Artificial intelligence and the future

Publication | April 2018

Introduction

Artificial intelligence (AI) is pushing innovation in new ways and accelerating with advancements in computing power, data and algorithms. AI tools are being used in previously unachievable ways to improve the entire food supply chain. This will become even more important as the world’s growing population creates an increasing demand for agricultural production.

What is AI?

AI is a field of computer science that includes machine learning, natural language/speech processing, expert systems, robotics and machine vision. It enables computers to perform tasks traditionally completed by humans by automating decision making, learning and recognition. In some instances, human subject matter experts provide feedback on results as part of training or testing process and in turn machine learning dynamically updates programming code to improve the algorithms. The methodology used by AI can be represented by various graph and network structures. For example, an artificial neural network or neural net is a system designed to process information by simulating the framework of biological brains to automate tasks.

How can AI apply to agriculture?

AI tools can augment human capabilities to improve agriculture productivity including crop monitoring, robotics and prediction analytics. In the future, we may eat food that has been planted, watered, monitored, picked, sorted and delivered without human intervention.

Machine vision can recognize crop diseases and pest damage. For example, an AI tool created for farmers can automatically detect diseases in cassava plants with 98 percent accuracy1. The neural network that powers this AI tool runs entirely on a smart-phone, without the need for expensive processing resources, making it more accessible to farmers. The tool was built using an open source machine learning library. An AI technique called transfer learning enables a neural network trained to recognize one type of object to learn to recognize other objects with less training data. In this case, transfer learning enabled the AI tool to learn to detect cassava leaf ailments using a small number of high quality leaf images which is important as high quality leaf images may not be abundant. As globalization accelerates, the spread of pests and pathogens making it increasingly important to have tools that can quickly detect disease to contain the problem.

Traditional fruit picking requires precision typically only capable by human hands and eyes which is why most fruit today is hand-picked. This is changing. A robotics company has developed an automated apple picker, to identify an apple, tell if it is ripe and then use a vacuum system to pick the apple without damage. This requires sensors and cameras to collect data to provide automated decision-making in real time to actuate robotic components. Systems are also being developed for other fruit such as oranges and strawberries.

Self-driving vehicles are not limited to streets. There are already self-driving tractors that can automatically navigate a farm on private land without the need for equipment makers to consider complex traffic and regulatory rules. A human driver can teach the farm layout to the tractor system and then take their hands off the steering wheel. There are also self-driving lawn mowers and automated weed whackers. The self-driving vehicles have data collection devices to capture navigation routes along with sensor data relating to internal components and the external environment.

Farming drones collect images and other statistics which an AI tool combines with imagery data, weather, soil, and expert plant pathology to deliver a visual farm dashboard with analytic overlays. Data from different systems can be uploaded to servers for fleet management and precision agriculture services.

AI tools can generate farming forecasts and drive operational decision making. For example, an AI product accurately predicted corn and soybean yields by processing satellite and plant images using a combination of machine learning, meteorological and reported agriculture data. Initial testing of the system was done against years of historical data. The “real world” prediction data shows the estimates shifting over time which highlights the importance of constantly collecting new data and refining the algorithms.

What are some legal considerations?

Intellectual property

Developers of a new AI agriculture tool should consider the following

  • An intellectual property (IP) strategy that layers IP rights to protect different aspects of the innovation.
  • Contributors to the technology should be identified and tracked.
  • Ownership and confidentiality should clearly be set out in a written agreement.
  • Companies should have policies for developers incorporating third-party IP, even if inadvertently, as it may impact ownership of the technology and freedom to operate.
  • Employees or a contracted developer, for example, may incorporate third-party source code without authorization, which may impact ownership and could create inadvertent liability of infringement of other’s IP rights.
  • Contractual terms with end users and third parties should clearly specify permitted use and ownership for collected data.

Copyright is an important IP asset for AI as it protects any new original works which can cover computer program code, application programming interfaces, compilations of data and graphics. This protects the technology product (code) from unauthorized use and reproduction. Digital locks on products and services can protect the code and data. Circumvention of digital locks is an offence in some jurisdictions.

AI systems can also generate new works protectable by copyright, such as creating new artwork or music. However, most copyright statutes do not yet clearly define who owns machine-generated works. It is currently a point of contention in respect of some such works whether the work is generated by a machine, and or the role played by the humans in creation of the work. To this end, agreements should attempt to clarify ownership when possible. Further, an AI system may act or operate autonomously in a manner that infringes third-party IP rights. If existing laws do not extend liability to a machine, then a related stakeholder (such as the owner, developer, operator or another supply chain participant) may be responsible.

A trade mark may consist of a combination of letters, words, sounds or designs that distinguishes one company’s goods or services from those of others in the marketplace. A strong brand helps companies differentiate AI products and services from competitors and establish a strong reputation in the market. Algorithmic accuracy can help a company develop goodwill for its brand. AI companies are often stewards of important data assets and documentation should consider these as valuable assets and document and register marks when possible. A reputable brand may be of paramount importance to customers.

An AI tool can be a “black box” device embedded within a finished product offered by a third party. This can make it difficult for the end customer to recognize the brand of the company supplying the “black box”. A co‑branding agreement can provide for use of the mark associated with the “black box” on the finished product offered by the third party. This can help the “black box” provider become recognizable by the end consumer.

Patents provide an exclusive right to make, use and sell his or her invention, which may help companies, obtain or maintain market share, and protect research and development investments. In contrast with trade secrets, granted patents may be enforced against third parties that make, use or sell the claimed invention, despite independent development. Patents may also be used defensively as a negotiation tool and patent publications can be cited against subsequently filed applications to prevent grant which can protect freedom to operate.

AI involves software which is increasingly difficult to patent and there is no clear delineation of what is patentable and what is not patentable. Highlighting salient technical features such as technical advantages and practical implementation details can increase the likelihood of success during patent examination. The description should highlight physical form factors and discernible effects generated by the AI innovation, such as moving a physical machine to pick up an object. Given the quickly evolving AI market, obtaining early priority dates is important in view of the“first to file” nature of the patent system.

A company making, using or selling AI tools should also consider its freedom to operate to avoid encroaching on existing IP. A landscape assessment and competitor monitoring are helpful to mitigate risk. In the AI context, the legislative protection has not yet advanced as quickly as the technology, which makes early and ongoing IP portfolio management of particular importance. A company may then better control the use of its IP rights, including permitted use under licensing and collaborative arrangements.

Privacy and data protection

AI tools can collect a large amount of data about farm operations for direct delivery to a server managed by the equipment maker. A farmer might not even have access to their own data due to technology protection measures by the equipment maker. Although agricultural information alone might not be considered to be personal data, a large amount of it in combination with different data sources might make it possible to identify an individual and breach data privacy laws. Privacy laws vary from country to country making it complex for both the equipment maker and farmer to navigate.

New data privacy laws, such as the EU General Data Protection Regulation (GDPR), are beginning to deal with AI explicitly. Under such privacy laws, key issues include whether all personal information used by an AI system has been collected with the data subject’s valid consent and such consent covers all purposes for which the AI uses the information.

AI tools are susceptible to hacking which can create both economic loss and physical damage. As a result of such incidents, businesses will need to ensure that all appropriate steps are implemented to guard against such risks and mitigate any breaches, in accordance with applicable legal requirements and industry practices.

Supply chain liability

A self-driving tractor might crash into a neighboring farm, damaging property and injuring people. Who might potentially be liable for the harm caused? The current legal system would not confer separate legal personality on AI tools. Different AI supply chain participants might potentially be attributed with liability. Example participants include: the commissioner of the AI system or the person paying for its design; the designer of the system; the person who prepared the technical and functional specifications; the programmer; the licensor or distributor; the integrator or installer; the trainer or tester; the owner of the system; and the operator of the system. Contractually, those in the supply chain may need to address more complex liability allocations than would be the case in a traditional supply arrangement. The autonomous nature of AI has the potential to shift that liability up the supply chain. The dynamic nature of AI makes it difficult to foresee how the software could evolve which may create unintended consequences.

Conclusion

Autonomous farm machines and sensors will continue to crop up on farms and agriculture data will continue to grow in quantity and scope, making farming processes increasingly data-driven and data-enabled using AI tools. Responsible development and deployment of AI tools requires careful navigation of the complex legal risks. This is particularly complex given the autonomous nature of the tools, the changing nature of the AI system by machine learning and novel uses cases. Businesses should create a defensible process for the use of AI and consult with experts.

Please visit our site www.aitech.law for a detailed review of the legal and ethical implications of AI.


  • 1 Using Transfer Learning for Image-Based Cassava Disease Detection, Amanda Ramcharan, Kelsee Baranowski, Peter McCloskey, Babuali Ahmed, James Legg, David Hughes, Frontiers in Plant Science 2017 vol. 8 p. 1852.
Top

Contacts

Maya  Medeiros

Maya Medeiros

Toronto Vancouver