Robots in the Food System
Today, the National Science Foundation announced that UC Davis will host a new $20 million research institute on artificial intelligence. The AI Institute for Next Generation Food Systems (AIFS) is one of seven new national artificial intelligence institutes, and one of two sponsored by USDA that focus on agriculture and food.
I am excited to participate in AIFS as leader of the Socioeconomics and Ethics group. A distinctive feature of the institute is that we view social, economic and ethical issues as foundational to the development of AI technologies in the food system, rather than something tacked on at the end. We recognize that human decisions and preferences are intricately linked to every stage of food system supply chains.
AIFS will develop generalizable, data-efficient, and trustworthy AI solutions to enable (1) Molecular breeders to discover and/or design the next generation of high yielding, high-quality, consumer-focused foods, (2) Agricultural producers to maximize food quantity and quality, while minimizing resource consumption and waste, (3) Food processors and distributors to deliver highly traceable and safe food, while minimizing resource consumption and waste, and (4) Consumers to rapidly and precisely assess the nutrition of a meal, quantify the food’s molecular composition, and predict the impact on their health. AIFS will extend these solutions through programs in Education, Outreach, and Workforce Development.
So, where do socioeconomics and ethics come in? Here are a few examples.
A successful AI technology is a trustworthy AI technology. It needs to be safe, fair, and to protect privacy. We will develop an ethical framework with clear answers to questions such as (i) what are AI developers asking people to trust them with and for what purpose? (ii) what practices are in place to assure trust?, (iii) how effective are those practices?, (iv) how vigilant are AI developers and to whom are they accountable?, and (v) how will we train grad students and postdocs in ethics of publishing and transparency?
A successful AI technology will be incentive compatible. If people don't see value for them in the technology, they will not use it. Or, they may use a technology to enhance their own welfare at the cost of worsening environmental outcomes. Establishing incentive compatibility requires modeling and understanding decisions such as how farmers decide what to plant and when to irrigate, how firms decide to adopt and use new technology, and how people decide what foods to buy and eat. Understanding these decision processes can mitigate unintended consequences.
A successful AI technology will account for worker wellbeing. Widespread adoption of AI technologies will reduce the demand for workers in fields and on production lines, for example, but will create new jobs operating the technologies. This technological transformation comes at a time when there is a shortage of agricultural labor. It is important to understand its effects on workers and the communities in which they live.
Those topics merely scratch the surface. I'm looking forward to a productive and interesting five years (or more) working with lots of amazing scientists.
