AI for all: The power and potential of democratising artificial intelligence
The concept of ‘democracy’ carries several interpretations. While the most popular one is a government representing all people, it also extends to things/resources that can be democratised and made available to the masses.The notion applies equally in the realm of digital technology. AI democratisation is a framework aimed at making the benefits of artificial intelligence accessible to all, even to those who do not have any knowledge of it.
What is AI democratisation? Why democratise?
The 20th century witnessed a surge in technological advancements and techies. This century, however, is all about data and intelligence and using them to recognise trends, make decisions, forecast, learn, improve, and above all, make data accessible to everyone in the organisation.
But the catch is that AI development demands a lot - resources, expertise, computing power, and money. In short, it is a technology that is limited in its reach for the rich and/or the tech-savvy population.
Others find themselves lagging - a situation that firms up the need for democratising AI. The three main benefits of doing so are:
- Reduced entry barriers for individuals and organisations: With little investment, they can both enter the world of AI and experiment with building AI models by accessing publicly available datasets.
- Accelerated adoption in the fields of business and academics: The use of NLP (Natural Language Processing) in businesses, ML (Machine Learning) and deep learning to augment human decision-making enables businesses to be more agile and hyper-productive.
- Lower overall cost of building AI solutions: Since data and many algorithms are available for free, developers can start extending them to build more powerful solutions.
Levels of democratisation
Which aspect(s) of the AI product adds the most value for the end user? This question must be answered by technology vendors before they release their products in the market. The gamut of AI democratisation includes data, algorithms (pre-trained models), data computing and storage, and the marketplace.
Data: It is the raw input for ML from which insights are generated, decisions made, and outcomes predicted. This level is easy to democratise, enables basic analysis of datasets and also helps in training AI models quickly.
Algorithms: At this level, data gets engaged. Democratising algorithms, though more complex than data democratisation, is still easy. However, it needs a basic understanding of computer science, statistics, and mathematics to make sense of the algorithm. Platforms like GitHub and Hugging Face boast of hundreds of AI code repositories and pre-trained models which provide a jumpstart and overall acceleration to the AI journey.
Data Computing and Storage: At this point, democratisation starts to become more complicated. With major AI/ML providers like Amazon Web Services (AWS), Google Compute Platform, Microsoft Azure, and IBM Cloud, transitioning to cloud-based platforms, access to essential hardware components like central processing units (CPUs) and graphics processing units (GPUs) has been simplified. Additionally, advanced storage solutions like Google Cloud Storage and Amazon S3 facilitate easy and secure data storage and management.
Marketplace: The democratisation of the marketplace is still in its early stages, indicating untapped opportunities awaiting exploration. Platforms such as Hugging Face Leaderboard, Kaggle, Algorithmia, and AI Hub are thriving AI communities of ML practitioners. These communities regularly conduct open challenges that promote the creation and enhancement of Large Language Models (LLMs).
Much like the printing press democratised information and the internet democratised knowledge – AI and Generative AI, representing the present and future of technology, are on this path now. The full potential of AI is yet to be realised and democratising it is the path forward to unlocking its true capabilities. As Kay Firth-Butterfield, the head of AI and ML of the World Economic Forum aptly noted, “The long-term success in the development of AI will depend on the agility of the collaboration, the diversity and integrity of the data, and the accuracy of the risk assessments.”
This article was first published on Business Insider