Information is not knowledge, knowledge is not wisdom, and wisdom is not foresight. Each grows out of the other, and we need them all.
– Sir Arthur C. Clarke
Noted science writer and doyen of the science fiction world who was also a keen proponent of space travel provided us with these famous words. Fame aside, they are proving to be an edict for the fast evolving world of technologies and tools.
No other domain is more appropriate and yet understated for an Artificial Intelligence (AI) driven innovation than EHS. AI is a transformational technology that harnesses machine or systems based intelligence instead of natural intelligence. This technology-learning method-complexity exhibit by McKinsey Global Institute accurately summarises as to what constitutes AI applications:
A simple definition that explains the elements described above lie within the premise of higher mathematics, algorithms, networks learning and statistics. What we’ve got to understand is how different sectors are using AI to leverage the speed, accuracy, and foresight on offer via these elements. In case of EHS, the difference lies in understanding the consequences of AI-driven decisions. The litmus test which may determine the shape and function of EHS frameworks, though not yet understood, concretely.
However, our immediate limitation of understanding shouldn’t be treated as gaps in knowledge, instead, they should be viewed as periods of opportunity. The opportunity of using our current on-ground technologies in EHS that can offer an effective piggyback learning mechanisms for the future AI – sample, aggregate and learn from the EHS ecosystem.
The piggyback learning mechanisms – mobile apps, proactive digital reporting platforms and knowledge repositories – armed with the latest JSA’s and updated Risk assessments offer the data driven chance to become the very foundations of future AI within the EHS domain. Let us now look at how AI tackles challenges in the real world.
Current breed of Problem Type functions commonly used within AI:
Classification – Image or pattern recognition from datasets.
Continuous estimation – Data dense operations ‘predicting’ or ‘forecasting’ what the future might hold based on curated precursor data.
Clustering – Problem sets with similar characteristics (for e.g. incident/accident – occupational health) recognizing, sharing and generating overall insights for EHS networks.
All other optimization – Single point or multi-point (trait – characteristic) optimization that is ultimately serving a specific objective. E.g.: Effective asset performance management by combining Permit to work-LOTO, Change management and EHS knowledgebase (without human intervention).
Anomaly detection – Error and failure point detection within a system based on the data it generates.
Ranking – Exercise in labeling the ‘top’ or ‘priority’ items in a data (information) retrieval scheme which is generally seen on your webpages while browsing at a retail marketplace and ‘suggested items’ pop up.
Recommender systems – Such systems provide recommendations based on the sets of training data.
Data generation – These systems dabble in novel data generation from the sets of data that has been provided to them.
There already exist several instances of using these AI problem solving systems, such as optimization in logistics, ranking and recommender systems in digital retail and anomaly detection in machine operations.
In case of EHS, the applications that will determine the usefulness of the final product will need a constant supply of data variety and volume. Current digital EHS systems fit this bill perfectly as they are acting as the information conduits of the EHS ecosystem, now.
The other factor that will allow AI to positively disrupt EHS would be partnerships. Now, these can be partnerships between the EHS automation providers and AI specialists or between the industry solution seeker and specialists. Regardless of the modality, such knowledge partnerships will need to align with the vision of data variety and volume.
In simple words, the more diverse and in-depth stories datasets can narrate, the better its corresponding AI functionality and response. These would especially be needed within EHS due to the simple fact that safety and health of human and mechanized assets is on the line.
To credibly express our opinion, there is a very strong likelihood of an AI based approach to become part of the native functionality within the EHS domain. The caveat being, to let AI researchers do their jobs and let EHS personnel theirs while exploring for opportunities of partnerships, actively.