Can AI-For-Good-Project Deliver its Results Up To the mark?

Artificial intelligence (AI) has been at the front and centre during the COVID-19 pandemic. The global pandemic has pushed governments and private organisations globally to propose AI solutions for everything from analysing cough sound to installing disinfecting robots in hospitals. Such efforts are part of a broader trend that has been picking up momentum- the deployment of projects by companies, governments, universities, and research institutes aiming to use AI for societal good. The goal of these programs is deploying cutting-edge AI technologies to solve crucial issues like poverty, hunger, crime, and climate change, under the 'AI for good' umbrella. But the bigger question is what makes an AI project good?

AI has the potential to address some of humanity's biggest challenges, such as poverty and climate change. However, as a technological tool, it is agnostic to the context of application, the intended end-user, and the specificity of the data. And for that reason, it can ultimately end up having both beneficial and detrimental consequences.

Here are some ways how to make sure if an AI for good project is actually good.

Does a Business Need AI Solutions?

Successful implementation and adoption of AI in businesses will require three things:

• Being able to distinguish between what's possible and what's still sci-fi with the help of an expert
• A problem or task that AI can be used to solve
• Data, data and more data

Although AI can accomplish a lot of things, it's limited by technology. Having someone involved in the project who can tell a business owner what's possible and what isn't will save a lot of time in the long run.

AI is very popular at the moment but not making the mistake of starting an AI project is advised. Working with a partner can analyse the business and honestly let the business owner know if AI is required for the business or not.

For any AI project to be successful, one needs enough data to feed into it for machine learning (ML) to happen.

Asking the Right Questions

Before involving into a project intending to apply AI for good, there are a few questions one should ask. The questions include what the problem is. It is impossible to solve the real problem at hand, whether it is poverty, climate change, or overcrowded correctional facilities. So a project inevitably involves solving what is, in fact, a proxy problem: identifying poverty from satellite imagery, detecting extreme weather events, producing a recidivism risk score. There is also frequently a lack of adequate data for the proxy problem, so one relies on surrogate data, such as GDP per census block, extreme climate events over the past decade, or historical data regarding inmates committing crimes when on parole. But what happens when the GDP does not imply the whole story about income, when climate events are progressively becoming more extreme and unpredictable, or when cops data is biased? One ends up with AI solutions that optimise the wrong metric, make erroneous assumptions, and have unintended negative consequences.

AI solutions are too complicated, too expansive, and too technologically demanding to be deployed in many ecosystems. Therefore, it is of paramount importance to take into account the context and challenges of deployment, the intended audience, and even more straightforward things like whether or not there is a reliable energy grid present at the time of deployment. Things that people take for granted in lives and surroundings can be very challenging in other regions and geographies.

Given the current ubiquity and accessibility of ML and deep learning approaches, one may consider they are the best solution for any problem, overlooking its nature and complexity. While deep neural networks are undoubtedly powerful in certain use cases and given a large amount of high-quality data relevant to the task, these factors are rarely the norm in AI-for-good projects. Instead, teams should prioritise simpler and more direct approaches like random forests or Bayesian networks, before jumping to a neural network with millions of parameters. Simpler approaches also have the added value of being more easily interpretable than deep learning that is a useful characteristic in real-world contexts where the end-users are frequently not AI specialists.