122. Moravec's Paradox for Ag Robotics

“Software is Feeding the World” is a weekly newsletter for Food/AgTech leaders about technology trends.

Greetings from Lake Tahoe! We did get some rain and snow flurries in Lake Tahoe on Saturday, but the afternoon was beautiful and gorgeous!

On to this week’s edition now.

Moravec’s Paradox for Ag Robotics

The rice and the chessboard problem is a well known fable about exponential growth.

What repeated consistent digital doubling means, exactly, is best demonstrated through a story about an emperor and a chessboard. As legend has it, when chess was invented in sixth century India, the inventor was given an audience with the emperor.

When asked to name his prize, the inventor asked for a single grain of rice to be placed on the first square of the chessboard, two to be placed on the second, four on the third, and so on, with the quantity of rice doubling every square.

What most people, including the emperor, fail to realize is that if this pattern continues, by the final square the emperor would owe the inventor eighteen quintillion grains of rice, more rice than has been produced in the history of the world.

Computing has had its own version of exponential growth, also known as Moore’s Law.

Moore's law is the observation that the number of transistors in a dense integrated circuit (IC) doubles about every two years.

The layman’s explanation is that computing speed will double every two years.

Moore’s law has been effective (more or less) for the last 40 years (We haven’t made all the way through the chessboard yet). We have seen tremendous advances in computing power. Increasing computing power is a never ending project.

II.

There are certain types of problems, which have proved especially hard, even though there is a significant talk of robots or AI (Artificial Intelligence) taking over the world to make humans redundant. In spite of many advances in computing power, machine learning, and deep learning, many problems are still quite intractable or have recently become a bit more manageable.

We have not seen broad adoption of general purpose robots like C3PO from Star Wars or Marvin from Hitchhiker's Guide to the Galaxy.

How can we explain the challenges encountered by ML/AI and huge amounts of computing power to solve seemingly simple problems?

One explanation is the Moravec’s Paradox.

Moravec's paradox is the observation by artificial intelligence and robotics researchers that, contrary to traditional assumptions, reasoning requires very little computation, but sensorimotor and perception skills require enormous computational resources. The principle was articulated by Hans Moravec, Rodney Brooks, Marvin Minsky and others in the 1980s.

Moravec wrote in 1988, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility”

If we apply this to agriculture, Moravec’s Paradox says it will be relatively easy for computers to solve problems like make prescriptions, analyze seed performance, but hard to solve problems which require perception and fine motor movements like picking a fruit or removing a weed.

But why does AI struggle with the simple? The explanation behind Moravec’s paradox revolves around evolution, understanding, and perception. (You can watch this video which explains the 5 different levels of difficulty for robotics in the context of Moravec's Paradox)

The skills that we define as ‘simple’ are acquired over years and years of evolution. So, while they may appear simple, it’s only because of thousands of years’ worth of tuning.

Things we consider simple are seeing things, recognizing them, lifting them, and moving them around. The complexity of the simple skills we take for granted is invisible, as we have learnt them through thousands of years.

As Erik Brynjolfsson, Director, Stanford Digital Economy Lab pointed out in his fantastic essay, “The Turing Trap: The Promise & Peril of Human-Like Artificial Intelligence”

Humans have evolved over millions of years to be able to comfort a baby, navigate a cluttered forest, or pluck the ripest blueberry from a bush, tasks that are difficult if not impossible for current machines. But machines excel when it comes to seeing X-rays, etching millions of transistors on a fragment of silicon, or scanning billions of webpages to find the most relevant one.

III.

Why is it difficult to do agriculture robotics?

In edition 119, “Agriculture Robotics is difficult AF” I had given examples of strawberry picking, which is considered a low to medium skill task for humans.

Picking strawberries at scale, and picking them consistently requires practice and skill. You have to judge if the strawberry is at the right ripeness for picking, grab the stem, twist it, and put the strawberry in your basket, without bruising or damaging the strawberry. Human beings are good at following the process to find strawberries with the right ripeness, grab the stem, twist it, and put it in a container.

Try to build a robot which can do the same job with human or better efficiency, and it is difficult AF.

The difficulty is due to Moravec’s Paradox, and the challenge to train your machine learning, and AI models to learn thousands of years of human evolution in a few years, and then apply the same learning to a different crop type.

IV.

In 2022, The Mixing Bowl did a study of the agriculture robotics landscape.

For the purposes of this robotic landscape analysis, we focused on machines that use hardware and software to perceive surroundings, analyze data and take real-time action on information related to an agricultural crop-related function without human intervention.

Source: The Valley of Death report on ag robotics by The Mixing Bowl

As can be seen from the chart above, vision-aided robotic pickers and vision-aided spot sprayers are some of the hardest problems to solve within agriculture robotics.

The Western Growers harvest automation report from 2021, reached similar conclusions.

Current adoption of automation in harvest and harvest related activities is low, but significant advancements are expected in the next 3-5 years. Due to Morevac’s Paradox, harvest technologies are not as far advanced as pre-harvesting, and harvest assist activities as can be seen from the chart below.

Primarily, the study finds that the overall advancement of harvest automation in the fresh produce industry is so far limited, mainly due to the technical difficulties in replicating the human hand to harvest delicate crops.

Western Growers summary from their 2021 report succinctly highlighted the limits imposed by the paradox.

If going past Moravec’s Paradox requires unwinding and learning the evolutionary process, which has happened over thousands of years, do we have a realistic chance of solving these problems in the near future?

Can we replay the evolutionary learning process to teach computers and robots what humans have learnt over thousands of years?

The diagram below is a schematic representation of the modern Homo sapiens, which has happened over 400,000 years. If we assume 400,000 years and 25 years being average for a generation, then we get 400000/25 = 16000 generations of evolution.

If we assume human evolution to have been going on for 4 million years, we end up with 160K generations.

Will we be able to replay 160K generations of evolutionary learning, teach robots, and build algorithms in a short period of time?

VI.

Are there examples where we have shown drastic improvements in our learning in a very short amount of time - maybe exponential growth, similar to Moore’s Law?

There are examples where we (as in humans) have made progress faster than Moore’s Law. For example, the cost per human genome has reduced from $ 100 million to $ 1,000 (and falling) in just 20 years.

Source: The falling cost of sequencing a human genome over two decades. NATIONAL HUMAN GENOME RESEARCH INSTITUTE

Another example is Training Compute (FLOPs) of milestone Machine Learning Systems over time. According to Wikipedia, floating point operations per second (FLOPS, flops or flop/s) is a measure of computer performance, and is useful in fields of scientific computations that require floating-point calculations.

Source: Compute trends across three eras of machine learning

Reference: Parameter, Compute and Data Trends in Machine Learning by Jaime Sevilla, Pablo Villalobos, Juan Felipe Cerón, Matthew Burtell, Lennart Heim, Amogh B. Nanjajjar, Anson Ho, Tamay Besiroglu and Marius Hobbhahn; 2021.

The report's summary shows extremely fast improvements in computing power over the last 10-15 years.

We identify an 18-month doubling time between 1952 and 2010, a 6-month doubling time between 2010 and 2022, and a new trend of large-scale models between late 2015 and 2022, which started 2 to 3 orders of magnitude over the previous trend and displays a 10-month doubling time.

Over the last few years, we are beginning to see AI go past the Moravec’s Parodox. We are beginning to see AI tools like image classification and facial recognition, learnt by a child naturally.

Will we be able to compress human evolutionary learning in a few years with sophisticated machine learning and AI tools?

VII.

Given some of the challenges with the Moravec's Paradox, should robotics, AI, and ML focus exclusively on automation and replacement of human tasks as performed today? It is tempting to think automation is the holy grail to solve problems associated with human labor, and efficiency.

A common fallacy is to assume that all or most productivity-enhancing innovations belong in the first category: automation. However, the second category, augmentation, has been far more important throughout most of the past two centuries. One metric of this is the economic value of an hour of human labor. Its market price as measured by median wages has grown more than ten-fold since 1820. An entrepreneur is willing to pay much more for a worker whose capabilities are amplified by a bulldozer than one who can only work with a shovel, let alone with bare hands. (From Erik Brynjolfsson’s essay)

We have seen the development and adoption of collaborative robots (or cobots) as a way to augment human capabilities rather than to replace them completely.

Augmenting humans with technology opens an endless frontier of new abilities and opportunities. The set of tasks that humans and machines can do together is undoubtedly much larger than those humans can do alone.

Some examples of augmentation in Ag Robotics are new robotic platforms successfully undertaking labor-saving tasks which are not very difficult. For example, the GUSS autonomous sprayer can work in orchards.

The GUSS machine navigates autonomously to adjust spraying based on its ultrasonic sensors.

Other examples include Burro’s smart farming system. The system is described as augmenting human capabilities and working side by side with farm workers.

We've built a smarter farming system using user-friendly, autonomous robots that work side-by-side with farm workers to make agriculture more productive and sustainable.

VIII.

In summary, Moravec’s Paradox is a real barrier to progress in Ag robotics, but we have made tremendous strides in going past it.

Automation of human tasks is important, but given augmentation of human capabilities can create new opportunities and abilities, researchers, and entrepreneurs should constantly look for ways to augment, rather than pure automation.

It is an interesting approach to get past Moravec's Paradox.

In the News

AgTech and Agronomy

Seed specialist KWS recently launched “TraitWay,” a website that provides information about available native traits, in addition to guidance on the process of obtaining a simple license.

Deere issues a satellite communications RFP to enhance satellite connectivity to further connect its fleet of intelligent machines.

Latest research from the University of Vienna. “It is important to characterize the substances released by the plant roots and to decode the interaction with the soil organisms. With the help of complex metabolomics analysis platforms, we can test the messenger substances of the roots and thus their potential to inhibit or prevent the nitrification process.”

A NEW research and development project from the University of Sydney will develop products to help growers map their soil constraints in three dimensions across the paddock. The products will use machine learning to map constraints such as sodicity, pH, salinity and gravel, and determine the depth at which these constraints become limiting and impact plant-available water capacity (PAWC). Once commercially available, the products will help growers and advisers to better predict crop-yield variability, both prior to planting and during the growing season, and make decisions on management options such as inputs and soil amelioration.

Current state of AgTech from AgNEXT conference in St. Louis a. Innovation needs a farmer first approach b. Collaboration is critical to innovation c. Investors are not going anywhere

Robotics and Automation

Grape growers: ultraviolet light therapy for powdery mildew.

Autonomous farm machinery could add up to $ 60 billion of global GDP by the end of the decade

The food and beverage industry in the US installed 25% more robots last year to reach 3,402 units in 2021

Supply Chain

“The agtech startup revolutionizing nitrogen fertilizer production, announced today the close of its Series A investment capital raise at $20 million. This fundraising round was led by Khosla Ventures and Fine Structure Ventures with additional participation from Energy Impact Partners, Lowercarbon Capital, and MCJ Collective. Nitricity electrifies and distributes the production of nitrogen fertilizer. The Nitricity approach uses a new technology for regionalized nutrient production using low-cost solar or wind. This marks a major difference from the existing nitrogen supply chain, which is highly centralized and uses fossil fuels and costly transportation.”

Amvac expands traceability

ULTIMUS is patented software technology that is the backbone for the SIMPAS system but applying it for use in spreader and sprayer applications. (AMVAC)

Digital Twins are virtual representations of living or non-living physical entities. Deployment of sensors that detect biological, chemical, and physical properties of objects in real-time, ensures that the digital counterparts of these measured objects are accurate and ‘live’. In such cyber-physical architectures, changes that occur in the physical system are modifying its virtual twin simultaneously and continuously.

Sustainability

Technology and incentives are essential for water management in cotton. “Variable rate moisture sensors may increase profitability by as much as $100 an acre. “The odd thing is the increase came not from saving water and not from increasing yield but from improved quality. It’s amazing to see that big a response in cotton quality from changing the water a little bit.””

Smallholder

Smallholders produce one-third of the world’s food, less than half of what many headlines claim.

E-Book: Global Perspectives on Agriculture Technology

The 12 conversations explore a wide range of topics like the venture capital model for Agricultural Technology (AgTech), different financing options for AgTech, the nature of innovation, automation, how farming will change in the future, the culture clash between technologists and agriculture industry veterans, history of US and Canadian agriculture, techno-optimism, role of policy and science, mental health issues for farmers, produce e-commerce, challenges with smallholder farming in countries like India, China, Indonesia, Malaysia, Kenya, Nigeria, and Zambia, and the potential of farm improvements to raise the standard of living of millions of farmers in the developing world.

All proceeds from the book (minus taxes, transaction fees etc.) will be donated to a non-profit or charity, working to improve the lives of people working in our food and agriculture systems. I will match the first $ 1001 of sales dollar-for-dollar with my own money.

You can get my ebook for free! Or you can donate an amount of your choice.

Current donation amount $ 2820.

Can you help push this to $ 3,000 by the end of the year?

What do you think?

💗 If you like “Software is feeding the world”, please share with a friend.

🙏 If you don’t mind answering 3 questions anonymously (2 are optional), I would love to get your feedback.

About me

My name is Rhishi Pethe. I lead the product management team at Project Mineral (focused on sustainable agriculture). The views expressed in this newsletter are my personal opinions.

Rhishi Pethe