How Artificial is your Intelligence?

Technology moves at a non-human pace, and the adoption of new tech follows a non-linear trajectory. Nearly two decades passed from the introduction of the first Nokia SMS-capable flip phone in 1997 to the omnipresent pocket supercomputers we all carry today. People transitioned from "I don't need a mobile phone" to "I can't leave home without a smartphone." This shift has predominantly been positive, especially in developed nations and, to a lesser extent, in the developing world. Across all areas, smartphones and the internet have significantly empowered internet technologies, making lives better, easier, more productive, and agile. While there are inevitably darker aspects to human endeavors, let's focus on the positives.

Today, we find ourselves on the steeper part of the technology adoption curve—mass adoption of new technologies no longer takes two decades but instead a couple of years. The mass adoption of AI technologies serves as a clear example. In a few short years, we will all carry AI companions that will be as indispensable to us as smartphones are today. AI, in its various forms, will also be omnipresent and critical in enterprise, government, education, agriculture, and beyond.

In the face of this rapid sea of change, the question arises: how can you adapt? Are you relying solely on your own intellect to perform your job, or are you willing to embrace early adoption? Take a moment to reflect in the mirror and ask yourself, "How artificial is your intelligence?"

A good initial step is to become proficient in utilizing these new AI tools effectively. This burgeoning area is now termed "Prompt Engineering," the practice of providing high-quality "Human Intelligence" inputs to your AI companion to yield high-fidelity results. After all, concerning data-trained machine learning models like Large Language Models (LLMs), the term "intelligence" is somewhat liberal. These applications do not truly "think"; that aspiration belongs to other AI domains like Artificial General Intelligence or deep learning, which are far less mature than machine learning today.

LLM models may appear to reason, but in reality, they merely make educated guesses at best. They lack common sense and understanding of real-world contexts. They are akin to the "auto-complete" function on steroids in word processors. Auto-complete predicts your next word or phrase based on your typing history. LLMs take this concept several steps further; you define the topic, goals, purpose, and style you desire in an article, along with a list of topics to cover, and they "guess" from an extensive database of text, images, and videos to produce the best possible three-page article, complete with grammar, punctuation, and style that aligns with your preferences—a true time-saver.

The encouraging news is that a skilled "Prompt Engineer" resembles a seasoned pilot. If you possess knowledge of the roads, the nuances, and the tacit knowledge of your areas of expertise or industries, you're well-positioned to leverage these new AI models to their fullest potential. You simply need the right toolkit! So, dive in and start experimenting with various models, tools, and datasets. Avoid becoming a one-trick pony (yes, I'm talking to you, ChatGPT-only fans). As impressive as OpenAI and ChatGPT are, they are not the sole players in the field. Depending on your work nature, different models trained on distinct datasets may be better suited for certain tasks. Models like Claude or conducting internet searches with Perplexity.io, which connects to multiple LLMs, are examples. And if you're a geek like me, you can begin using open-source foundational models to train models on your own data—many of them are open source, such as Llama2, Palm 2, and Falcon 40B.

So, at the next cocktail party, don't hesitate to pose the question to your friends, "How artificial is your intelligence?". Just make sure they do not excuse themselves to the bathroom to get an answer from ChatGTP...off their Smartphones ;)