. While digital twin technology typically involves creating exact virtual replicas of specific entities or datasets to model and predict behaviors, Simulacra uses survey data—ranging from hundreds to hundreds of thousands of observations—to synthetically generate new data or incorporate new knowledge. He believes this approach allows Simulacra to adjust and predict outcomes with more mathematical accuracy and statistical relevance. Rather than producing textual outputs like those from large language models (LLMs), Simulacra returns quantitative and categorical data that companies can use for rigorous statistical analysis.

Looking Ahead: The Future of AI in Consumer Research

As AI technology evolves, Cohen envisions a future where AI-driven consumer research—including synthetic data—is the norm rather than the exception. He predicts that tools like Simulacra will help companies reduce the high failure rates associated with new product launches by providing more reliable data and insights earlier in the development process.

Despite the transformative potential of this technology, Cohen is quick to dismiss concerns that using AI model and synthetic data will lead to consumer product homogenization.

“The idea that this technology is going to be a convergent force across different product development cycles, I don’t think that’s the case,” he said. Companies will still have different goals, constraints, and consumer segments, leading to diverse outcomes even when using similar technologies.

You can watch Cohen’s full interview below. If you’d like to hear him talk about Simulacra and meet him in person, he will be at the Food AI Summit on September 25th!

The Spoon Talks with Analytical Flavor Systems

Shin Starr Brings Robot-Made Korean BBQ to Bay Area With Olhso Food Trucks

A few years ago, the Bay area was the go-to place if you wanted to eat at a robotic restaurant. Whether your meal of choice is burgers (Creator), bowl food (Mezli or Eatsa), or pizza (Zume), chances are if you wanted food made by a high-tech, venture-funded robot, you were in the right place. Fast-forward to today, however, and all of these companies have shut down.

So, where do you go if you’re looking to get a robot meal in what was once the nascent robot restaurant capital of America? If the founders of Shin Starr have any say in the matter, the answer is you stay where you are.

Based in South Korea, the company is operating two fully automated robotic kitchens housed in their mobile Olhso Korean BBQ & Seafood food trucks in the Bay area. If you’d like to order a meal from Olhso, you can download the app and order a meal delivered to your home or office.

The Birth of an Idea

The inspiration behind Shin Starr stemmed from the co-founders’ desire to blend the tradition of Korean barbecue with robotics. The two founders—Jay Shin, a long-time technology investment portfolio manager,and Kish Shin, who previously managed the US franchises for one of the largest Korean barbecue franchises in the country in Kang Ho Dong Baekjeong—got together in 2019 and started working on building a company that would utilize robotics to reduce labor costs and make authentic Korean cuisine.

“We wanted to create something unique and sustainable,” Jay Shin told The Spoon in a recent interview. “Kish and I saw an opportunity to introduce Korean barbecue to the U.S. market in a way that leveraged robotics to enhance efficiency and scalability.”

The Olhso Korean BBQ & Seafood Food Truck

After five years of development, Shin Starr introduced its first Olhso Korean BBQ & Seafood food truck in May of this year. These trucks, which currently operate in Foster City, are outfitted with robotic woks, known as “auto-woks,” that prepare and assemble meals with minimal human intervention. Shin says the only human employee required is a driver, as the robotic kitchen handles the rest, ensuring that meals are served hot and fresh, timed perfectly through a smartphone app.

“Our trucks are very simple,” says Shin. “There’s only one worker, which is a driver, who is also the store manager. The robotic kitchen does everything else.”

The menu features four dishes, including galbi short ribs and kimchi shrimp with pork belly, with prices ranging from $25 to $55, catering to customers looking for a premium street food experience.

According to Shin, the company has filed for nine patents and has five issued so far.

Moving Beyond Trucks

Shin Starr’s ambitions extend well beyond food trucks. The company plans to open a flagship restaurant in downtown San Mateo by November 2024. This 5,000—to 6,000-square-foot restaurant will incorporate the company’s robotic kitchen technology, particularly for preparing side dishes, while still employing human staff for other aspects of the operation.

In addition, Shin Starr is working on a vending machine version of its robotic kitchen, which it plans to launch by mid-2025. This version will target high-traffic locations such as airports and corporate campuses.

Lego-Block Modularity From The Get-Go

Shin says what sets Shin Starr’s robots apart is their modularity. The company has developed its own robotic refrigeration units, cooking devices, and dispensing systems, which can be assembled to create various kitchen setups. This innovative system has already attracted the attention of major players like Samsung, which has ordered 31 robotic systems for its cafeterias following a successful trial.

“From the beginning, we took a modular approach,” Shin explains. “It’s like Lego blocks—each piece can be integrated to form different kitchen setups, allowing us to adapt to various needs and environments.”

Samsung Takes Notice

Shin Starr’s robotic kitchen technology has caught the attention of major corporations, including Samsung. According to Shin, after a successful trial at Samsung’s headquarters in South Korea, where four robotic systems were initially installed, the tech giant has placed an order for 31 additional units to be deployed across its campuses. These robotic kitchens are designed to operate within corporate cafeterias, preparing meals for Samsung’s workforce.

Market Potential and Future Outlook

With an average ticket price of an eye-popping $90 per order, Shin Starr hopes to position itself as a premium brand in the food robotics space. The company’s focus is on building its own consumer-facing brand, but it is also considering franchising its Olhso trucks, vending machines, and brick-and-mortar restaurants.

If you’re in the Bay area and are craving some good Korean fare, download the app and try the Olhso truck. If you do, let us know what you think.

The Food Tech News Show: Is Smell-O-Vision Finally Ready for Prime Time?

Welcome to the Food Tech News Show!

This week, Mike and Carlos discuss the following stories:

Osmo is trying to digitize the world of smells and use AI to finally help us create Smell-O-Vision

The company’s The Principal Odor Map (POM) created by Osmo’s model, outperformed human panelists in predicting the consensus scent of molecules, marking a significant advancement in olfactory science and demonstrating that AI can predict smells based on molecular structure better than individual human experts in many cases.

How’d You Like a Nice Glass of 2D Printed Oat Milk?

This week, Milkadamia, known for its range of macadamia-based milks, announced its first oat milk. However, this isn’t just any oat milk; the company is introducing Flat Pack oat milk, which are printed sheets of plant-based milk that are designed to be rehydrated in water overnight or blended for an instant beverage.

According to the company, these sheets are created by printing oat milk paste onto flat sheets using a proprietary 2D printing process. Each package contains eight of these lightweight sheets, reducing both packaging and weight.

Anova Informs Community Its App Is Going Subscription, and It’s Not Going Well

Last week, Anova CEO Steve Svajian announced that the company will begin charging a subscription fee for new users of its sous vide circulator app starting August 21st, 2024. However, existing users who have downloaded the app and created an account before this date will not be impacted by the change. These users will be grandfathered into free access to the app’s full features.

Svajian explained that the decision to introduce a subscription fee stems from the fact that “each connected cook costs us money,” a cost that has become significant as the number of connected cooks now numbers in the “hundreds of millions.” The new Anova Sous Vide Subscription will be priced at $1.99 per month or $9.99 per year.

Unsurprisingly, the news has sparked discontent among Anova users. 

We’ve got a new humanoid robot that not only performs kung fu, but will also make a meal.

The S1 robot assistant is claimed to have unparalleled agility, dexterity, and accuracy, which help it perform all kinds of tasks. Launched by Stardust Intelligence, a Chinese company, the robot has a human-like upper body structure mounted on a wheeled base. In a video, the robot can be seen feeding cats and making waffles.

Chick-fil-A Is Launching a Streaming Service. Yes, You Read That Right

Extra sauces and some unscripted content, please. Chick-fil-A — yes, that Chick-fil-A — is looking to launch a streaming platform. The fast food chain has been working with Hollywood production companies and studios to create family-friendly, mostly unscripted original shows. The chicken house is also in talks to license and acquire content, according to a source that’s pitched a project.

You can listen to this episode on Apple Podcasts, Spotify, clicking play below or wherever you get your podcasts.

The Idea of Smell-O-Vision Has Been Around for Over a Century. AI May Finally Make It Work

Since the early 1900s, the entertainment industry has been attempting to pair the experience of smell with video entertainment.

In 1916, the Rivoli Theater in New York City introduced scents into the theater during a movie called The Story of Flowers. In 1933, the Rialto Theater installed an in-theater smell system. Hans Laube developed a technique called Scentovision, which was introduced at the 1939 World’s Fair. A decade ago, Japanese researchers were also exploring “Smell-O-Vision” for home TVs, working on a television that used vaporizing gel pellets and emitted air streams from each corner of the screen into the living room.

However, none of these efforts took off, primarily because they didn’t work very well. These attempts at Smell-O-Vision failed because we’ve never been able to adequately recreate the world’s smells in an accurate or scalable way, largely because we’ve never been able to digitally capture them.

This doesn’t mean the fragrance and scent industry hasn’t been robust and growing, but it’s a very different task to create a singular fragrance for a consumer product than to develop something akin to a “smell printer” that emits scents on command. The latter requires a comprehensive digital understanding of scent molecules, something that has only recently become possible.

The digital understanding of the world of smells has accelerated in recent years, and one company leading the way is Osmo, a startup that has raised $60 million in funding. Osmo is led by Alex Wiltschko, a Harvard-trained, ex-Googler who received his PhD in olfactory neuroscience from Harvard in 2016. Wiltschko, who led a group at Google that spent five years using machine learning to predict how different molecules will smell, founded Osmo in early 2023 with the mission of “digitizing smell to improve the health and well-being of human life” by “building the foundational capabilities to enable computers to do everything our noses can do.”

Osmo employed AI to explore the connection between molecular structure and the perception of smell, demonstrating that a machine can predict scents with remarkable accuracy. They developed a machine-learning model using graph neural networks (GNNs), trained on a dataset of 5,000 known compounds, each labeled with descriptive smells like “fruity” or “floral.” This model was then tested on 400 novel compounds, selected to be structurally distinct from anything previously studied or used in the fragrance industry, to see how well it could predict their scents compared to human panelists.

The model’s capabilities were further challenged in an “adversarial” test, where it had to predict scents for molecules that were structurally similar but smelled different. Osmo’s model correctly predicted scents 50% of the time in this difficult scenario. Additionally, the model was able to generalize well beyond the original training data, assessing other olfactory properties like odor strength across a massive dataset of 500,000 potential scent molecules.

The Principal Odor Map (POM) created by Osmo’s model outperformed human panelists in predicting the consensus scent of molecules, marking a significant advancement in olfactory science and demonstrating that AI can predict smells based on molecular structure better than individual human experts in many cases.

We’ve been able to digitally capture and categorize other sensory categories, such as vision, which has led to massive new industry value creation in robotics and autonomous vehicles. The biggest leaps have been a result of machine learning models, and now we’re seeing another massive leap forward in capabilities and product innovation through the application of generative AI.

One potential application Wiltschko describes is “teleporting scent,” where we’ll be able to capture a smell from one part of the world and digitally transfer it to another. To do this, he envisions a world where a local AI-guided molecular sensor could instantly identify the molecular makeup of any scent. From there, his odor map can create what is essentially a formula ready for teleportation without significant manual intervention by scent experts.

This idea, using AI to recreate scents based on a digital framework quickly, could lay the foundation for what film and TV makers have long dreamed of: creating technology that can recreate odors and smells at scale. In other words, we may finally enter a world where Smell-O-Vision becomes a reality. The potential for video entertainment, virtual reality, and other experiences in food service, travel, and more would no doubt lead to a multitude of new applications, much like we’ve seen over the past couple of decades with advances in computer and machine vision.

Sous Vide Specialist Anova Informs Community Its App Is Going Subscription, and It’s Not Going Well

Last week, Anova CEO Steve Svajian announced that the company will begin charging a subscription fee for new users of its sous vide circulator app starting August 21st, 2024. However, existing users who have downloaded the app and created an account before this date will not be impacted by the change. These users will be grandfathered into free access to the app’s full features.

Svajian explained that the decision to introduce a subscription fee stems from the fact that “each connected cook costs us money,” a cost that has become significant as the number of connected cooks now numbers in the “hundreds of millions.” The new Anova Sous Vide Subscription will be priced at $1.99 per month or $9.99 per year.

As Digital Trends noted, this announcement comes on the heels of Anova’s decision to sunset app connectivity for older Wi-Fi and Bluetooth sous vide circulators.

Unsurprisingly, the news has sparked discontent among Anova users. There are currently 195 comments on the Anova post announcing the new subscription, the majority of which express dissatisfaction, with many users stating, “I’m done with Anova.”

For instance, one user commented:

“I liked the product and bought it for friends and family as a gift. I will no longer be using this product and regret ever supporting this company.”

Another user remarked:

“You must have watched Sonos app troubles and thought, ‘Hold my beer.’ Charging your customers for your inability to innovate is a doozy!”

As a long-time Sonos user, I can relate to the frustration expressed in the Sonos comment, having witnessed how the music streaming hardware pioneer damaged its reputation with a glitchy app. While the Anova app may not be as central to the user experience as the Sonos app (I personally prefer using the on-device controls for the Anova), it highlights how upset customers become when a company alters or disrupts a previously satisfactory experience.

However, it’s important to recognize that smartphones have taught us that connected devices have a limited shelf life. Over time, products age, and companies like Apple, Samsung, and now Sonos and Anova, have made it clear that they can’t support old hardware indefinitely, particularly when maintaining apps incurs ongoing costs related to development, web services, and customer support.

The challenge for companies like Sonos and Anova is that consumers don’t perceive all connected electronics the same way, especially those that were initially free to use and expected to have a long lifespan. We’ve become accustomed to paying substantial sums for our phones and their associated monthly service fees, and despite this investment, most of us have accepted the forced obsolescence model that the smartphone industry has ingrained in us.

In contrast, when it comes to other devices, like connected cooking appliances, we tend to expect them to work indefinitely without additional costs. We assume that this new experience—connected cooking—will continue without requiring us to pay for the same level of service we previously enjoyed for free.

Considering the broader trajectory of Anova and its parent company, Electrolux, this news is not entirely surprising. Electrolux, like many appliance companies, has faced challenges in recent years, including laying off three thousand employees last fall. Despite these difficulties, they have continued to operate Anova as a relatively independent entity. Unlike other major brands that have shuttered their smart kitchen acquisitions, Electrolux appears to be making a concerted effort to keep Anova going in a tough economic environment.

It remains to be seen how this move will affect the brand. The backlash is predictable, but I wonder if the outrage is primarily coming from a vocal minority. I suspect that the “100 million connected cooks” figure is somewhat exaggerated, as Anova claims to have powered over 100 million cooks on its website. I also believe that many of these cooks, like me, are from users who simply plug in the device and use it directly without relying on the app.

Is Amazon Working on Something Big in Virtual Restaurants and Digital Food Halls?

What is Amazon up to now?

That was my first thought upon reading a LinkedIn update from Nick Avedesian, a longtime ghost kitchen and virtual food hall executive (and a speaker at The Spoon’s ghost kitchen virtual event in 2020). According to Avedesian, he has just taken a new position at Amazon titled Senior Program Manager, Industrial Launch & Execution, where he will “be supporting projects across our Fresh Food Productions and Amazon Grocery Logistics initiatives!”

Before Amazon, Avedesian held roles at several startups in the ghost kitchen and virtual restaurant space, most recently as the Head of Growth for Byte Kitchens, a startup that raised $6 million in 2022 to build out a digital food hall business. Before Byte, he was with Local Kitchens, which raised $40 million this June to expand its virtual food hall business. Prior to Local Kitchens, Avedesian was the Head of Development and Operations for DoorDash Kitchens.

So, Avedesian, who has spent the past five years overseeing the physical build-out of kitchen spaces tied to onboarding new restaurant brand partners for virtual food halls, is now going to work for Amazon to support projects across Fresh and grocery logistics. Could this mean Amazon plans to launch its own multi-tenant, multi-brand food halls? And if they do, how will those food hall restaurant brands be presented to customers?

While it’s still too early to be certain, I suspect there’s a strong possibility that where there’s smoke, there’s fire. One potential scenario is that Amazon might be planning a combination of in-venue buildouts for new restaurant menus and food offerings from licensed partners at Amazon Fresh stores—similar to what Wonder is starting to do at Walmart. Given Avedesian’s expertise in building both new and retrofit kitchen facilities for multi-brand food halls, I can easily imagine these kitchens, along with customer-facing food hall offerings, being implemented at Fresh locations and possibly other sites as well.

It’s not as if Amazon is entirely new to the idea of ghost kitchens or virtual restaurants. The company was an early investor in Deliveroo, a ghost kitchen and virtual brand pioneer. They are also an investor in Grubhub and recently announced that Amazon Prime users get free Grubhub+ and can order directly from within the app.

Naturally, one update isn’t a guarantee of a new project within Amazon, but there are other small indicators that something is afoot. One of Avedesian’s new coworkers is Kaitlin Garton, a project launch specialist at Amazon, as well as Gavin Worsdale, a manager for industrial launch and execution for Amazon’s worldwide grocery business division.

Amazon’s grocery business has been scrutinized closely as of late, partly because the company hit the brakes earlier this year on the expansion of its Fresh stores and the use of its Just Walk Out technology. However, the company has signaled that they are not getting out of or downsizing their food initiatives, and now you have to wonder if they see an opportunity in using their own kitchens in a virtual food hall business as a growth driver, both in-store (at Fresh and possibly Amazon Go storefronts) and perhaps in centralized kitchens (such as those that produce food for Whole Foods).

The company also has a robust grocery and food delivery business and could begin offering a variety of new home delivery options featuring meals from chefs or restaurant chains.

Whatever they’re up to, we’ll be keeping an eye on things. I’ve reached out to Avedesian to see if he has any specifics on what he’ll be working on, but given Amazon’s notoriously secretive ways, my guess is he’ll likely decline to comment.

Loch Launches Second-Gen Tiny Dishwasher With the $299 Capsule Solo

Ever since Tetra unexpectedly got people excited about the idea of tiny dishwashers in 2018, the category has had mixed success.

Some, like the Bob, have shipped but, unfortunately, haven’t made it stateside. Others, like the Shabosh, seem to be stuck in development hell, with backers wondering when they will ship. As for the dishwasher that kicked off our obsession with small-footprint countertop dishwashers, it never made it out the door and sold its assets last year.

However, those in the US looking for a tiny dishwasher have had an option: The Capsule dishwasher, which shipped to backers in December 2022 and is now available in the US. Soon, tiny dishwasher users will have another option as Loch Electronics, the Scottish company behind the Capsule, is launching a Kickstarter campaign for its second-generation countertop dishwasher, the Capsule Solo.

According to an announcement sent to The Spoon, the Capsule Solo was “designed in response to customer demand for a more affordable option.” Solo pricing will start at $299 at launch, which is $100 less than the first-generation Capsule dishwasher. This pricing includes a bottom rack with a cutlery basket, a clean water tank, and inlet and outlet hoses.

The Solo looks quite similar in size and form factor to the original Capsule, including the signature tall-standing design (reminiscent of a gaming PC casing) that allows for cleaning a frying pan or other taller dishes. According to the just-launched Kickstarter page, the company plans to ship the Solo to backers starting in December of this year.

And just like the original, users can either plumb the Capsule into their kitchen or simply place the device near the sink, where it will drain when the wash cycle is finished. The Solo is also extremely portable, with a handle to carry it from room to room or out to your RV.

While the company isn’t disclosing any shipment figures for the original Capsule, they did tell us that the original Capsule campaign was the most successful Scottish crowdfunding campaign of all time. It raised £566,605, which is $721 thousand. According to the company, across all platforms, the Capsule raised $1.2 million.

Given the lower price and previous track record for success, it wouldn’t be surprising if the company’s newest tiny dishwasher exceeds the funding totals of the original.

Food AI Bulletin: Google’s Robot Breakthrough & Wendy’s Spanish-Speaking AI Drive-Thru Bot

While it’s mid-summer, and while most of Europe (and a good chunk of the American workforce) is taking some well-deserved time off, the AI news hasn’t slowed down one bit.

This week’s Food AI bulletin has updates on a new Google breakthrough on enabling better contextual understanding of our homes (including our kitchens), how Gemini is powering new features in Google’s smart home products, Wendy’s release of a Spanish-language edition of its AI drive-thru assistant, Amazon’s AI refresh of Just Walk Out, a new AI-powered digital tool called the NOURISH to help those living in food deserts make better food choices, a Danone and Microsoft multiyear deal to upskill employees on AI tools, and a survey that shows South Korean students prefer AI-generated healthy food options over more conventionally developed products.

Here we go:

Google’s New Robot Breakthrough Could Make It Easier to Train Your Robot Butler to Cook or Grab You a Cola

In the past, robots were challenged in doing useful tasks with autonomy, in part because they didn’t generally understand what they were seeing and how it related to a person’s specific living situation, etc.

That’s begun to change in recent years, in part because we’ve seen significant advances in robot navigation as researchers using new tools such as Object Goal Navigation (ObjNav) and Vision Language Navigation (VLN) have allowed robots to understand open commands such as “go to the kitchen.”

More recently, researchers have created systems called Multimodal Instruction Navigation (MIN), which enable robots to understand both verbal and visual instructions simultaneously. For example, a person can show a robot something like a toothbrush and ask it where to return it using both the spoken request and the visual context.

Now, Google researchers have taken things a step further by creating what they call Mobility VLA, a hierarchical Vision-Language-Action (VLA). This is a “navigation policy that combines the environment understanding and common sense reasoning power of long-context VLMs and a robust low-level navigation policy based on topological graphs.”

In other words, showing a robot an exploration video of a given environment will allow it to understand how to navigate an area. According to the researchers, by using a walkthrough video and Mobility VLA, they were able to ask the robot and have it achieve previously infeasible tasks such as “I want to store something out of sight from the public eye. Where should I go?” They also write that they achieved significant advances in how easily users can interact with the robot, giving the example of a user recording a video walkthrough in a home environment with a smartphone and then ask, “Where did I leave my coaster?”

One of the biggest challenges around having robots be useful in a food context is that the act of cooking is complex and requires multiple steps and contextual understanding of a specific cooking space. One could imagine using this type of training framework to enable more complex and useful cooking robots or even personal butlers that will actually be able to do something like fetching you a cold beverage.

You can watch a robot using this new Gemini-enable navigation framework in the video below:

“You’re Food Delivery Is Here”: Google Bringing Gemini Intelligence to Google Home

Speaking of Google, this week, the company announced a new set of features coming to their suite of smart home products that their Gemini model will power. The new features were revealed as part of an announcement about a new version of the company’s smart thermostat and its TV streaming device. According to the company, they are adding Gemini-powered capabilities across a range of products, including their Nest security cameras and its smart voice assistant, Google Home.

By underpinning its Nest camera products with Gemini, the company says its Nest Cams will go from “understanding a narrow set of specific things (i.e., motion, people, packages, etc.) to being able to more broadly understand what it sees and hears, and then surface what’s most important.” Google says that this will mean that you can ask your Google Home app questions like “Did I leave my bikes in the driveway?” and “Is my food delivery at the front door?”

During a presentation to The Verge, Google Home head of product Anish Kattukaran showed an example of a video of a grocery delivery driver which was accompanied by an alert powered by Gemini:

“A young person in casual clothing, standing next to a parked black SUV. They are carrying grocery bags. The car is partially in the garage and the area appears peaceful.”

After what’s been a somewhat moribund period of feature-set innovation for smart homes over the past couple of years, both Google and Amazon are now tapping into generative AI to create new capabilities that I’m actually looking forward to. By empowering their existing smart home products like cameras and their smart home assistants with generative AI models, we are finally starting to seeing leaps in useful functionality that are bringing the smart home closer to the futuristic promise we’ve been imagining for the last decade.

Wendy’s Pilots Spanish-Language Drive-Thru AI Voice Assistant

This week, Wendy’s showed off its new Spanish-language capabilities for its Fresh AI drive-thru voice assistant according to announcement sent to The Spoon. The new assistant, which can be seen in the Wendy’ s-provided b-reel below, has a conversant AI bot that seamlessly switches to Spanish, clarifies the order, and upsells the meal.

Wendy's Demos Fresh AI Drive-Thru in Espanol

According to Wendy’s, the company launched its Fresh AI in December of last year and has expanded it to 28 locations across two states.

This news comes just a week after Yum! Brands announced plans to expand Voice AI technology to hundreds of Taco Bell drive-thrus in the U.S. by the end of 2024, with future global implementation across KFC, Pizza Hut, and Taco Bell. Currently, in over 100 Taco Bell locations, the company believes the technology will enhance operations, improve order accuracy, and reduce wait times.

Amazon Previews New Generative AI-Powered Just Walk Out

Last week, Amazon gave a sneak peek at the new AI model that powers its Just Walk Out platform.

In a post written by Jon Jenkins, the VP of Just Walk Out (and, as Spoon readers may remember, the former founder of Meld and head of engineering for the Hestan Cue), we get a peek at the new AI model from Amazon. Jenkins writes the new technology is a “multi-modal foundation model for physical stores is a significant advancement in the evolution of checkout-free shopping.” He says the new model will increase the accuracy of Just Walk Out technology “even in complex shopping scenarios with variables such as camera obstructions, lighting conditions, and the behavior of other shoppers while allowing us to simplify the system.”

The new system differs from the previous system in that it analyzes data from multiple sources—cameras, weight sensors, and other data—simultaneously rather than sequentially. It also uses “continuous self-learning and transformer technology, a type of neural network architecture that transforms inputs (sensor data, in the case of Just Walk Out) into outputs (receipts for checkout-free shopping).”

Academic Researchers Creating AI Tool to Help Americans Living in Food Deserts Access Better Food Options

A team of researchers led by the University of Kansas and the University of California-San Francisco is tackling the issue of food deserts in the U.S. with an AI-powered digital tool called the NOURISH platform. According to an announcement released this week about the initiative, the group is supported by a $5 million grant from the National Science Foundation’s Convergence Accelerator program and the U.S. Department of Agriculture. The project aims to provide fresh and nutritious food options to the estimated 24 million Americans living in areas with limited access to healthy food. The platform will utilize geospatial analyses and AI to identify optimal locations for new fresh food businesses, linking entrepreneurs with local providers and creating dynamic, interactive maps accessible via mobile devices in multiple languages.

Danone Announces Multiyear Partnership with Microsoft for AI

An interesting deal focused on bringing AI training to a large CPG brand’s workforce:

Danone has announced a multi-year collaboration with Microsoft to integrate artificial intelligence (AI) across its operations, including creating a ‘Danone Microsoft AI Academy.’ This initiative aims to upskill and reskill around 100,000 Danone employees, building on Danone’s existing ‘DanSkills’ program. Through the AI Academy, Danone plans to enhance AI literacy and expertise throughout the organization, offering tailored learning opportunities to ensure comprehensive training coverage. The partnership will initially focus on developing an AI-enabled supply chain to improve operational efficiency through predictive forecasting and real-time adjustments. Juergen Esser, Danone’s Deputy CEO, emphasized that collaboration is not just about technology but also about fostering a culture of continuous learning and innovation. Microsoft’s Hayete Gallot highlighted the significance of AI in transforming Danone’s operations and the broader industry, aiming to empower Danone’s workforce to thrive in an AI-driven economy.

My main critique of a deal like this is that it essentially brings training and curriculum to train employees from an AI platform provider with skin in the game in Microsoft. As someone who’s long weaned myself off of most of Microsoft’s software products, I’d hate to go into a curriculum that will mostly be largely Microsoft AI tools training, not really broader AI training.

It is a good deal for Microsoft, with a smart focus on upskilling by Danone. Let’s hope Microsoft’s training brings a broad-based AI tool belt to the Danone workforce that is not entirely walled-gardened within Microsoft’s products.

Survey: Korean Students Prefer AI-Driven Health Foods

While some Americans are becoming more concerned about AI’s impact on our lives, it appears that at least some South Korean students are embracing AI in the development of healthier food options.

According to a recent survey conducted by Korea University Business School, young South Koreans are more likely to trust and purchase healthy functional foods (HFF) developed using artificial intelligence (AI) than those created through traditional methods. The study involved 300 participants and revealed that AI-developed HFFs scored higher in trustworthiness, perceived expertise, positive attitude, and purchase intention. The AI model, NaturaPredicta™, uses natural language processing to analyze botanical ingredients, significantly reducing the time and cost required for new product development. However, researchers noted the potential bias due to the relatively young demographic of the participants and suggested broader studies for more representative results.

Food Tech News Show: Blackbird Launches Pay, DoorDash Delivers Big Numbers

This week on the Food Tech News Show, we are joined by Kristen Hawley, long-time restaurant tech journalist who writes for Fast Company, Eater and her own site, Expedite.

The Food Tech News Show

Blackbird Launches Blackbird Pay

This week, Blackbird, founded by Resy and Eater’s Ben Leventhal, launched Blackbird Pay. Blackbird is a blockchain-based guest loyalty and payments platform designed to better connect restaurants and diners. Diners can use the app to pay at participating restaurants with stored credit card information or loyalty points, and Blackbird charges a 2% processing fee per transaction, lower than standard restaurant POS systems.

Andrew Simmons Ends Pizza Automation Experiment

Andrew Simmons announced the end of his Pizza Roboto project after investing over $300K. His post on Linkedin details the challenges he faced around spinning up new locations and the mixed success of his pizza subscription plan. This news coincides with RestoGPT launching a pizza subscription program based on Simmons’ plan.

DoorDash Sees Digital Orders Surge Despite Industry Headwinds

Amid financial pressures, DoorDash reported a 19% year-over-year increase in total orders and a 23% rise in revenue in its Q2 2024 earnings report. The company noted that digital ordering remains resilient compared to declines in other restaurant industry channels.

Yum Brands to Expand Voice AI Across Taco Bell Locations

Yum! Brands announced plans to expand Voice AI technology to hundreds of Taco Bell drive-thrus in the U.S. by the end of 2024, with future global implementation across KFC, Pizza Hut, and Taco Bell. Currently in over 100 Taco Bell locations, the company believes the technology will enhance operations, improve order accuracy, and reduce wait times.

Amazon Unveils New Just Walk Out AI Model

Amazon revealed details of its new AI model powering the Just Walk Out platform, despite removing the feature from Fresh stores earlier this year, indicating the platform’s continued development and potential.

You can listen to the full episode on Apple Podcasts, Spotify,, or by clicking play below. Don’t forget to like, rate, and review!

Amazon Gives a Peek at the New AI Model Powering Just Walk Out Platform

This week, Amazon gave a sneak peek at the new AI model that powers its Just Walk Out platform.

In a post written by Jon Jenkins, the VP of Just Walk Out (and, as Spoon readers may remember, the former founder of Meld and head of engineering for the Hestan Cue), we get a peek at the new AI model from Amazon. Jenkins writes the new technology is a “multi-modal foundation model for physical stores is a significant advancement in the evolution of checkout-free shopping.” He says the new model will increase the accuracy of Just Walk Out technology “even in complex shopping scenarios with variables such as camera obstructions, lighting conditions, and the behavior of other shoppers, while allowing us to simplify the system.”

The previous Just Walk Out AI model was built in 2018 using generative AI and machine learning models available at the time. This technology, while advanced for the time, could only power a sequential processing of key variables – shopper movement and location in the store, what they picked up, and the quantity of each item—each action processed one after another. This sequential processing worked in most simple scenarios, but complex scenarios like multiple shoppers accessing the same area at a given time, would lead to potential errors that would need to get sorted out at checkout.

The new system differs from the previous system in that it analyzes data from multiple sources—cameras, weight sensors, and other data—simultaneously rather than sequentially. It also uses “continuous self-learning and transformer technology, a type of neural network architecture that transforms inputs (sensor data, in the case of Just Walk Out) into outputs (receipts for checkout-free shopping).”

Jenkins writes that the new system will be better at navigating these complex situations that would result in potential errors with the previous system. He detailed a scenario where a shopper picks and puts down multiple varieties of yogurt, and while doing so, another customer reaches for the same item or the freezer door fogs up and obscures the cameras’ view. In this scenario, the new system processes inputs from various sources such as weight sensors on the fridge shelves and continuously learns from these inputs, eventually deciding which are most important in order to accurately sort out who took what.

The post also gave an update on the current installed base of Just Walk Out technology. According to Amazon, Just Walk Out is currently in 170 third-party locations, including airports, stadiums, universities, hospitals, among other locations. The system is installed in the US, UK, Australia, and Canada and the company plans to double the number of third-party stores with the technology in 2024. 

How a Non-Scientist & Former Lobbyist Started a Future Food Company (Podcast)

Paul Shapiro had spent most of his career as a lobbyist, working for organizations like The Humane Society. While he found his work advocating for animal welfare extremely gratifying, around 2016, he started to reevaluate.

He started thinking he could effect change in a way that he couldn’t as a lobbyist by helping to build one of these new technologies that could make animal agriculture—in other words, raising cows, pigs, chickens, and other farm animals for human consumption—obsolete.

So he decided to write a book. Paul had never published a book before, but he had done plenty of writing. Op-eds, columns, and other pieces as a lobbyist. And because he was so early to the topic and pitched a book no one had written before, it didn’t take long before a publisher became interested.

While writing the book Clean Meat—the first to focus on the nascent cultivated meat industry—he spent a lot of time talking to folks on the front lines of the future food industry. It was during this time he had a realization.

“What I learned was these people were mere mortals like myself,” said Shapiro on the latest episode of The Spoon Podcast. “I thought if you were going to start your own company, you had to have some kind of special skillset that commoners like myself lacked. It was at that point that I decided that instead of writing about these people who I thought were going to solve this problem that had been animating my life for the past couple of decades, I would become one myself.”

In this episode of The Spoon Podcast, Shapiro talks about their journey to start The Better Meat Co., his decision to pursue hybrid meat solutions instead of technologies like cultivated meat, his first big win with Purdue, and his company’s eventual transition to developing mycoprotein based products.

You can listen by clicking play below, or you can download it on Apple Podcasts, Spotify, or wherever you get your podcasts.