An Ethical Awakening in AI

On unleashing innovation and human capital towards social and environmental impact

Artificial Intelligence meets Tech for Good

Recent increase in computing power, availability of data and technical talent have contributed to the third wave of AI. Scientists and engineers, attracted by high salaries, challenging technical problems and motivated by ambitions of changing the world gathered within the research departments of large technology companies.

In these environments full of bright peers, computing resources and an innovative culture, many projects such as deep learning for breast cancer detection are developed and support the narrative that progress in AI is obviously in the best interest of the public.

Problems with the narrative

In 2018, cracks have begun to appear in the tech for public good narrative and recent developments have raised a host of ethical questions as well as doubts about how AI is used in practice.

Prominent examples include the Facebook’s Cambridge Analytica scandal, clashes at Google over project Maven, a $10 billion Pentagon contract over the use of computer vision for drone strikes and similar tensions at Amazon over the use of facial recognition technology to power mass surveillance.

Universities watching these clashes unfold, are now part of a movement pushing for an ethical awakening to happen within the AI community. Responses to these developments on college campuses have taken two forms: top-down institutional and bottom-up student-led responses.

At the institutional level, leading engineering institutions, such as MIT, are reshaping to put ethical and societal questions at the center of their work. The newly announced Schwarzman college of computing at MIT will hire 50 new faculty and focus on “unlocking the massive potential of AI and explore its ethical and economic impacts on society.”

This means new research questions, ethics classes and creation of many forums to engage with these issues. By doing so, universities want to plant the seeds among future technology leaders to think about “what products they choose to develop, what policies they adopt around user data”.

At the individual and associative levels, students are banding together to create tech for good associations. As Vicki Niu , a leader of CS+Social Good @Stanford explains “we wanted to create a community of people that valued impact to remind computer scientists that they have skills to do things that matter and remind folks who are already doing impact-oriented work how technology is a tool they could also leverage.”

Government and International Institutions fail to attract talent

Recent progress in the area of blockchain and AI leaves many multilateral organizations, such as the World Bank and United Nations, puzzled about effective internal and external strategies they can put in place to keep up with this progress.

For example, at a recent gathering in Addis Ababa, representatives from the UN and various African governments investigated how they could use satellite imagery and GIS for monitoring the environment, coordinating and tracking efforts and capital allocation towards the SDGs.

Given student interest in technology for social good and the demand for such skills in the public sector, it is worth asking whether the demand is sustainably met and whether, after graduation, students keep engaging with these questions and lead adoption of AI in the public sector.

Unfortunately, there are barriers to this progress. Due to internal structures, bureaucratic processes and lower salaries, many organizations have failed to attract the kind of expertise and talent sought after to address these problems. The talent instead flocks toward private and public technology companies.

For example, 22% of the MIT class of 2017 went to work for either a technology startup or a large technology company after graduation. Only 0.3% went to work in the public sector. Top STEM talent cares about innovation.

While the public sector’s institutions market their drive for innovation, they aren’t structured to create an environment conducive to this. In addition, the perceived (and actual) bureaucracy of these institutions combined with the strong magnet for-profit technology companies exert on students make it even more difficult for tech talent to flow into these organizations. Yet these institutions play a vital role in shaping society.

As a result, talent does not always get allocated towards these impactful goals and humanity misses out. There’s room to put more of our brightest minds in positions where they can lead the charge towards mitigating the effect of climate change.

Need for a new model: Impact startups working with international Institutions

There is a clear need to change the paradigm. What we need are organizations conducive to innovation and intentional in their goals of allocating the best talent towards impact (i.e. solving the UN’s SDGs).

Creating this ecosystem requires new organizations (i.e. impact tech startups), funding models (impact capital) and financial instruments to reward impact (i.e social impact bonds) .

The responsibility for impact technology startups is to maintain a high standard of innovation, rapid iteration while working on commercial projects which are informed by and add value to the public sector which will pay them if a certain set of social outcomes are achieved.

Impact investors can catalyze these changes through investment in these technology startups, making them global role models in the innovation community. Such investments can spur entirely new type of companies accountable for allocating top talent towards solving the SDGs.

The same technology powering mass surveillance can be used to preserve the environment and planet. The reward structures we put in place now will shape how these skills will be used.

Sela, technology for impact measurement

Sela is a technology company sitting at the intersection of innovation and impact. With a team that has worked on oil pollution cleanups in Nigeria as well as machine learning for autonomous vehicles, Sela understands how to align social impact and innovation.

Following the UNEP report on Ogoniland, the issue of the oil spills in Nigeria has gained global attention and monitoring the progress of oil spill cleanups is vital for correctly allocating capital. As a first project Sela is using satellite imagery for mapping and monitoring oil spill cleanups.

Value proposition: Streamlined impact measurement

The value proposition of such a project is strong. With high resolution satellite images captured daily by companies such as Planet Labs , there is a rich set of machine learning techniques which can be deployed for :

  • Real time monitoring of oil pollution across entire states in Nigeria
  • Keeping track of oil cleanups over time, and building quantitative models to assess impact

Given the data generating mechanism, satellite imagery offer avenues for accurate and real-time accountability of oil spill cleanups. Even the most remote areas of the world can now be accessed with a few API calls.

Creating institutional environment propitious to innovation and impact

In order to execute such a project, various steps have been taken by Sela to both attract the best talent and allocate it toward impact-oriented projects.

A first step has been kickstarting a collaboration with the IDIAP Research Institute, a Swiss Artificial Intelligence Lab working closely with École Polytechnique Fédérale de Lausanne, a leading European engineering school. This collaboration will happen under the Innosuisse program designed to facilitate collaboration between startups and academia.

A second step, to complement this research collaboration will be offering externships to MIT students in January 2019. This will allow Sela to prove the hypothesis that impact tech startups can lead in innovation, attracting the best talent towards solving the SDGs.

Technologists want to deploy their skills towards impactful projects in the public interest. However, bureaucratic barriers of entry deter the sustainable allocation of tech talent towards these projects. Overcoming these hurdles requires governments, private companies and impact investors developing a new ecosystem.

New financial instruments such as social impact bonds and impact investing can help with incentivizing such efforts. Programs such as Innosuisse, Microsoft AI for earth grants can facilitate collaboration between startups, academia, the public sector and large tech companies. Given the urgency of the crisis we’re facing, we have to act now to make sure our best talent is working towards these goals.

Hassan Kané is a recent graduate of MIT. He’s a tech entrepreneur and AI researcher currently serving as the CTO of Sela