ChatGPT take 1/4

data24

3 things I learned from reading the patent application publication

There is a dominant narrative that wants us (Africans) to believe that innovation only happens elsewhere and that innovation is happening so fast we can't keep up, so we need to embrace or take what's been given to us. But before we embrace innovation from outside, we need to truly study it. By studying, I don't mean watching and reading every piece of marketing material out there. I mean having a healthy dose of skepticism when studying these innovations. Here is what I've done so far to learn about Open AI ChatGPT. On the patent first.

3 things I learned from reading the patent application publication

  1. ChatGPT is about taking communication to the next level by automating some or all of it. ChatGPT is really improving existing automated response generation tools, like Gmail Smart Reply. A few drawbacks of the existing automated response tool include: (1) short phrases (predefined) (2) The existing tools are confined to a specific channel (i.e., Gmail), and (3) the current tools are not robust, i.e., lack context or ability to contextualize communication.
  2. The patent is very vague when it comes to the nitty-gritty, i.e., what type of datasets are used to train the model, and how big are they? What are the privacy considerations? permission?—it just referred to the Internet. Given that the Internet is not representative of Africa, one can argue that these models won't be particularly robust if there is not enough data on the continent.
  3. What is fascinating about this tool is its ability to generate natural language responses based on context while also performing computation at the same time. I can't imagine the kind of power needed to run such engines. Do we have that kind of power in Africa?

What does it mean in Africa?

1) AI requires data (training), and we still don't produce enough data. So before jumping to queries on ChatGPT, let's be aware that the model is only as good as the training data. One can argue that because we've really spent our time copying and pasting from the west or outside, it really doesn't matter. The flip side is that we can now learn at a faster rate with the data available from the west. More on that in my last take (4/4)

2) You can only appreciate the disruption if you have context. If you've written a poem before, you will be amazed by the speed and style of ChatGPT, but only if you've done it before. This is important for our youth, who are now flocking to Chat GPT to solve their assignments or essays.

3) If you run a startup in Africa (fintech, etc.) where you collect data, check your IP and IP ownership quickly. You may be in for a shock :) In the digital world, true ownership resides in the IP. I think by now you've realized that your cap table is another indicator.

That's it for now for Take 1. Tomorrow we will talk about the true test of ChatGPT in Africa.

If you want to read the open AI patent, its here: https://www.patentguru.com/US20220229999A1

Contact our team today to explore how our consumer intelligence and brand intelligence can empower your decision-making process. Win with confidence with Kasi insights https://www.kasiinsight.com/request-demo/


Recent posts

See all

Segun Sunmonu

Narrowing the AI Divide in South Africa: Skills, Access, and More

Catherine Wangari

Zambia's Telecom Landscape: Gender-Driven Preferences and Spending Habits

Wilson Busaka

Unmasking Kenya's AI Landscape: Tackling Gender Disparities Amid Job Loss Concerns

Subscribe to our free newsletter