The Oxfordshire based company first submitted its proposal to the IBM Watson AI XPRIZE competition at the beginning of 2017, along with many other hopefuls. Their proposal to crowdsource development skills from around the world to reduce food waste and improve food supply was accepted by the prizes independent panel of experts to move forward onto the next phase in the competition.
The IBM Watson AI XPRIZE is a $5 million global competition, challenging teams to develop powerful AI based applications and demonstrate how humans can collaborate with AIs to tackle the world’s grand challenges. AI Gaming’s project aims to ‘Crowdsource an artificial intelligence solution for efficient food production and distribution’ and has been judged against four criteria; its importance to humanity, the solution’s progress and probability of success, its positive broader implications and its technical novelty.
From an initial 147 teams from across the globe that entered the competition over a year ago, AI Gaming is one of only 3 teams from the UK that have been selected to continue to Round 2, leaving a total of 59 teams in the competition. Company founder, Paul McDonnell commented that “As a start-up business based in rural Oxfordshire, this achievement comes as the result of a lot of hard work and a very eventful year”. With the acclaim of its IBM Watson AI XPRIZE success, AI Gaming has been able to grow considerably in 2017, engaging with new partners, and creating a technical platform that can host the crowdsourced development it aims for. The development and implementation of a number of hackathon events is helping the company to reach their potential AI developer user base and is moving the company significantly closer to delivering their final solution.
According to recent reports 1 , in the EU alone, the manufacturing, wholesale and retail supply chain accounts for 39m tonnes of waste food per annum, and 44% of food waste occurs in manufacturing and distribution, compared to 42% from the household. Global food waste is estimated to be 2 billion tonnes per annum. AI Gaming is implementing a novel AI approach, combining the efforts of hundreds of AI solutions, to more accurately predict future demand so that at every stage from farming through to retail, food waste can be minimised.
Paul states his view of this issue as “Part of the problem is the overproduction of food due to poor prediction of sales levels. Our research shows that at each stage of the production and distribution process, inadequate solutions are currently in place, leading to unnecessary wastage. By taking information such as historic sales, weather, day of week or year, public holidays, trends in complementary products and categories, social media trends etc, we aim to produce a system which can predict food demand up to one growing season away from the point of sale”.
Paul goes on to say that “We are confident our AI solution will be able to produce estimates that are much more accurate than the current state of the art which in our research only takes a limited input data set. AI is well suited to taking a multitude of inputs which may or may not have an impact on the result and finding correlations which no human could expect to identify.”
Paul’s goal is to affect an area that will have significant impact and benefit for everyone. “We aim to reduce food waste in the supply chain by 25%, saving approximately 10m tonnes of waste per annum in the EU with its associated environmental impact in land use, transportation and disposal”.
To date, AI Gaming have created a platform to build a developer community and have been able to partner with a large UK food retailer to understand the issue and source historic data to train and test the AI being developed by their users. The platform creates an environment where AI developers compete in ongoing continual gamified challenges where better performance is rewarded and a form of natural selection occurs. Once qualified, developers can participate in solving real world problems such as food sales level prediction. The outputs of the participating bots are weighted according to their previous performance and they share in a micropayment made by the customer in a form of crowdsourcing. Paul’s view of this is that “Our approach is a kind of ‘meta AI’ where we harness the wisdom of the crowd of AI development expertise from all around the world”.
The crowdsourced solutions are inherently subject to “survival of the fittest” – developers whose code consistently doesn’t perform as well, will try alternative approaches, switch to alternative challenges or retire. This creates an environment where evolution of ideas and approaches will occur, with a persistent incentive for the developers’ coding expertise to improve.
AI Gaming plan to expand on their reach into this development community by introducing a certification scheme and the development of an education program that will guide and incentivise developers to tackle progressively more challenging projects until they are advanced enough to take part in the food sales level prediction challenge.
All of these initiatives are being rolled out now with scheduled events and training courses at universities throughout the UK reaching over 2,500 students in the coming months. Their first lecture series will be held in conjunction with the Oxford University Coding Society early in 2018.
1 AMI SEDGHI. 2015. UK tops chart of EU food waste, The Guardian. 22 May 2015
About AI Gaming
AI Gaming is an innovative, technical startup based near Oxford, UK. The company is building a platform to harness the skills of software developers all over the world. After a year of proving the concept, they have developed a robust and flexible platform that has seen over one million games played, and is being used to crowdsource solutions to challenging problems, educate users in AI and Machine Learning and as a tool to provide proof of technical competency for potential employers.