Over the last few months, very few concepts have captured the world’s imagination as much as that of AI. Naturally, this fascination has engulfed Ghanaians in more ways than one. As a software developer, I have had the advantage of being at the forefront of this technological revolution. In addition, having early access to open AI’s ChatGPT before it became the topic du jour in many circles allowed me to experiment with and study the implications of AI.
A raging debate about the existential and economic threats posed by the rapidly evolving technology, fanned mainly by the media, has successfully hidden the fact that many of us do not know what AI is besides the fact that it refers to the term “Artificial Intelligence”.
What is Artificial Intelligence?
Artificial Intelligence refers to computers and machinery mimicking (or exceeding) a human being’s ability to acquire skills and knowledge. This happens largely through the provision of voluminous labelled training data, which are analyzed for patterns and correlations and are then used to make predictions about future states. ‘Training data’ is information ‘scraped’ from online sources – essays, blog posts, digital art, articles etc. – used to model format, content and style to AI.
Narrow AI or “Weak AI” refers to any AI application that can outperform humans in a narrowly defined or structured field.
Let us imagine we have a friend, ‘Kwaku Ananse’, who was just employed as a customer service representative at a fictitious company called ‘High Heels Factory Ghana’. His job is to respond to any enquiries that come to the company through their WhatsApp line.
Kwaku Ananse has no ambitions except to be the best WhatsApp Customer Service Representative for ‘High Heels Factory Ghana’. Kwaku Ananse is blessed with a wonderful memory he intends to use on the job. Fortunately for him, ‘High Heels Factory Ghana’ also has millions of previous messages between customers and customer service representatives that were considered excellent responses by the company management.
Kwaku Ananse spends all night memorizing each question and answer and notices that the questions are either related to the availability of a particular pair of shoes, the availability of the shoes in a given size, the cost of the shoes, the cost of delivery or the ability of “High Heels Factory Ghana” to deliver to the person’s location.
He “labels” each question based on the categories above and notices a pattern of words in a sentence suggesting the customer wants to know the price of the shoes.
Kwaku Ananse may have noticed the phrases, “How much are the shoes?”, “How much?”, “The shoes, how much?”, “How expensive is it?”, “Ei, how expensive!”, “Is it expensive?” all receive a response of “It’s only [price of shoe]”. Recognizing this pattern, when he first sees a customer ask, “How much is it?” he recognizes the pattern as similar to “is it expensive?” and all its labels and thus responds with the price of the shoes.
Kwaku Ananse is our Narrow AI. Within the narrow confines of “High Heels Factory Ghana”, he can consume lots of data to respond “automatically” to the questions. Usually, his work is “supervised” by a human who will assess his ability to “guess” the question and respond appropriately to make adjustments or corrections in his assumptions where needed.
Narrow AI is quickly becoming a huge part of our lives. From chatbots to autonomous vehicles to well-known brands like Siri and Alexa, a method not dissimilar to what Kwaku Anase went through is what forms the basis of their technological feats.
The Good in Narrow AI
From the perspective of the CEO of High Heels Factory, Kwaku Ananse, the Narrow AI can be a huge asset. Customers will no longer have to wait several minutes or hours to find out if a particular pair of shoes in a given size is still in stock since that can be answered without human intervention. The Customer Service Unit can now focus on enhancing customer experience strategically without devoting human resources to endlessly answering repetitive questions.
Stepping away from our example, there are remarkable opportunities in healthcare, communication, education, transportation and logistics, which could see some advances that would benefit the bottom line of forward-thinking companies in these sectors.
The Hype of Narrow AI
Based on the media narratives, one could be forgiven for thinking that the progress made by Narrow AI necessarily means civilization as we know it is over, especially as people on the African continent; we have a perspective and a reality far different from what exists.
It is evident that since many of the feats are still based on previous work and human assessment of them, there is little risk of human beings playing second fiddle to Narrow AI.
We must, however, concede that there is a chance that access to Narrow AI, which will primarily be based on financial ability, could accelerate and increase the divide between the “haves and the have-nots” in society.
Narrow AI could displace those unwilling or unable to change whilst empowering those well-positioned to take advantage of it.
The Ugly Side of Narrow AI
The most visible unknown effect is what will happen to employment. Occupations that rely on the ability of professionals to retain vast amounts of specialized knowledge whilst performing repetitive tasks are indubitably at risk. I am convinced, however, that there is room for growth in jobs that require creatively combining existing knowledge to birth new solutions. After all, Narrow AI relies heavily on these solutions to add to its knowledge base.
What effect does this have on the nature of education? Here lies my biggest concern.
I am old enough to remember how in the late 90s, technologists looked forward to the so-called information age. It was meant to usher in a period of outstanding human achievement in which easily accessible knowledge and information could emancipate many people from poverty.
Whilst this has happened to some extent, there have been many who, I would argue, have been mentally impoverished by the ubiquity of mindless entertainment through technology.
How does this manifest when it comes to Narrow AI? I believe that early access to AI would have an effect similar to introducing calculators to toddlers before they can learn arithmetic. There will be an overdependence on the answers Narrow AI gives. As long as there is a human element that fine-tunes the responses in the background, there is an opportunity for nefarious actors to warp reality.
Indeed, these AI sources themselves often have trouble distinguishing fact from fiction. A phenomenon that is not fully understood but results in many people being misled when using them as ‘search engines’.
The future will belong, however, to those who choose to use Narrow AI to augment their productivity and increase their knowledge base by using it not as a tool for reference but rather as a tool for learning.
Enter two software developers, Serwaa and Yaa.
Serwaa has learned the right prompts to get her AI engine to generate code in bits and pieces to build applications that satisfy her client’s needs. Within an hour or two, she can develop and deploy minimum viable products without spending significant time thinking about the algorithms needed to solve each problem.
Yaa’s approach is slightly different. She uses her AI engine as a tutor in her software development journey, prompting it for explanations where applicable, asking for references for concepts, and complementing the knowledge gaps with online forums and documentation for each concept or programming language being studied.
Serwaa may have just put herself on a path to accidental ignorance. As her project expands, she loses her ability to understand her code base until the day it inevitably “breaks”. Unfortunately, if her AI cannot detect the issue, Serwaa is woefully unqualified to diagnose and detect the problems herself. Yaa will, however, have no trouble doing this if every line of code was generated in her brain and thus probably matches her thought patterns in a way that facilitates troubleshooting.
I am thoroughly convinced that actual gains in the workplace will come from the employees who are such experts in their fields that they can detect the moments when an AI engine is wrong and thus can prevent themselves or their organizations from potentially fatal errors.
The fear of losing one’s job to AI can be alleviated with expertise always committed to being more advanced than any Narrow AI. For any ambitious professional, utilizing AI to stretch one’s brain and expedite learning would be an excellent skill to hone.
Artificial General Intelligence.
What happens if AI can expand beyond a specialized field to become an omniscient being? Is this even possible? What happens if it can do this unsupervised, continuously increasing and refining its knowledge base? Can we regain control of it? Would it lead to the extinction of humankind?
The idea of an all-knowing, infinitely learning artificial intelligence model or General AI elicits awe and fear amongst the intelligentsia. Will the AI model acquire human “emotions” or, more realistically, human motives whilst pursuing knowledge?
Will the AI model prioritize its survival and thus begin to evade some or all of the fail-safes humans would try to place on it? Will it attempt to acquire power after it is capable of out-thinking the most intelligent humans and thus conclude that it would rather not accept tasks from intellectually inferior creatures?
For now, hardware and software limitations prevent these fears from being realized; however, things could get ugly pretty quickly for the “powers that be” as the gap between science fiction and reality closes with each passing day. Should Africans be worried? I would argue not.
On the other hand, one could argue that the only entities that should fear the rise of General AI and its potential desire to attain power are those who currently wield it over the general population. For those suffering under the current regime, we should probably welcome this paradigm shift as an opportunity for us Africans to compete equally with the rest of humanity.
However, that should probably be discussed in detail regarding AI’s implications for governments and governance.
This writer is, convinced, however, that the very monopoly that our lawmakers possess over the laws of the land will be utilized to prevent their power from ever being threatened. I am willing to bet that our leaders will enact all the laws necessary to ensure their survival and thus prevent General AI from usurping their authority.
The author is a software developer and technology & business development consultant with over 19 years of experience in the industry. He specializes in helping small and medium-sized businesses leverage technology to drive rapid revenue growth and maximize market share.
+233 507409529
