The Confusion Stops Here: AI Explained

I’ve been in tech for a long time, and I’ve never seen anything shake up the tech world like the emergence of AI. Along with the excitement around these new tech breakthroughs comes confusion about what AI can and cannot do.

This confusion is rooted in a lack of understanding about how it works. I recently read a book about AI, AI Snake Oil (A. Narayanan & S. Kapoor, 2024) that beautifully articulates the problem using this analogy:

So, let's demystify its magic so we can use, build, and lead AI into the future.

What exactly is Artificial Intelligence?

At its core, AI is a branch of computer science that enables machines to perform tasks traditionally requiring human intelligence - such as learning, reasoning, problem-solving, and decision-making. However, AI has evolved beyond its simple definition, capturing the public's attention and becoming a focal point of global curiosity, innovation, and contention.

At one time, spellcheck and autocorrect were considered complex AI solutions. Today image generators, like Midjourney and Dall-e, and chatbots, like ChatGPT and DeepSeek, are cutting edge AI but as it continues to advance AI will be redefined again as new functions are rolled out.

There are dozens of flavors of AI, just like in the vehicle analogy, each developed for specific functions. The most talked about and misunderstood types are Generative AI and Predictive AI.

Let's Focus on Generative AI and Predictive AI

Generative AI

What it does: Creates new data (text, images, audio, etc.) based on patterns in training data.
Key goal: Creativity and content generation.
Examples:

• ChatGPT generates human-like text.
• DALL·E creates images from text descriptions.
• Music generation models compose new songs.

How it works: Often uses models like Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), or Transformer-based models like GPT.

Predictive AI

What it does: Analyzes historical data to predict future outcomes or trends.
Key goal: Decision-making and forecasting.
Examples:

• Predicting stock prices or market trends.
• Weather forecasting models.
• Recommendation systems (e.g., Netflix suggesting what you should watch next).

How it works: Uses techniques like regression, decision trees, or deep learning (e.g., LSTM or time-series models) to identify patterns and make predictions.

How does Machine Learning fit in?

Machine Learning is a branch of AI that enables computers to learn patterns from data and improve their performance on tasks over time without being explicitly programmed. In essence, it's the algorithms and models that process data. ML is used in both Predictive and Generative applications (and all of the others).

Similarities in Implementation

Data Collection: Both types need large datasets to learn from.

• Generative AI: Text, images, music, etc.
• Predictive AI: Historical data, customer behavior, market trends, etc.

Training Process: Both train models to minimize errors.

• Generative AI: Learns to generate outputs that resemble the training data.
• Predictive AI: Learns to make accurate forecasts or classifications.

Pattern Learning: Both aim to identify relationships in the data.

• Generative AI: Learns creative structures (e.g., grammar, color schemes).
• Predictive AI: Learns patterns for prediction (e.g., price fluctuations).

Model Use: After training, both apply learned patterns to new input.

• Generative AI: Produces new, original content.
• Predictive AI: Produces insights or forecasts.

See, AI isn't Scary. But there are a couple of things to watch out for. 

1. Predicative AI still doesn't work very well. 

Unfortunately, it's being sold as a "Magic Wand" to replace complex analysis used in decision making. At this point, there is very little data to back up results of many of these tools.

2.  AI models and ML algorithms are a blackbox.

Creators of these tools have immense power to manipulate the output without any accountability.

We need regulation and ethics around AI should focus on these pitfalls, instead of fears of it convincing your spouse to divorce you or worse AI waking up and taking over SkyNet.

Wrap Up

You shouldn't be afraid of AI , it's our friend. There's nothing scary about it and the migration is happening as we speak. There's incredible research and development going on right now creating new and exciting jobs that will need highly experienced technical expertise. Keep learning, my friends to be part of the tech future.