AI can improve public health
Published:Wednesday | October 11, 2023 | 12:05 AM
THE EDITOR, Madam:
As the adage goes, “It’s all fun and games until someone gets hurt.” Such is the case with the recent and alarming incident that saw over 60 primary-school students hospitalised due to consuming THC-infused candies – products that looked strikingly similar to regular sweets.
This grave episode underscores an urgent need for innovative solutions to ensure public health safety, particularly concerning consumable goods.
At StarApple AI, we believe artificial intelligence (AI) can play a crucial role in eliminating such threats, specifically through the use of multimodal AI models.
AI MODELS CAN SEE
Multimodal AI models are designed to analyse, interpret, and generate insights from multiple types of data, such as text and images. Their multifaceted capabilities make them ideal for complex tasks like identifying harmful substances in seemingly innocent candies, based on visual cues alone.
AI DETECTS CANDY THREATS: OUR STUDY
Our specialised research group, Section 9, conducted an experiment using GPT-4 – a state-of-the-art multimodal model – to address this pressing issue. We tested 24 popular candy brands to discern which could potentially be harmful to children due to the presence of THC. The AI model achieved an impressive accuracy rate of 88 per cent – a remarkable feat, considering the complexity and variability in candy packaging.
While human testers yielded a slightly higher accuracy rate of 92 per cent, the near-parity in performance indicates the immense potential of AI-driven solutions in ensuring public health safety.
BEYOND CANDY: AI’S EXPANDING ROLE
The applicability of multimodal AI models goes far beyond identifying harmful candies. These systems can also analyse packaging for compliance with local regulations, generate safety recommendations, and more. As AI technologies evolve, their role in public health safety is poised to expand significantly.
A GLOBAL CONCERN
It’s imperative to note that the issue of misleading packaging is not localised – it’s a global problem. Since 2019, incidents of THC poisoning have been reported in multiple countries, including the US, Canada, and the UK. Hence, AI models like the one developed by Section 9 can serve as an invaluable tool for public health workers, regulatory bodies, and citizens alike.
THE INTERSECTION OF AI AND PUBLIC WELFARE
We stand at a pivotal moment where technology and public health are intersecting in meaningful ways. The work conducted by Section 9 is just a glimpse into the transformative power of AI in addressing real-world challenges that impact public safety.
In summary, as we continue to witness advancements in AI, its role in safeguarding public health is not just promising – it is inevitable.
ADRIAN DUNKLEY
AI developer and researcher
CEO, StarApple AI