It is not news that Artificial Intelligence (AI) is shaking up the world, unlocking new frontiers in addressing business and societal challenges. The AI progress metrics maintained by the Electronic Frontier Foundation show that in several applications, including tasks previously thought to be doable only by humans, AI has superseded human ability. Businesses and countries are leveraging this to improve their bottom-line and make life easier for their citizens respectively. They are also putting their money where their mouths are.
Globally, there was an investment of over 6 billion dollars in AI companies in just 2017. While a lot of discussion on AI, especially on the African continent focuses on its propensity to take away jobs in the context of huge unemployment burdens, there is also a strong consensus on the potential of AI for good?—?to address the enormous socio-economic challenges on the continent.
At the heart of the AI revolution is the data explosion. The massive amounts of data coming from increasingly pervasive sensors, social networks, and mobile devices are unlocking new opportunities. This explosion has fortuitously coincided with recent success in the application of a technique called deep learning that uses graphics processing unit (GPU) accelerated computing. The current state of the art in AI is driven by the success of deep learning, and the explosion in the 3 Vs of data?—?velocity, volume and variety and the parallel computing advantage that GPUs provide.
To say that AI is unlocking new frontiers, even in well established industries, is not a hyperbole. The ability of an autonomous vehicle to detect and label objects in its surrounding environment, or identify lane markings, for instance, is crucial to its success.
Deep learning techniques have been shown to exceed human level accuracy in image classification, and are today applied successfully in autonomous vehicles by the likes of Google, Uber and Tesla among many others in the space. In a similar vein, predictive policing is no longer fictional. The New York Police Department’s and many police departments in the US are leveraging AI to significantly enhance their ability to predict where and when crimes are more likely to happen and who may commit them.
Global outlook on AI in healthcare
AI is being applied to numerous industry verticals and healthcare is not left out. The pace of adoption of technology in the healthcare industry is often constrained by policies and regulations and for good reason. That said, the wave of AI has also been felt in the industry. Machine vision is being applied widely to the diagnosis of various health conditions. From detecting the potential presence of cerebral bleeds to identifying cardiac illness from echocardiography data, AI is introducing efficiency and accuracy in healthcare. This is in addition to the increasing pervasiveness of health chat-bots. A doctor can only see so many patients daily, but a conversation agent has no physical limit on the number of people it can interact with simultaneously. A company called Woebot, based in the US, recently launched a chat-bot to address mental health issues. Babylon, headquartered in the UK, uses a conversation like agent for triaging patients and to perform simple diagnoses.
Current AI endeavors and potential in Africa
With a population of over 1 billion, plagued by serious institutional deficits, the African continent presents a tremendous opportunity to leverage AI for good and to accelerate solutions to fundamental problems like quality education or access to healthcare. Focusing the lens of AI endeavors in Africa, IBM Research Lab in Kenya is applying AI to determine optimal interventions for eradicating malaria specific to a given location and leveraging knowledge from other disciplines such as game theory. In South Africa, their sister lab is researching the automated extraction of clinically relevant information from pathology reports.
Furthermore, a Nigerian company is attempting to detect birth asphyxia?—?a condition arising when the body is deprived of oxygen by analysing the sound of a baby’s cry at birth. The opportunities to apply AI in healthcare on the continent are numerous. There are fewer regulatory bottlenecks around patient data privacy compared to other regions. There are pros and cons to this situation, but in the long run, optimism is the best strategy for regret. For starters, conversation agents are able to extend access to care to millions of people through the convenience of their mobile devices. In addition, computer vision can be used and is being used across the world to diagnose eye or skin conditions using images from the cameras of everyday smartphones.
Given the dearth of ophthalmologists and dermatologists in many parts of the continent, this has substantial benefits. In addition, pharmacogenomics (the branch of genetics concerned with determining the likely response of an individual to therapeutic drugs) is an interesting area to apply AI. Considering that Africa is said to contain the most genetically diverse population, advancement in this area have profound ramifications for the continent. The application of deep learning to accelerate genomics research is gaining considerable momentum. The implication of genomics and pharmacogenomics in healthcare is that disease could be treated according to genetic and specific individual markers with potentially better outcomes.
Exploiting AI for business or social value
Despite the potential of AI in unlocking business or social value, expertise is sorely lacking. “…people don’t know how to do this stuff” according to a Wired article. Any organisation or individual looking to leverage AI in healthcare or any other industry would typically need to go through a number of steps. What follows is a sequential enumeration of the key steps. The enumeration is not formalism and the steps are by no means de facto but rather come from personal experience.
1. Decide to leverage AI: Clearly, a conscious and deliberate decision needs to be made to leverage AI for whatever reasons an entity may have.
2. Determine the specific vertical in your sector or business to which you want to apply AI: AI may not be directly applicable to all parts of a sector or business. Deciding the vertical to which AI should be applied can be driven by the bottom line, limitations of current approaches or potential for AI in light of factors such as automation propensity, availability of data, cost or current degree of inefficiencies.
3. Map out the goals for your AI: Define what success looks like. If you don’t know where you are going, any road will lead there. Ensure to establish that ethical standards are met, especially in sectors like healthcare.
4. Draw out/plan a data collection strategy: Ensure to take into consideration potential biases, local data policies and laws.
5. Implement your data collection strategy.
6. Try out multiple models: Despite the unreasonable effectiveness of certain AI techniques, a one-size-fits-all strategy is doomed from the start. You’ll likely spend a lot of time on this step. Be aware of the no-free-lunch theorem and ensure your assumptions are well documented.
7. Test early: Carry out end-to-end testing as early as possible.
8. You may need to refine your data collection strategy given the initial results.
9. Roll out your model: Ideally starting with a small group of end users.
There is no doubt about the impact of Artificial Intelligence as a technology. I hope this article has helped you to think of how AI can help you meet your goals.
The writer is a Master’s student at Carnegie Mellon University in Kigali and a technology enthusiast.