Emerging Trends: The Future of Food

Vertical Farms.jpg

With the world population projected to reach 8.5B people in 2030 by the UN from today's population of 7.3B, a crop of foodtech startups and technologies are under development to tackle this immense challenge of feeding such a huge population.  This challenge is further exasperated by the decline of clean accessible water and depletion of nutrients in crop soils global.  Several technology trends are arising to tackle this challenge in different ways.

Indoor Farming

This set of technologies is based on increasing crop density and yield by having perfect control of growing conditions (humidity, lighting, air composition) as well as making use of vertical farming.  The idea, which has yet to be proven economically viable, is to reduce waste of the inputs (i.e. fertilizer, water, energy, etc) while dramatically increasing the speed of crop growth and completely eliminating plant diseases and rot.  More crops more quickly for less cost then current methods is the goal.  The poster child for this set of technologies is startup Plenty which is backed by a number of big name and deep-pocketed investors.  They advertise yields up to 350 times conventional farming techniques with only 1% of the water.

New Methods of Producing Meat

It is well documented that cattle rearing for the world's ever growing hunger for meat products is horribly inefficient and damaging to the environment releasing high amounts of greenhouse gases.  A number of startups have arisen over the last few years seeking to grow artificial meat from cells in a lab environment.  Of these the one that has received the most press and funding is San Francisco based Memphis Meats.  The technology here is basically to use animal stem cells and bioreactors to grow meats in a lab setting that is indistinguishable from the traditionally raised and processed meat we eat daily.  It's a much more complex scientific challenge than it initially sounds.  The advantage is it likely will be much faster, cheaper and efficient pound for pound to produce any type of meat once the technology is developed and scaled.

Farm 2.0 - The Digital Farm

There are a ton of startups focused on improving traditional farming with everything from using drones and software to analyze crops in the field (i.e. Farmers Edge), robotics to replace human workers, software to better manage farm operations (AgCode), and many other areas.  While not as likely as the above two categories to dramatically increase food yields, this is a much faster to implement and less difficult set of technologies to bring to market.  In a word the Digital Farm is a sustaining set of technologies rather than a revolutionary set of technologies.

Investor Discussion Series: Evangelos Simoudis of Synapse Partners

Synapse Partners.png

Evangelos Simoudis is the founder of Synapse Partners, a VC firm that focuses on AI and Big Data investing.

How do you really identify and due diligence AI companies, versus companies just doing statistics or have vaporware?

There is a lot of hype in the space about what is possible with the technology available today.  Anytime there is hype in a sector you get pretenders along with the startups that are developing important IP. 

At Synapse we partner with large corporations in automotive and transportation, financial services, and telecommunications. We work with Sr. Executives from corporations in those industries to understand what they view as strategic problems for their companies and we then determine which of these can be addressed via data and AI.

Since Synapse Partners invests exclusively in early-stage startups developing enterprise applications combining big data with AI, my personal background in AI proves to be very helpful when we consider new investment opportunities.  We also tap into our firm’s advisory board that includes senior AI and data scientists.

 

What kinds of moats do AI companies have from what you’ve seen?  Does it really just come down to the data/all the algos are mostly the same right?

Identifying and gaining access to the right data sets for solving important enterprise problems, selecting the right AI approach to exploit that data, properly preparing the data for processing by the AI system, and finally making sure that the results of this exploitation are correct are the prerequisites for creating such moats.

It is an oversimplification to think that all one needs is data and that more data is always better. The moat is created by the uniqueness of the data and its quality, as well as the ability to exploit  it in a smart way.  It’s a fallacy to think that by taking some open source machine learning software and presenting a large data set means that one can create an important product.  This approach may have been used in the past but the low-hanging fruit has already been plucked. 

For example, properly labeling data before presenting it to machine learning systems turns out to be a difficult and expensive task that today is mostly performed manually. As corporations work with very large data sets, such as those generated by autonomous vehicles, such manual labeling becomes prohibitive. We have invested in a company called Understand.ai that uses AI to automatically annotate that type of data.    

 

What Trends are you currently investing in?

●      For the past couple of years, we have been investing in companies that develop AI-based software to enable autonomous mobility.  More recently, we developed and currently pursue an investment thesis around using big data and AI to monetize autonomy. For example, we are looking at startups developing fleet management and commerce-related AI systems for passenger transportation and logistics where autonomous vehicles will have an advantage. This includes tasks like scheduling, and maintenance of autonomous vehicles aimed at increasing a vehicle’s uptime.   

●      Intelligent software agents (not chatbots) that operate within larger software or hardware systems, for example warehouse robotics. We are interested in systems that understand natural language so that they can collaborate with humans, and can learn from such interactions.

 

What’s Overhyped today from an investment standpoint?

AI, autonomous vehicles, blockchain and cryptocurrencies, augmented and virtual reality are all hot areas but are all overhyped right now in terms of their potential impact and the speed with which this impact will occur. It will take us longer than the popular press talks about to really see the impact from these but I remain optimistic that we will have important changes as a result of using these technologies to address enterprise problems. We will continue to see significant pilots and experiments being done by corporations using these emerging technologies.  However, people need to keep in mind there is a big difference between experimenting with a technology and extracting insights and being able to have broad deployments.

On the other hand, I am very optimistic about the accelerating proliferation of cloud computing in the enterprise. But you can see how long it has taken cloud computing applications to permeate the enterprise and for the market to become as large as it is today. 

What’s the key signal or two you look at when thinking you want to invest in an early enterprise startup, what ultimately convinces you?

I start with the team and I like to see how driven they are for the startup. I pay attention to how complementary the team members are, since I consider well-rounded teams to be an important ingredient to a startup’s success. I don’t want to see a team that consists only of engineers.  I like to understand the team’s background and how they got where they are.  Of course, we always pay attention to the market opportunity, which can often be challenging when the startup is trying to address a brand-new market.  Lastly, at Synapse we always syndicate. It is therefore important to understand who are the other investors.

 

What are some resources you use to stay up to date on a space?

I spend a lot of time reading and working with large corporations. Technology- and startup-related conferences are also important sources of information.  These days I find myself connecting a lot more with PhD and academic colleagues.  The biggest challenge I have is finding good enough filters to discard the overhype of the AI field right now.

 

Any advice to young venture capitalists and angel investors out there in sourcing deals?

It’s not only about writing checks for new investments; it’s about making money for your investors.  Finding a company that wants your money is the easy part, understanding if this is a good company and important investment opportunity to make you want to be part of the company for the next 5-10 years is something you need to focus on. 

Sourcing these days is very difficult. There are so many startups being created and capital is a non-factor.  Having deal flow today is not hard. Having the right deal flow is the important part.  You get the right deal flow by being able to show to the entrepreneur and your co-investors that you have something unique and important to offer.

Today local networks are not as important as they used to be.  You need to be able to tap into global deal flow and be willing to invest globally even as a small investor. 

 

Anything else you think AI  investors or enterprise entrepreneurs should know?

AI is a complex field. It includes several different areas.  Not only machine learning.  People don’t have as much of an understanding of AI as they think.

 

Looker

Looker logo.png

Overview

Looker is a data analytics and business intelligence startup.  Their software is used for exploring data and making decisions for everything from marketing, sales, operations, logistics, web, etc.  It is able to pull and digest data in real time from numerous sources including Salesforce, Slack, SQL, Hadoop, web APIs and produce a number of dashboard visualizations and analytics. 

Looker's business model is a standard SaaS subscription model.

The team is based in the San Francisco bay area.

Why I like Them

Companies currently are flooded with data in the "Big Data Age" but understanding that data and being able to use it to make better decisions is something many firm's struggle with.  Today, to make any sort of headway with their data, specialist data scientists with years of advanced education and expertise in tools like SQL, R, etc. are needed.  However, every job function in a company needs to be able to quickly analyze and make decisions off this data without going through the bottleneck of a limited number of data scientists.  Looker provides this ability to every person in the company without the bottleneck and without needing advanced training to use and analyze this data.  In our data driven age this is a critical ability and it's easy to see Looker becoming a standard tool utilized across every levels of a company just like Microsoft Office Suite.

Despite how crowded the Business Intelligence space is, Looker has a lot of tractionI at over 1,000 paying customers and more than 50,000 users.  These include some large names such as Square, IBM, WeWork, Nordstrom, and Amazon. 

 

Disclosure:  All information is from publicly available sources, I have not had any contact with a member of the company or its investors.

2.png

Some Thoughts on Voice

Voice-Activated-Technology-1024x289.png

Lately, I've been thinking about the rapidly emerging voice technologies, a subset of artificial intelligence and machine learning.  Most of the big technology companies now offer a voice assistant through a mobile phone or smart speaker such as Amazon with Alexa, Apple with Siri, Microsoft with Cortana, etc.  Forecasts predict that 50% of all searches done online will be through voice by 2020 (driven by mobile), with an estimated 13% of US households owning a smart speaker today.

My take is that voice is another interface, not a platform in and of itself or a new paradigm in computing.  Voice is inherently a low bandwidth medium making it strong for certain types of activities, specifically giving commands and getting information to basic questions.  “Alexa play music” or “Google what is the weather going to be today?”.  This is borne out by the plethora of studies being released lately examining how consumers use voice technologies.  The most common uses of voice today by far are asking for directions, asking a quick question, calling someone, checking time, or playing music.  

I expect to see the voice interface dominate when it comes to certain use cases such as home automation but I suspect it’s overhyped for a lot of other common digital use cases.  A big one in this bucket would be online shopping.  Humans are inherently visual creatures and the ability to see a product and read about it will trump being able to purchase purely through a voice dialog.  The only exception to this would be repeat purchases of the same product or if it’s a really basic item that you are agnostic to its brand (“Alexa, order me a stapler”).  Obviously, Amazon disagrees with me having stated it has over 5,000 employees currently working on Alexa and Echo technologies but I remain bearish on voice for eCommerce.

In terms of search, due to its low bandwidth I doubt voice will be very useful for discovery in general or for new customer acquisition by brands and advertisers.  It’s just not a good interface for comparison or discovery.  That of course won’t stop Google and Amazon from offering a new type of ad unit in the very near future where you can pay these companies to be a featured product in a voice search.

Noted technology pundit Scott Galloway for the last year has been predicting the death of brands due to voice but I have to disagree with him here and think he is greatly exaggerating the impact this technology will have.  Voice will simply becoming one of several interfaces users have access to as it seeps into common usage, but it certainly won’t replace touch, keyboards, etc. 

DeepScale

EGuKv9wy_400x400.jpg

Overview

DeepScale AI is a computer vision startup focused on real time automotive perception systems that give cars self driving capabilities.  The team deploys onto cars pre-trained neural networks able to take information from any type of sensor system (including cameras, LiDar, etc.).  They are focused on "efficient" deep learning that is high powered but doesn't rely on expensive and specialized hardware.

DeepScale has a standard software licensing model that charges per sensor and per car to the manufacturer.  Their technology has already been licensed by a number of manufacturers and will be in customer cars in the next few years

The team is around 20 employees and is based in California. 

Why I like Them

What most people don't realize is that existing autonomous vehicles such as Waymo's are massively expensive.  The industry isn't cost-sensitive currently but it will have to become extremely cost conscience as it scales.  The current modus operandi for solving deep learning challenges is to throw more expensive hardware at it.  Automotive manufacturing is a low margin business and would never reach close to profitability with autonomous cars cost structure as they are being developed today.

 I like that DeepScale has thought ahead on this and is focused on creating computer vision systems that work on cars existing electronics (known as ECUs) rather than requiring additional specialized and expensive chips to be installed.  Under the current cost structures of automobile manufacturers, autonomous vehicles will not be able to be sold even close to current car prices, so the team's focus on the business side of the technology is particularly noteworthy.  By using systems already in the latest vehicles, they help automotive manufacturers save on cost by doing more with less.  Think of DeepScale as building efficient automotive AI.  

I also like the competitive nature of the product.  Their systems continue to gather data in the field and are able to be continuously updated with upgraded neural networks pushed to cars remotely that have already been sold.

Disclosure:  I have spoken to members of the team.