Generative AI: Lack of Information Architecture increases misinformation risks

Imagine AI as a powerful engine. It can take you anywhere, but without a proper roadmap (Information Architecture), you're likely to get lost. Thankfully, businesses are overflowing with data – the fuel for this incredible journey.

This year has been a whirlwind for Generative AI. As we’ve moved through 2024, upskilling in GenAI has surged, and here are some of my personal insights from the journey.

Q1: It was like the early days of the internet – everyone was buzzing about chatGPT, but we were still figuring out the basics. We even learned about "hallucinations," where AI confidently makes things up! (Think of it like a mischievous storyteller who gets carried away.) Few top items which I learnt:

There is a great course to understand in general which doesn’t require any prior knowledge - https://go.nocode.ai/30-day-email-course

Prompt engineering and Tuning -

https://www.deeplearning.ai/short-courses/chatgpt-prompt-engineering-for-developers/

https://www.dhirubhai.net/learning/prompt-engineering-how-to-talk-to-the-ais

https://www.dhirubhai.net/posts/armand-ruiz_how-to-prompt-engineer-like-a-boss-here-activity-7239226197732118528-3A0s?utm_source=share&utm_medium=member_desktop

Q2: The excitement grew as we discovered how to use LLMs for text retrieval to find specific information (like a skilled librarian) and started discussing responsible AI – ensuring this powerful technology is used ethically.

https://github.com/IBM/watson-machine-learning-samples/blob/master/cloud/notebooks/python_sdk/deployments/foundation_models/RAG/Use%20watsonx%2C%20Chroma%2C%20and%20LangChain%20to%20answer%20questions%20(RAG).ipynb

https://www.ibm.com/think/topics/responsible-ai

Q3: No-code solutions for text retrieval became the hot topic, like a shortcut to finding the perfect book in a massive library. Small Language Models (SMLs) also emerged, offering more focused and efficient AI assistants.

https://developer.ibm.com/tutorials/awb-watsonx-prompt-lab-build-rag-app-llama405b/?sf198248324=1

https://developer.ibm.com/tutorials/awb-build-rag-application-llama405b-watsonx/https://www.dhirubhai.net/pulse/generative-ai-planning-analytics-watsonx-akram-ali-1emwc/

?

Q4: InstructLab changed the game. It's like teaching the AI your personal preferences, making it a truly customized and helpful companion. (If you're on a Mac or running Fedora, give it a try today!)

https://docs.instructlab.ai/

This has been an incredible year for Generative AI. We still need to examine the unit economics in the new year but according to me the top ticking box for organisations in 2025 will be, governance on the existing Data Science pipelines and new projects and there was a huge announcement of from IBM on WatsonX.ai being available in ANZ in Q4 2024 and more importantly how the small language models (SML) in agentic AI will shape enterprises in this AI journey.