What is the difference between vertical farming and machine learning?

What you will learn

  • What is Vertical Farming?
  • Why does this author compare it to machine learning?
  • Is there really a connection between these two things?

Sometimes inspiration comes in the weirdest ways. I love watching CBS News Sunday Morning because of the variety of stories they air. Recently they did one on “vertical farming – a new form of farming” (see video below).

CBS News Sunday Morning recently published an article on vertical farming which gave rise to this article.

For those who haven’t watched the video, vertical farming is basically an indoor farming method using hydroponics. Hydroponics is not new; it is a subset of hydroculture where crops are grown without soil. Instead, plants grow in mineral-enriched water. This can be done in conjunction with sunlight, but usually an artificial light source is used.

The approach is useful in areas that don’t provide enough light, or at times or places where the temperature or outdoor conditions wouldn’t be conducive to plant growth.

Vertical farming is hydroponics taken to the extreme, with stacks upon stacks of trays with plants under an array of lights. Lights these days are usually LEDs because of their efficiency and ability to generate the most useful type of light for plant growth. Automation can be used to streamline planting, support and harvesting.

A building can house a vertical farm anywhere in the world, including in the middle of a city. Although a lot of water is needed, it is recycled, making it more efficient than other forms of agriculture.

Like many technologies, the opportunities are great if you ignore the details. This is where my usual contrary nature came into play, as I pursued my original interest in looking for limitations or issues with vertical farming. Sure, I found quite a few and then noticed that many of the general issues applied to another topic that I cover a lot: machine learning/artificial intelligence (ML/AI).

If you’ve made it this far, you know how I see the difference between machine learning and vertical farming. They obviously have no relationship in terms of technology and implementation, but they have a lot in common when looking at potential problems and solutions related to these technologies.

As designers and developers of electronic systems, we are constantly confronted with potential solutions and their trade-offs. Machine learning is one such generic category that has proven useful in many cases. However, one must be wary of the issues underlying these flashy approaches.

Experts wanted

Vertical farming, like machine learning, is something you can try your hand at. To be successful, however, it helps to have an expert or at least someone who can quickly pick up that experience. This tends to be the case with new and renewed technologies in general. I suspect a lot more ML experts are available these days for a number of reasons such as hardware cost, but demand remains high.

Vertical farming uses a good deal of computer automation. The choice of plants, fertilizers and other aspects of hydropic farming are essential to the success of the farm. Then there is the maintenance aspect. ML-based solutions are a way to reduce the expertise or staff time required to support the system.

ML programmers and developers can also get easier-to-use tools, reducing the amount of expertise and training needed to take advantage of ML solutions. These tools often include their own ML models, which are different from the generated ones.

Profitable for only a few types of plants

Hydroponics works well for many plants, but unfortunately for many others it doesn’t. For example, crops like microgreens work well. However, a cherry or apple tree often struggles with this treatment.

ML suffers from the same problem in that it is not applicable to all computational tasks. But, unlike vertical farms, ML applications and solutions are more diverse. The challenge for developers is to understand where ML is and isn’t applicable. Trying to forcibly adapt a machine learning model to handle a particular problem can result in a solution that provides poor results at a high cost.

High energy consumption

Vertical trusses need energy to light up and move liquid. ML applications tend to be computationally intensive and therefore require a lot of power compared to other computational requirements. A big difference between the two is that ML solutions are scalable and the hardware tradeoffs can be significant.

For example, ML hardware can improve performance by several orders of magnitude over software solutions while reducing power requirements. Likewise, even software-only solutions can be efficient enough to do useful work even using little power, simply because the developers have made the ML models work within the limits of their design. Vertical trusses do not have this flexibility.

High investment and operating costs

Large vertical farms require a major investment and they are not cheap to operate due to their scale. The same goes for cloud-based ML solutions using the latest disaggregated cloud centers. These data centers leverage technologies such as SmartNIC and intelligent storage to use ML models closer to communication and storage than was possible in the past.

The big difference between vertical farming and ML is scalability. It is now practical for several ML models to work in a smartwatch with a dozen sensors. But that doesn’t compare to agriculture that has to adapt to the rest of the demands of the physical world, like the plants themselves.

Yet nowadays ML requires a significant investment in development and development of experience to properly apply ML. Software and hardware vendors have worked hard to reduce start-up and long-term development costs, which have been further increased by the plethora of free software tools and low-cost hardware now generally available.

Technology Failure Can Lead to Major Problems

Turn off the power on a vertical truss and things come to a halt pretty quickly, although it’s not like an airplane losing power at 10,000 feet. Still, plants need food and light, although they are used to changes over time. Nevertheless, responding to system failures is important to the long-term usefulness of the system.

ML applications tend to require electricity to operate, but this tends to be true for the entire system. A more subtle issue with ML applications is the input source, which is usually sensors such as cameras, temperature sensors, etc. Determining whether input data is accurate can be difficult; in many cases, designers simply assume that this information is correct. Applications such as self-driving cars often use redundant and alternate inputs to provide a more robust set of inputs.

Evolution of technology

Vertical farming technology continues to evolve and refine, but it continues to mature. The same goes for machine learning, even if the comparison looks like something between a piggy bank and Fort Knox. There are simply more ML solutions out there, many of which are very mature with millions of practical applications.

That said, ML technologies and applications are so varied, and the rate of change so great, that keeping up to date with what’s available, let alone how things work in detail, can be overwhelming.

Vertical farming benefits from advances in technology, from robotics to sensors to ML. The ability to track plant growth, germination, and pest detection are just a few tasks that apply to all of farming, including vertical farming.


As with many “What’s the difference” articles, the comparisons aren’t necessarily individual, but I hope you picked up something about ML or vertical farms that you were interested in. Many issues are not well mapped, such as pollination issues for vertical farms. Although the release of vertical farms will likely fuel some ML developers, ML is likely to play a bigger role in vertical farming given the level of automation possible with ML sensors, robots, and monitoring now available.

Sherry J. Basler