Artificial intelligence or superficial lip service? How humans mandate governance over AI innovation now can seal our fate — and AI’s destiny
Despite significant advancements, AI — particularly generative AI (GenAI) — remains primarily in the “hobbyist” stage.
The reason for this perspective lies in the current state of AI development. While impressive, many AI solutions resemble high-powered gadgets – impressive feats of engineering, but ultimately impractical or too far-fetched for everyday use.
With businesses facing a multitude of challenges, and most mainstream AI solutions often falling short due to their limitations and programmed directives (to please their masters), what lies ahead and how can companies effectively scale their AI adoption?
Diminishing returns
GenAI has reached a point where massive models are no longer just the answer, particularly in domain-specific models like large language models (LLMs) where we have likely reached the point of diminishing returns from larger parameter counts.
While massive models have propelled the AI golden age, they come with drawbacks. Only the largest firms can afford to train and maintain energy-hungry models with hundreds of billions of parameters. One estimate asserts that a standard day of ChatGPT queries rivals the daily energy consumption of 33,000 households.
On the other hand, smaller language models democratize AI by lowering barriers to entry for individuals and organizations with limited resources. These models require less computational power and memory to tune and deploy. Likewise, their ability to run locally on smaller devices may help address privacy and cybersecurity concerns. Processing data on-device, rather than sending it to centralized servers, minimizes the risk of data exposure and unauthorized access.
Until now, GenAI applications have demanded substantial resources and energy. But downsizing need not interfere with performance. On the contrary, smaller models excel in specialized domains such as Finance or HR.
At first, harnessing the benefits of AI while going green seems to be two conflicting goals. However, this is not a pipe dream. It can be realized with the smaller language models that are entering the market and demonstrating competitive performance across multiple tasks, especially in specialized domains.
No single AI model fits all needs
The fact that small language models can give their large language counterparts a good fight, shows that specialization has its place in the development of AI. This means going beyond the generic, one-size-fits-all approach to a different future where every enterprise can deploy custom models that align precisely with their goals and regulatory requirements.
Enterprises need to recognize that understanding foundation models is crucial for optimizing AI initiatives, as the former constitute the backbone of such systems. However, more organizations also realize the importance of personalizing AI models to mirror their unique values and operational contexts.
For example, imagine a bank from Vietnam developing a customer service virtual assistant for the region’s customers. Using AI Singapore’s SEA-LION open-source large language model to complement its own customer service data can help the bank discern overseas-customers’ needs and deliver personalized responses —because it has tailored its foundational model to understand the language, cultural nuances and needs of its current and prospect clients.
AI with good governance is a threat
The future of AI begins and ends with better governance. Only with responsible oversight can AI adoption become widespread and trusted.
People are wary of AI. Well-known risks include more sophisticated privacy breaches and potential bias in AI’s recommendations and outcomes. Developing ethical AI policies; safeguarding data privacy through the entire lifecycle; tracking data provenance, changes in data and model versions are some of the most pressing dictums for enterprises and regulatory bodies to ensure use of trustworthy AI.
This mandate really cannot be emphasized enough. Without ethical principles embedded from the outset, AI truly has no future. This culture does not just belong in IT teams. It starts at the top with every CEO.
