© The Indian Express Pvt Ltd
Tags:
Journalism of Courage
A volcano in Iceland erupted on Saturday (March 16) for the fourth time since December, the Icelandic Meteorological Office said, spewing smoke and molten lava into the air. One question that arose immediately was: Will it impact air travel? A 2010 eruption in Iceland had halted around 100,000 flights in Europe as ash clouds and haze enveloped the skies around the Arctic Circle, a key international flight path.
That’s where data analysis for pattern searches using Artificial Intelligence comes in. Beyond commercial applications, Moscow-based Yandex has introduced a service capable of monitoring volcanic ash movement and mitigating its impact on communities and ecosystems.
Given that volcanic ash presents a significant hazard, affecting regions across the globe with far-reaching consequences, Yandex, using advanced mathematical models and neural networks, has developed an interactive map that allows the real-time monitoring of ash clouds after eruptions. The idea is to empower authorities and communities to respond swiftly to ashfall and safeguard public safety and infrastructure.
According to Anna Lemyakina, Director of Strategic Projects at Yandex Cloud: “In projects such as forecasting volcanic ash dispersion, seamless and swift access to services for hypothesis testing and model training is crucial. Our project is readily scalable to monitor volcanoes worldwide, addressing the urgent issue of volcanic eruptions and their aftermath.”
Even as text-based generative AI platforms such as OpenAI’s ChatGPT and Google’s Gemini have caught the fancy of people, companies and researchers around the world – including oil and gas majors, pharmaceutical companies, and manufacturing entities – are increasingly relying on generative AI for all kinds of purposes, including oil exploration, drug discovery and worker safety, among others.
For companies, the advantage is the trove of their own historical data which can become useful fodder for generative AI tools for predicting things. For instance, an oil drilling company with wells around the world has the historical geological data on the regions where all oil drilling has led to sucessful finds, and an AI system trained on this historical data could predict where a new oil well could be located.
Although, given the hype around generative AI, the term has also become a marketing buzzword. For instance, a number of digital wearables’ companies have repackaged some of their personalisation options as AI-first features. While the personalisation basis of a user’s typical habits is still a function of AI, it is nowhere near the sophistication – both in hardware and software – that generative AI platforms need.
A majority of the top 20 global oil and gas producers are learnt to be working on an AI strategy for their upstream (exploration) and midstream (pipeline and logistics) businesses. A recent EY survey indicated that more than 92 per cent of oil and gas companies around the world are “either currently investing in AI or plan to in the next two years.”
AI algorithms are being programmed to seek solutions for a variety of desired outcomes, especially for data-led interventions to sift through the large amount of data generated by past surveys and explorations to identify patterns and correlations that may escape other forms of analysis. AI for oil and gas can leverage the data produced by active wells during extraction to make predictions about probable reserves, provide predictions about the best ways to access known reserves, and extrapolate the lifetime yield of current wells.
Take for instance oil major Shell. Last year it forged a partnership with the AI software service company SparkCognition to develop a proprietary generative AI tool which can generate subsurface images for potential oil exploration opportunities.
As per the company, the traditional approach to subsurface imaging and data analysis is time-intensive and costly, relying on terabytes of data, high-performance computing, and complex physics-based algorithms to analyse and identify exploration opportunities.
However, the generative AI solution being developed by Shell and SparkCognition uses deep learning to generate reliable subsurface images using far fewer seismic shots—as little as 1per cent in completed field trials—than traditionally necessary while preserving subsurface image quality. The company says this will allow it to explore resources at a much lower cost.
Earlier this month, Saudi Aramco, the world’s largest oil producer, showcased its metabrain generative AI. According to Amin H Nasser, the company’s president and CEO, metabrain is helping Aramco to analyse drilling plans and geological data as well as historical drilling times versus costs and recommend well options. The model will also have the capability to provide precise forecasts for refined products, including pricing trends, market dynamics, and geopolitical insights.
Then there is the field of applying deep neural networks in drug discovery, which happens to be one of the most promising areas of research currently. “Predictive models are central to our work,” according to Friedrich Rippmann, Director, Computational Chemistry & Biology at Merck. “These are statistical models that predict whether a compound idea – a not-yet-synthesised molecule – will produce a desired activity.”
“The technologies we’re using mostly relate to machine learning. In particular, we’re using various types of deep neural networks. But we’ve also explored other more classical statistical techniques, with funny names like random forests and support vector machines.”
So far, in terms of practical benefit, the research carried out by Friedrich Rippmann’s team and Merck’s partners has made available almost 300 new models for assessing the properties of a compound, which can help predict their ability to bind to a specific disease-relevant target. “These models are already being used by our chemists to judge their compound ideas before deciding on whether to synthesise them,” according to Rippmann.
AI analysis is only as good as the quality of the datasets in use. For this reason, the pharmaceutical industry is increasingly seeking to collaborate by pooling data. A recent initiative to facilitate, the ‘MELLODDY Project’, involves the EU Innovative Medicines Initiative and around ten pharmaceutical companies in a project aimed at improving predictive models through so-called ‘federated learning’, by using a novel blockchain system to store data on a secure ledger while protecting the trade secrets of individual companies.
