Simply put, GenAI/LLM is fundamentally based on tokenizing information to estimate associations and similarities between tokens (weights and biases). Tokens are discrete units, but our world and human experiences are continuous. If you understand how things work inside GenAI then you should be better positioned to tell the differences between the current hype and the real abilities/limitations of GenAI. It’s really fun and can be useful for video/audio/text generation though.

However, if you are on Linkedin or Youtube, it’s not hard to find many self-claimed GenAI experts are clearly fueling the hype instead of communicating the true abilities & limitations, either intentionally or unintentionally. Why? Glory and fortune? But how long can this last?

[The above is human (myself) generated content]

Below is Google Gemini’s answer to a question of mine.

I find it useful.

Question: New self-claimed AI experts often seem to only focus on GenAI and ignore the classic machine learning use cases and operations already in place for decades. Why?

Answer from Google Gemini:

1. “Wow Factor” and Media Hype:
   • GenAI is more tangible and visually impressive: Tools like ChatGPT, Midjourney, and Sora can generate human-like text, stunning images, and even videos from simple prompts. This immediate and creative output is incredibly engaging and easy to demonstrate, making for great headlines and social media content.
   • Traditional ML is often “behind the scenes”: While foundational and critical, classic ML applications (e.g., fraud detection, predictive analytics, recommendation systems, spam filters) often work silently in the background, optimizing processes or making predictions that aren’t as directly visible or “creative” to the average person. Their impact, though immense, is often less immediately “shareable” or exciting.
2. Lower Barrier to Entry (for using GenAI):
   • User-friendly interfaces: GenAI tools are designed for broad accessibility. Anyone can open a browser, type a prompt, and get a result. You don’t need to write code, understand algorithms, or manage data pipelines to get started.
   • Traditional ML requires deeper technical skills: To implement classic ML, you typically need to understand data preprocessing, feature engineering, model selection, training, evaluation, deployment, and ongoing monitoring. This requires programming skills (Python, R), knowledge of libraries (scikit-learn, TensorFlow, PyTorch), and statistical understanding.
3. Perception of Novelty and “The Future”:
   • GenAI feels cutting-edge: It represents a new frontier in AI, pushing boundaries in content creation and human-computer interaction. It feels like “the future” of AI.
   • Traditional ML feels “established”: While continuously evolving, the core concepts and many applications of traditional ML have been around for decades. Some might perceive it as “old news” or less innovative, even though it forms the bedrock of most current AI applications.
4. Simpler Narrative for Non-Technical Audiences:
   • GenAI is easy to explain: “It creates new stuff!” is a simple and compelling message.
   • Traditional ML is harder to simplify: Explaining the nuances of a gradient boosting model for churn prediction or a CNN for image classification is far more complex for a general audience. The “experts” targeting business leaders or the public find it easier to discuss GenAI’s transformative potential without getting bogged down in technical details.
5. Monetization and Consulting Opportunities:
   • Immediate business buzz: Companies are scrambling to understand and implement GenAI, creating a huge demand for “experts” who can explain it, consult on strategies, and lead workshops. This often translates to higher speaking fees, consulting rates, and job opportunities compared to the more mature and often specialized traditional ML market.
   • “Disruption” narrative: The idea that GenAI will “change everything” appeals to businesses looking for a competitive edge, leading them to seek out anyone who claims to understand this “disruptive” force.

It’s important to clarify: Generative AI is a subset of machine learning, typically utilizing deep learning architectures (like transformers for LLMs) trained on massive datasets. It builds upon decades of ML research and breakthroughs.

The danger of this GenAI-only focus by “self-claimed experts” is that it can lead to:

• Misapplication of technology: Trying to force GenAI solutions onto problems where traditional ML is more efficient, accurate, or robust (e.g., using an LLM for simple classification tasks when a classical model would suffice with less computational overhead and higher reliability).
• Ignoring existing value: Overlooking the massive, proven value that traditional ML systems already deliver across industries for critical operations like fraud detection, credit scoring, demand forecasting, quality control, and medical diagnostics.
• Lack of foundational understanding: Without a grasp of core ML principles, it’s hard to truly understand the limitations, biases, and ethical implications of even GenAI models, which can lead to unrealistic expectations or dangerous deployments.

Ultimately, genuine AI expertise requires understanding both the novel capabilities of GenAI and the robust, long-standing power of traditional machine learning, and knowing when and how to apply each appropriately.