AI: The Boom

AI: The Boom

Happy Monday!

The verdict is out: AI is undoubtedly shaping our reality at an unprecedented pace. But would it kill us, take our jobs, or make our lives much better? What does it even mean for us as Africans? And how do we take advantage of its enormous potential?

Here is the second in a series of four on my fully-fledged thought about Artificial intelligence.


If you missed the last edition (AI: The Gears), here's a summary:

- AI is the simulation of human intelligence in computers, allowing them to think and learn like humans.

- In traditional computing, computers follow instructions to execute tasks predictably, even in complex applications like gaming.

- AI, however, shifts from using algorithms to solve problems to employing machine learning, which enables computers to learn and adapt.

- Using football as an analogy, traditional computing approaches playing football using step-by-step algorithms, while AI learns from watching several matches to optimize for winning.


TL;DR

  • Machine Learning is a part of AI that lets systems learn from data instead of being explicitly programmed. It figures out patterns and predictions by analyzing data, as seen in the example of a football AI. Unlike traditional programming, it derives its own rules based on inputs, leading to creative and unexpected outcomes.
  • The rise of Machine Learning is due to improved computing power and the shift from CPUs to GPUs for AI tasks. GPUs can handle multiple tasks at once, boosting efficiency and enabling exponential growth in AI capabilities, even defying Moore's Law. This power has incredible potential but also raises several concerns.


What is Machine Learning?

Machine learning is a subset of artificial intelligence that enables systems to learn and improve from experience without being explicitly programmed. It works by analyzing large sets of data to identify patterns and make predictions. But this is too much big English, let's come to the basics.

Staying on our football example from AI: The Gears, an AI system doesn't grasp the rules of the game; it learns by watching numerous matches and deducing winning strategies. It optimizes for success without necessarily understanding the underlying principles of penalties or goal kicks, but by observing, learning, and adapting from the data of several football matches it has been shown.

No alt text provided for this image

With machine learning, you do not give a computer a rigid set of rules that says if this happens then do this else do that, otherwise do this other thing, and these are the possible outcomes. You, instead, give a computer a set of inputs without outputs or an end goal and allow it to create the rules that go from one step to the other until it reaches the end goal.

No alt text provided for this image


This means that it may come up with rules you never thought of. Or may not even understand - completely transforming AI from a rule-bound machine into a learning entity. This freedom allows it to come up with rules or pathways that we didn't even think were possible. Sometimes, we may not even understand its way of reasoning.

Machine Learning is undoubtedly the cornerstone of modern AI. But it only became possible quite recently because the tools used in training computers have become exponentially more powerful.

But Why?

The surge in Machine Learning can be attributed to advancements in computing power and a shift in AI training techniques. One pivotal factor is the transition from using Central Processing Units (CPUs) to Graphics Processing Units (GPUs) for AI tasks and training.

CPUs are like the brains of a computer. They are capable of handling a wide range of tasks but at a relatively slower pace. On the other hand, GPUs are designed to handle complex calculations simultaneously, making them ideal for parallel processing.

In effect, while a CPU would process tasks one after the other, a GPU can tackle multiple tasks at once, dramatically boosting efficiency.

This exponential growth in computing power has defied Moore's Law, which predicted a doubling of transistor density in integrated circuits approximately every two years. AI organizations like OpenAI have witnessed an astounding trend, announcing that the computing power needed to train the largest AI models has doubled every three months!

This exponential growth in computing power explains why AI can do insanely incredible things these days. But with great power, comes great responsibility, and this speed makes the promise of AI so huge and yet so scary.

Find out why in AI: The Nuances next week.


Ponder Goodies

Top AI tools to take advantage of:

Category: Design

  • Canva: Create social media posts, presentations, posters, videos, logos and more.
  • DALL-E: Create realistic images and art from a description in natural language.
  • Midjourney: Same functionality as DALL-E, but better
  • RemoveBg: Remove images from their backgrounds with AI
  • Adobe Firefly: Create stunning art from text, edit photos to be anything you imagine with generative fill, and so much more.


About Ponder

Ponder, my blog, is my attempt to share my random thoughts with you to inspire you, cause you to challenge your thinking and even make you laugh (Hopefully).

With ponder, you get to explore Minimalism, Christianity, Science, Graphic Design, an annoying appreciation of detail and out-of-the-box thinking. You get to explore all of these from the perspective of an African.

There’s more on my website: joshwordey.org/blog

To view or add a comment, sign in

Insights from the community

Others also viewed

Explore topics