David Speller (OPTIfarm): "It's all about a hybrid model"

The Animal AgTech Innovation Summit will be live online on October 19-20, 2021 for two days of interactive live-streamed sessions, roundtable discussion groups and 1-1 video meetings.

David Speller, OPTIfarm CEO/Founder, will be one of the distinguished speakers at the event and in this Engormix interview, he touches on some of the topics of his upcoming presentation.

 

We have seen huge disruption to what is normally such a highly planned and forecasted sector. The shift of product flow from food services to retail and the impossible task of packing plants to suddenly adapt to different demands has led to disruption right through the supply chain.

As we control the risks of COVID in many parts of the world, we are still seeing a lasting impact of the removal of cheap, unrestricted global travel. I myself would spend 200 days per year travelling and now very little. Technical experts in their sector have lost the ability to go and visit a situation and give in-person advise, this restricts the ability for a company to spread technical knowledge and makes supporting a product in the field difficult. This is leading to a change in procedures, whether it be deciding the item to purchase, reviewing how it is supported after purchase or looking at how invested technologies can interact with existing teams that are on farm.

Labour availability has also come into the spotlight post-COVID as many countries see a shortage of their traditional labour sources. Restricted travel, quarantine, a reassessment of life goals are all post-COVID realities for the labour force. We definitely see more questions around automation in the supply chain and also how technology investments might attract those staff that are needed. Can new innovations allow a broader selection process for staff, no longer needing years of experience but instead being supported and trained closely on the farms and in the supply chain? Tech solutions can help, this is not just about replacing them all, as there are some things humans definitely do better than technology today.

In general, we see more acceptance of digital solutions that offer reliable remote information gathering and remote interaction with the supply chain.

Simple data capture solutions were in demand before but more so now to get data from as many locations as possible, as cheaply as possible, including those facilities that don’t have hard wired internet is essential. Technology solutions must be fit for purpose across a wide portfolio of scenarios, not just the brand new farm with perfect internet connections.

An ongoing challenge will still be how to engage with on farm staff who are not naturally adopters of early technology innovations.

 

The more controlled the environment, the more able to fully automate. This points towards intensive poultry, pig and aqua as possible first adopters of fully automated systems.

However, for any animal species, I have my doubts as to how successful the fully automated approach may be. In reality, a fully automated farm will move from employing stock persons to employing engineers and programmers to maintain the systems. These may or may not be easier to recruit, train and retain but there is still a need for some humans somewhere and at what point is it easier and more cost effective to have a hybrid model utilising the best of both human and computer/robots.

Strengths of an automated scenario have to be consistency, a more secure investment than transient staff, possible ease of recruiting engineers rather than stock persons.

But the limitations are once we go beyond giving fresh air, feed, water and lighting it is very difficult to teach an algorithm to assess animals and cope with the constant unpredictability of farming animals, there are simply some interpretations a human does that are analogue assessments that are too complex to digitalise fully. To digitalise what I can do within 2 minutes of entering a poultry barn takes thousands of lines of code, where binary judgements are interacting and conflicting with each other, yet as an experienced poultry person, I do it within a couple of minutes. Any automated farming system would also rely 100% on clean data that, from our experience, is very difficult to get from a farm. Even when the hardware is very reliable where it is located in the facility, or the interaction of the animals with it or the failure of the communication infrastructure, means we have to do a lot of pre-screening of data.

Aside from technological challenges a lot of consideration would need to be given to digital security of a fully automated farm. When farming animals, we are dealing with living creatures and any systems disruption, intended or not, can lead to the loss of many lives and this offers its own brand risk to anyone attempting it in the near future.

Finally, a totally automated livestock facility may struggle with consumer acceptance, many consumer decisions are made based on an emotional judgement or desire. If you are farming animals autonomously, would I rather divert my purchase to a meat free product that is made autonomously?

For myself and why I set up OPTIfarm was all about a hybrid model. A human in the loop concept, this allows a farm to get closer to going automated by ensuring continual assessment and optimisation of the production process but ensuring when the artificial intelligence and algorithms are not clear in an evaluation the query is passed to trained humans that can remotely in real time investigate the cause, evaluate what is best for the animal, give advice and support to the farmer if needed and ensure the systems returns to a planned norm swiftly. Once the situation is stabilised again, a hybrid production facility can return to being under the influence of automated controls and assessments.

 

Before we consider the timelines, let’s remember those workers that are good with animals are not naturally good with digitalisation. If I wanted to be a data analyst or computer user why would I opt for a career working on a livestock farm tending to the needs of the animals? And vice versa, why employ a digital-focused person to work on your farms?

This fundamental fact means we will definitely see different timelines for different global regions, different companies and different individual workers.

From our experiences, we must also consider: do the workers see digitalisation as a threat to their jobs and therefore resist the change or something that can make their lives better and therefore something they want? This often needs considering when introducing digitilisation, will workers still get the bonuses if the computers are now delivering the results, not them? Will the new digital world highlight their personal skills shortages and indicate all the areas that they need to improve in? I am sure my new computer-aided car knows all my bad driving habits and it could, if it wanted to, inform the police about me occasionally, not something I like to think about or welcome, but when the navigation system tells me of an accident and sends me a new way home to avoid delays, I really like the smart car.

 

It's all about the potential of AI and robots. Today, in my opinion, some modelling is used to forecast and some new data is being gathered and some early work shows robot interactions can be beneficial but there are still few if any physical tasks done by robots interacting directly with the animals. Robotics does offer new ways to assess the situation, such as the faeces assessments by the Chicken Boy robot and the robots like the Tibot moving amongst the birds offering an ability to effect and enhance activity/behaviour reducing dominance but I don’t particularly see in the short term a robot culling a sick animal or supervising the catching and loading teams at the end of a cycle.

Items such as planning feed orders or predicting weights to the factories will definitely help future efficiency and sustainability but again, I don’t yet see a piece of AI prescribing a medicine to a sick animal, we are still not seeing any advising or giving instructions never mind interacting with robots and physically doing things.

Today we are still building large data sets that can be used for teaching and establishing AI and machine learning models but without an ability to go to the next step and at least give advice, we just keep building more information with no means to monetise it. We must have very large training data sets, however, to ensure we are not writing algorithms based on limited data, this is especially true given all the variability in livestock facilities, nutrition, processing demands, etc.

There is a place and combined with some human skillsets AI and robotics will help drive consistency in performance and enhance our knowledge, which in turn will drive sustainability around productivity and welfare.