#1. Understanding Generative AI (GenAI) in simple language.

Generative AI is revolutionizing the way we approach artificial intelligence, particularly through its ability to predict and generate text. Here are two fundamental concepts to grasp:

1. Neural Network Predictions: GenAI uses a complex neural network to predict the next word (often referred to as a token in AI terminology) based on preceding words.
2. Context Window: The prediction is influenced by a context window, which defines how many previous words the model considers. For example, in 2024, Meta developed a model with a context window of 128K words, meaning it can analyze 128,000 words before making a prediction.

Training and Achieving Artificial General Intelligence (AGI)

Companies train these models using vast amounts of internet data to strive for Artificial General Intelligence (AGI), which aims to replicate human reasoning. The idea is that human reasoning is based on experiences—data points collected and processed by our neurons. When reasoning or making decisions, neurons process these data points. Similarly, AGI systems require extensive data (experiences) and computational power to emulate general intelligence.

Challenges in Achieving AGI:

• Stress on Computational Systems: Just as intense learning stresses our neurons, training models require powerful computational systems and substantial energy. This is akin to the mental drain we experience during focused learning.

The Importance of Open Source Models

While making predictions (inference) is less resource-intensive, training models from scratch is not feasible for individuals due to the enormous computational and energy requirements. Therefore, it’s crucial for companies to provide open-source versions of their trained models. This allows individuals to fine-tune these models for specific domains without the need for extensive resources.

Domain-Specific Models:

• Training on the Job: Just as employees are trained on the job to enhance their college education for specific company tasks, domain-specific models can be fine-tuned for specialized areas. This approach supports the growth of AI and the normalization of GenAI in various business sectors through Small Language Models (SLM).

Conclusion

For AI and GenAI to thrive and become a standard in business, we need domain-specific models that can be adapted for various fields. This ensures accessibility and practicality, allowing more people to benefit from these advanced technologies without the need for extensive resources.

In the next blog, we will understand few use cases of GenAI. It will help us in building thought proces

---

If you’re interested in learning more about data science, AI, and GenAI, follow me on my social media handles. The views shared here are my personal opinions.