The Next Wave: Artificial Intelligence In Alternative Proteins

“Technology is a useful servant but a dangerous master.”

– Christian Lous Lange

Image credit: Yexl - stock.adobe.com

A Glimpse of 2050

A hovercar parks itself outside the house. Lights brighten as the doors to the house open. As I step inside, a sleek humanoid robot, Astra, springs to life.

Astra: Hope you had a great day! What would you like for dinner?

Me: Thanks, Astra! I’m starving. What are my options?

Astra: My scans reveal you have low amino acid levels, which we can address with additional protein. As a result, I suggest two dinner options:

• Meal 1: Hybrid meat with cultivated pork and pumpkin seed protein along with fermented white wine

• Meal 2: Mycelium-fused chicken with cultivated pork cells and fermented ice-cream for dessert

A holographic display of both these meals is available in the kitchen. I’ll prepare the meal once you let me know your preference.

Me: Sounds good! I’ll go with Meal 2.

I take a seat at the dining table while Astra seamlessly glides to the kitchen, maneuvering between countertops and appliances. The kitchen comes to life as Astra’s robotic arms and sensors work in harmony.

Me: So, how’s everything been today?

Astra: Well, we had a breakdown in our internal heating system. I was able to uncover the root cause and fix the issue. All systems are functioning properly now.

Me: Thanks for taking care of it!

Astra: You’re welcome!

Astra completes the finishing touches and serves dinner ­– a new dish prepared to my culinary preferences and nutritional needs.

Me: This looks amazing!

Astra: I'm glad you like it! Please enjoy your meal.

A Future We Can’t Predict

Humans can’t help themselves when it comes to predicting the future. History reveals that these premonitions range from the uncannily prescient to the wildly inaccurate. Once upon a time, the president of the Michigan Savings Bank declared that, “The horse is here to stay, but the automobile is only a novelty, a fad,” while the founder of Digital Equipment Corporation, a tech giant of its time, surmised that, “There is no reason for any individual to have a computer in his home.”

Whether we embrace or dismiss it, a fundamentally transformative technology is impossible to ignore once it appears on the horizon. Artificial Intelligence (AI) is one such development, and it’s likely to shape our future in ways we cannot foresee. AI’s ability to analyze colossal amounts of data at unprecedented speeds, make deductions, generate solutions, and even interact with us represents a paradigm shift. It holds the power to transform industries, economies, and societies akin to some of the greatest inventions in human history. A few notable ones include:

• Printing Press (15^th Century): Facilitated the mass production of books, democratizing knowledge and accelerating the spread of information

• Steam Engine (18^th Century): Powered machinery and transformed industries during the Industrial Revolution

• Electricity (19^th Century): Its invention and distribution fundamentally transformed society, in turn catalyzing other inventions

• Personal Computers (20^th Century): Brought computing power to individuals and small businesses, democratizing access to technology

• Internet (20^th Century): Enabled instant communication, real-time collaboration, and unprecedented access to information

Just as these inventions transformed society, AI is not just disrupting sectors, but going a step further by learning from data patterns, making predictions, and displaying cognitive ability to make decisions. This will reshape industries, solve major challenges facing humanity (e.g. climate change) and bring about a more sustainable future.

From Inception to Intelligence

The evolution of AI did not happen overnight; it is the result of years of incremental progress with the emergence of capabilities spanning several decades, including:

• Big Data: Processing of large datasets in real-time, providing crucial inputs for AI models

• Machine Learning (ML): A rules-based system that involves training machines to learn from data to improve their performance

• Deep Learning: Neural networks, with multiple layers of interconnected nodes, capable of processing vast amounts of data, recognizing intricate patterns, and making highly accurate predictions 

• AI: An evolution of ML capable of learning from data and improving performance over time without human intervention, thus displaying an ability to perform cognitive functions

AI, simply put, is the simulation of human intelligence processes by machines. AI can thus solve complex problems at incredible speeds. It excels in the decision-making process which can be leveraged across industries to diagnose medical conditions, optimize supply chains, even to analyze human behavior and preferences. This has resulted in an ecosystem of new startups that are leveraging AI to be a first mover in devising a wide array of applications across industries. Though nascent, the alternative protein industry is already being impacted.

Artificial Intelligence in Alternative Proteins

AI holds immense potential to revolutionize the alternative protein landscape. Not only can it help in enhancing productivity, but AI can also spur innovation across the value chain. The impact will be pervasive:

• Research & Discovery: AI can shrink the R&D cycle dramatically. It can recommend optimal strains for recombinant protein production, predict outputs based on simulations of experimental variables and improve accuracy of R&D processes. Shiru, a California based company is leveraging Google DeepMind's AlphaFold¹ to discover plant-proteins with specific functionalities via its Flourish platform. AI can also help us design new and improved proteins. Biotech startup Cradle is using generative AI to train models to natively understand molecular structures and design better proteins. AI can be leveraged for virtual human modelling to understand how proteins interact with the human body. The learnings can then be used to create superior proteins with better functionality and digestibility. Talent is starting to flow into the industry with AI and ML experts such as Noa Weiss helping startups use AI to accelerate their R&D in the alternative protein sector.

• Product Formulation: AI can be used to analyze vast datasets of ingredient properties, nutrition, and consumer preferences to optimize formulations for taste, texture, and nutritional content. It can identify unique combinations of ingredients to create novel recipes. It can also recommend substitutes to existing ingredients in formulations to achieve cost efficiencies without compromising on taste and texture. Using AI, scientists have built a tool to predict the odor profile of a molecule based on its structure which can have applications in alternative proteins². NotCo, a Chilean alternative protein company, has identified proteins for milk replacement using ‘Giuseppe’ – its AI platform that identifies food patterns at molecular level to develop combinations of desired flavors and textures. The day is not far when AI will be used to understand the complex interplay between the entire food matrix (and not just a single component) to formulate food that can evoke desired emotions or feelings.

• Process Optimization: AI can provide greater control of process parameters leading to real time data optimization. This will result in enhanced food safety, accuracy, and productivity through precise management and adjustment of key variables. For example, integrating AI into bioreactors can allow to tightly control the entire production process. AI can be used to determine the optimal media and growth environment as well as the right duration for cell growth and differentiation leading to valuable cost savings. Bioreactors can even be replicated virtually, similar to servers provided by Amazon Web Services. AI can also be utilized in other aspects of the production process such as texturization and modelling cooling die parameters for extruders. Recognizing AI’s potential, meat giant Tyson has invested in an AI-enabled robotics startup, Soft Robotics, to boost worker productivity³.

• Supply Chain: AI can serve as a transformative tool to genetically modify crop seeds to increase yields. It can analyze data around weather patterns, soil conditions and plant health to offer guidance on irrigation, fertilization, and post-harvest management. AI can identify cost efficiencies by predicting quality, yield, waste, and throughput loses along the supply chain. This will enable transition to more sustainable practices by optimizing resource usage (e.g. land, water, energy) and minimizing environmental impact. We will see smarter inventory management by better predicting consumer demand. Thus, it is not surprising that McDonald's has embraced AI integration throughout its supply chain to improve its operations⁴.

• Go-to-Market: AI can spot trends to optimize the release of new products to market. It can analyze large data sets to identify the ideal target consumer segment based on consumer behavior, preferences, needs and income levels. AI can improve customer segmentation to enable companies to tailor their marketing message for each segment, thereby raising customer conversion and retention via predictions for each based on previous patterns. By comparing competitor prices and consumer purchase patterns, AI-enabled food manufacturers and retailers can discover opportunities to offer discounts. Nestle is said to be exploring the use of GPT-4 and DALL-E 2 to better market its products⁵.

• Consumer Feedback: Given the ongoing evolution of consumer preferences and demands, AI can monitor social media, online reviews, and purchase patterns to gather insights, which can be utilized to formulate products to meet consumer needs. In addition to gathering direct customer feedback, it will detect underlying information consumers are communicating via pattern analysis to close the ‘say-do’ gap. For example, AI Palette analyzes online data points to provide real-time predictions about consumer preferences and market trends by categorizing the trends as dormant, emerging, growing, mature, declining, and fading.

For all of this to be possible, data is foundational. The more extensive and representative the data set, the better the results. It is worth noting that the alternative protein industry may have some lengths to go before data required for intensive AI models becomes readily available. In addition, the underlying computational models can be resource intensive, requiring substantial energy for their operation. For an industry that is addressing Greenhouse Gas (GHG) emissions, the alternative protein sector must prioritize the use of renewable resources to power these systems.

A Call for Mindful Progress

Powerful as it can be, technology can be problematic if we lose control of it. This is particularly precarious when we are dealing with technology whose intelligence can eclipse the human mind. There must be thoughtful engagement with this technology to ensure proper checks and balances. Potential risks and possible mitigations include:

• Avoiding Biases: AI models are heavily reliant on high quality training data and prone to inherit biases that may lead to errors. Using diverse datasets can avoid these pitfalls, while conducting audits can help identify and fix them.

• Data Security: Sensitive information, such as proprietary recipes and production processes, demand robust data security measures to ensure both protection and efficacy of this information.

• Reliability and Transparency: AI models must be rigorously stress-tested to ensure reliability with fail-safe measures for critical systems. Transparency, where possible, can help build consumer trust and provide inputs to improve these systems.

• Regulatory Compliance: Rapid advancements in AI run the risk of outpacing regulatory frameworks (e.g. food safety and labeling). Startups must adhere to government guidelines and work closely with regulators as the technology evolves.

• Equitable Access: Technology, for all its merits, poses the risk of deepening the digital divide. Governments, research centers, and charitable institutions can play a key role in ensuring small producers in emerging countries have access to AI tools.

The risks associated with AI implementation can be mitigated through ethical practices, robust frameworks, transparent algorithms, and ongoing monitoring and evaluation. This is particularly important for a nascent industry. Before AI applications become more prevalent in alternative proteins, proactive measures should be taken to address risks associated with its development.

The Second Half of the Chessboard

Disruptive technology requires innovative thinking and approaches for its utilization. AI is advancing at an unprecedented rate even by technology standards, doubling at a pace that far exceeds Moore’s law (2 years) or evolutionary biology (2 million years). The concept of stacked exponential acceleration means AI not only benefits from the advancements brought about by Moore’s Law and evolutionary biology, but also contributes to its own acceleration, creating a continuous loop. ‘The second half of the chessboard’⁶ refers to the impact of exponential growth reaching a scale that goes beyond common human understanding. We’re at the cusp of a new era. The AI revolution (in food) is just getting started.

1. Google DeepMind AlphaFold

2. University of Reading, 2023

3. Food Dive, 2021

4. Forbes, 2022

5. Reuters, 2023

6. Wikipedia