OpenAI has announced the release of GPT-4, its next-generation language model; and with it, more discussions around how every job and everything is in danger of falling victim to AI and how every job, everywhere, will soon be taken over by some AI smarter and faster than every human that has ever been born.
While this makes for some interesting conversations, the reality is somewhat more nuanced and deeper to understand, especially in the context of an environment like Nigeria’s. What exactly does ChatGPT mean for us and why should we even care?
Let’s dial back a little bit or a second. Who is OpenAI and what are language models? OpenAI is an American artificial intelligence research laboratory consisting of the non-profit OpenAI Incorporated and its for-profit subsidiary corporation OpenAI Limited Partnership.
OpenAI conducts AI research with the declared intention of promoting and developing a friendly AI. ChatGPT is an artificial intelligence chatbot developed by OpenAI and launched in November 2022. It is built on top of OpenAI’s GPT-3 family of large language models and has been fine-tuned using both supervised and reinforcement learning techniques.
A language model is a probability distribution over sequences of words. Given any sequence of words of length, a language model assigns a probability to the whole sequence. Language models generate probabilities by training on text corpora in one or many languages. In a nutshell, ChatGPT is a chatbot that has been trained to respond in a natural way to the questions that people throw at it by using the natural language it has been trained on by the folks at OpenAI.
That’s simple enough; we know that chatbots essentially respond to us based on the information they have been trained on. This is why in their former general narrow senses, they could get things wrong and we were okay with just being handed to a customer service representative, for example.
But things have moved on; computing power has become so ubiquitous and easy to access, for the right dollar amount, of course, that it has become a lot easier to develop models to train bots, and programs, to do all sorts of interesting things. In the not-so-recent past, we saw only the behemoths in the name of Google, Meta, Amazon and Microsoft deploying huge computing power to drive AI implementations in a not-so-publicly accessible and impressive way.
Autocomplete in Gmail, which has been available for years, is an example of an AI-driven utility, and so are Grammarly and other text-correcting utilities. The exciting thing was this mass market access to an AI which essentially made it look like you were chatting to your really smart friend to get an answer to a question.
Read also: ChatGPT: The new normal in technology today
The variation was also something new; on the one platform, you could ask about general information, coding, history, etc and you would get a well-formatted response. No wonder the temptation to use it for essays and other types of generic tasks became too hard to resist for some. CNET recently had to admit it was using AI to write some of its tech articles and had to pause the exercise due to the factual errors found in more than half of the AI-generated stories.
So, what about us? Will Nigeria suddenly see a spike in unemployment as more chatbots and special-purpose AIs are released into the public domain to create headaches for HR and management? As I suggested earlier, the reality is somewhat more nuanced. The models which drive the bots, ChatGPT, for example, are all based on information that they gather from the Internet and other sources they may be fed.
This means very simply that they contain the same biases, which the humans who generated the data have passed on and in some cases, they are just plain wrong. This second fact is possibly the one we should be very weary about. ChatGPT and other AI will get things wrong because they are only as good as the information they are trained on. In the Nigerian space, we currently have large gaps in the data that can be used to train models and there are very few local technology firms that have access to the computing power to develop bots like ChatGPT.
By some accounts, it is costing OpenAI $100,000 a day to develop ChatGPT, hence the need for significant investments from Microsoft. There are other significant ethical issues that need to be addressed before AI moves from very good recommendation engines for more nuanced users who can shape context to full-blown replacements for intelligent human beings.
We are safe for now but this certainly feels like an inflection point in the injection of AI into our everyday lives, the scope and application of which still remain unknown.
Mordi is a former chief operating officer of Carbon, a lending fintech, and immediate past head of digital lending at Access Bank. Twitter: @epmordi
