In an ever-evolving digital landscape, having a robust framework is pivotal for any business or project. Today, I’d like to introduce you to a framework encompassing six core elements that we’ll explore in detail. These are: Mindset, Team, Objective, Data, Models, and Infrastructure.
- The Mindset Shift
Changing our mindset demands that we let go of traditional methods and embrace new ways of working. Just as the advent of the internet necessitated skill upgradation, so does the AI revolution. I recall the drastic shift in my previous company when Mark Zuckerberg steered Facebook towards a “mobile-first” approach. Engineers had to learn not just new programming languages but also a new way of working. The same transition is visible with AI. Over the past decade, numerous companies have adopted an “AI-first” strategy and are already reaping its benefits. Essential to this shift is recognizing the opportunities and risks that vary based on the industry. To initiate change, launching pilot projects can be instrumental, offering tangible outcomes from theory to practice. This change demands adaptability; without it, transitions become convoluted.
- Team Assembly
In terms of resources and competencies, domain experts collaborating with various professionals, both technical and non-technical, are crucial. From system design, database management, data cleaning, model development, to deploying and maintaining infrastructure - be it cloud or on-premise - various roles beyond data scientists are vital for AI technological development.
- Define a Clear Objective
Once the pilot project and the team are in place, a clear and shared long-term goal is paramount. Drawing inspiration from external AI systems can be beneficial. Tasks for the AI system need clear definition, and potential risks require strategies for mitigation. While setting these in stone isn’t mandatory - with feedback, objectives and tasks can undergo necessary modifications.
- Data – The New Oil
Identifying data formats (be it images, text, or tabular) and sources (internal, external, open data, etc.) for model training is essential. Data quality and legal assessments, considering ethics and privacy, are pivotal. Notably, with the data explosion facilitating the growth of “big data”, AI systems thrive on clean, quality data. Hence, industry experts emphasize “Good Data” over “Big Data”. The Economist in 2017 aptly pointed out that data has surpassed oil as the world’s most valuable resource. Consequently, countries like Saudi Arabia are diversifying their economic reliance on oil, investing heavily in technology like AI.
- Model Selection
In AI, the model selection phase is where the blueprint for problem-solving takes shape. This phase evaluates algorithms based on the specific problem at hand. It could fall into one of the categories: supervised, semi-supervised, unsupervised, reinforcement, or the increasingly relevant, self-supervised learning
- Infrastructure
We won’t delve into the technicalities here, but it’s worth noting that creating the model is just the beginning. Ensuring system scalability and maintaining accuracy involve several essential post-creation activities. Deciding on updates with new data becomes crucial, with platforms like PagoPa’s Data Lake serving as infrastructure examples.
In conclusion, this six-element framework provides a holistic view, emphasizing that a harmonized blend of the right mindset, team, objectives, data, models, and infrastructure is key to leveraging AI’s transformative potential.