On a dusty university campus on the outskirts of Nairobi, Kenya, 750+ eager minds gathered at the annual Deep Learning Indaba - this being the third edition of its kind. The conference brings together researchers and practitioners from over 30 different African countries with an interest in machine learning and artificial intelligence (AI).
Having attended last year’s version in Stellenbosch, South Africa I was pleasantly surprised to see how much more expansive the Indaba had become. It is clear to see that the hunger for knowledge and community has not diminished. Among the various conferences that I attend, it holds a unique space as potentially the only one where it truly feels like a family that is coming together again after some time apart.
I can’t over-emphasize how important I think this community is. Machine learning has been re-ignited over the past few years as improvements in processing power and algorithmic efficiency has allowed us to take the training wheels off and work more intelligently with incredibly large datasets. As a result, the technology is a hot property right now - for good reason.
The hubs of Silicon Valley, Beijing, London, etc. attract a lot of attention as most of the key technical breakthroughs and influential technology companies emerged from those locations. They control the nexus of AI decision making. This concentration enables focused, fast-moving development but it also creates a bit of a bubble that has some nerve-wracking externalities (I would argue).
So in response to that challenge, the Indaba is Africa’s call for a seat at the table. It celebrates the excellence that abounds in African research (Don’t doubt that for a second) and brings together the right people to have the important conversations needed to solve uniquely African problems.
This for me is the greatest opportunity, hiding in plain sight. When we consider the types of problems that our current machine learning methods are best at solving, Africa has them all. It really is the very best place to be working on AI because these obstacles are such powerful opportunities for the technology to make a real tangible impact. They are rich and complex and any breakthroughs that make a dent in them will reverberate around the world. The future of AI is in Africa.
“We have guided missiles and misguided men.”
This quote from Martin Luther King encapsulates one key transformation that I want to see. The conference is dedicated to high-level machine learning research and rightly so - advanced mathematics and computer science are the bedrock of these advancements.
However, this technology’s power is ripe for misuse, both malicious and incidental, if we don’t think carefully about the ethical principles we are using to develop, maintain and interpret the results. The ethical and societal impacts of the decisions made by the developers sitting in those sessions will be incredibly important in the years to come and yet there was very little said on any of the ethical topics that worry those of us who are outside of the field.
This is not a slight on the organisers at all, because they dedicated a significant portion of one day of the conference to ethics. However, the turnout and engagement at that session was underwhelming. This was somewhat predictable because there were various other exciting sessions that were happening in parallel - which drew the majority of the crowd.
I think that this is indicative of a larger concern in the industry which is dominated by the discipline of computer science. I think it is important that we inject a helping of history, philosophy and economics into the discussion. We need more multi-disciplinary thinkers to be working on things like transparency, explainability, alignment and other ethical topics that don’t factor nearly enough in the minds of the developers who often find themselves racing ahead at all costs. This is not true of all developers of course, I had a number of encouraging conversations with fellow delegates about these topics, but it is true of a larger proportion than we often care to admit.
One key example is the desperation for large datasets. In Africa, we are very constrained when it comes to datasets that are useful, structured, labeled and large enough to run many of the models that we’d like to - and as such - any dataset that you can get your hands on is worth its weight in gold.
This lack of supply inevitably lowers the skepticism, rigour, and testing that might be applied to ensure that a dataset is fit for use. The bias and social context that is present in these data must be taken seriously and when we are celebrating too early, it’s easy to overlook these parameters. This is especially true when the pressure to keep up with the Western world is immense.
Again, this is not an Indaba-specific critique but an industry-wide one. We are not thinking carefully enough about the societal implications of the algorithms that we are creating.
Being the Stupidest Person in the Room
My last few comments refer to a wonderful epiphany that I had while sitting in the hall at Kenyatta University. In most respects, I had no right to be in those sessions. My computer science knowledge is nascent, at best, and while I’ve been working very hard on up-skilling myself over the last few years, I was still well out of my depth at a conference of this stature.
But that was magical. I can’t explain the joy of finding something that you find intellectually fascinating but immensely difficult and spending a focused week trying to swim in that pool. For one week I did nothing but eat, sleep and breathe machine learning with some of the brightest minds on the continent. In every discussion I had, I was the stupidest person in that discussion.
The opportunity for growth and learning is a drug that I live for and by putting myself in those situations, I felt truly alive.
I look at AI from a different perspective than a born-and-bred computer scientist. I bring my own expertise from seemingly unrelated fields to the table and that’s why I believe I have something to bear on these conversations.
I’m not there yet. I’m still the underdog. But every day I’m learning more and getting a little bit less stupid. Don’t underestimate me, you’ll regret it.
As one of my favourite South African proverbs says: “Knowledge is like a lion, it cannot be gently embraced.”
See you in Tunisia in 2020!