Usually the FT Tech for Growth Forum looks at technology within the business context. This report, nonetheless, examines how next-generation artificial intelligence generally is a force for good.
We will take a look at the ways wherein AI can reduce humanity’s impact on the planet and convey about societal improvements. Our focus this time is on agriculture, healthcare and environmental conservation.
The damaging effects of mankind’s exploitation of the earth’s resources have gotten clear. Our use of energy could have to be less polluting if we’re to carry the rise in global temperature inside the limits set by the Paris Agreement in 2015.
Climate change has caused extreme weather events to develop into more frequent. Germany, Austria, Hungary and Spain have all suffered storms and floods. In the US, Hurricane Helene caused severe damage much farther north than is common and the LA wildfires which broke out in January 2025 were exacerbated by unusually dry weather for the season.
Managing resources, notably food and water, will develop into increasingly essential, especially because the world’s population continues to rise. The Covid-19 pandemic and its effect on national economies showed the importance of maintaining a healthy global population.
The way forward for agriculture
The unpredictability of the weather has not only caused acute lack of life and assets, it’s having a chronic effect on farm yields on every continent. The pressure on agriculture is anticipated to extend.
With population growth highest in a few of the lowest-yielding geographies, “feeding the world” shall be more difficult than ever.
This yr Professor Patrick Schnable, who directs the Plant Sciences Institute at Iowa State University, told the BBC: “All projections show major losses in crop yields resulting from climate change. A ten per cent or 20 per cent reduction in corn yields could be catastrophic.”
Increasing agricultural yields shall be essential. This is where AI can assist: improving farming techniques will lift output in a single area at the same time as one other suffers from storms, heat or floods.
We must contain climate change in an effort to avoid exacerbating the results of maximum weather. Lower emissions may be achieved if farming methods change. The UN Food and Agriculture Organization says more judicious farm management could offset as much as a 3rd of agricultural emissions.
Water, water all over the place
Our use of water also must be rethought. Lakes and rivers are already drying up resulting from the mixture of climate change and our careless approach to water usage. These are usually not singular issues that affect some countries and never others: within the US alone, some water sources are disappearing at the same time as floods threaten more communities than before.
Demand from agriculture, commerce and consumers — all used to unrealistically low prices — combined with reckless management of a resource too often considered inexhaustible, implies that water usage doesn’t reflect the realities of its distribution, especially in developed markets.
For instance, drought-prone California is the best agricultural state within the US nevertheless it relies on intensive irrigation for crops resembling nuts. A 2024 study, Field-scale Crop Water Consumption Estimates, used data and machine learning and concluded that the Central Valley’s demand for water could fall by 93 per cent if farmers switched to less common crop types.
Such a technique is unrealistic since it would cut back profitability but the potential for letting some fields lie fallow could make a difference and be more commercially acceptable.
The issue, in California not less than, is prone to be forced with the renegotiation of the operating guidelines for the Colorado River Compact. The details of this 1922 treaty between seven US states are resulting from expire in 2026. Squabbles are breaking out over which state must reduce usage and the way much each should pay.
In central California, the San Joaquin “breadbasket” valley experiences many problems related to an excessive amount of water being taken, starting from drinking water contamination, to species and habitat loss, to general land subsidence.
Overall demands on farmland
The demands we place on farmland are huge. Although overall agricultural output continues to grow, the speed of expansion is slowing as overworked soils struggle to provide. The higher news is that technology and improved efficiency are helping to maintain output afloat. Farmers can have to depend on technology increasingly: farm labour forces in mature markets are declining, by as much as 2 per cent by 2033 within the US.
Regenerative agriculture could help, with data gathered through the same approach to that taken by the “field-scale crop water” study. It could guide farmers to aim for greater sustainability, encouraging a holistic approach based on replenishing the goodness in soils.
Supporters of this concept say higher soil will each draw down carbon and store more water. Coupled with AI-directed irrigation of plants, this might make water go further and reduce farming emissions.
Drones, sensors and phone apps all help
Technology powered by AI is behind sophisticated “precision agriculture” in markets resembling the US. It provides farmers with detailed information in order that they’ll goal areas of cultivation that need greater care.
Image evaluation has a use at each macro and micro levels. Armed with data from drones and tractor-mounted sensors, machine-learning algorithms can monitor crops and supply advice — for example that nutrients are short in a specific a part of a field — while phone apps resembling Agrio help with pest identification.
As the agricultural labour force diminishes, AI can fill the gap in some ways, taking over repetitive or structured tasks resembling sorting potatoes, planting or weeding. Equipment from John Deere, the machinery maker, offers automatic guidance on managing seed and nutrient placement and it provides updates on equipment performance. The Smart Cultivator from Stout, a Californian AI company, weeds fields without affecting crops. It claims to have 99 per cent accuracy.
AI also has a job in optimising the crops themselves. The AI Institute for Resilient Agriculture, founded in 2021 under the banner of Iowa State University, goals to enhance the resilience of farm systems by creating digital twins to model plants and fields.
In aquaculture, Bosch Business Innovations uses microphones and AI to analyse the behaviour of shrimps. Its system can detect when the crustaceans are hungry, allowing for precise feeding. Additionally, it could actually avoid overfeeding and flag up signs of illness, which allows early intervention. Productivity has doubled for farmers who use this technology.
Such initiatives are usually not limited to advanced markets. India, the world’s largest producer of milk and pulses, has embraced AI in areas from irrigation to pest management. Given the fragmented nature of the sector in a rustic with smallholder farms and 1,600 languages and dialects, technology is best served in ways which can be locally relevant and digestible.
Raising yields, improving quality and higher market prices
Digital Green, a world development organisation based in San Francisco, is working to realize this in India, Kenya and Ethiopia. Its multilingual AI chatbots may be operated by agents and deliver specific advice to individual farmers.
The company is behind Saagu Baagu, an initiative in Telangana, south-east India, run under auspices of the World Economic Forum. This has raised yields for 7,000 chilli farmers, reduced use of resources and improved quality, pricing and incomes. KissanAI also offers access to voice-based knowledge about pests and crop management.
In other regions, advanced insights may be hard to deliver with consistency. David Bergvinson, founding father of BeSustainable.io and a former leader of the digital agriculture division of the Bill & Melinda Gates Foundation, says poor infrastructure in some parts of Africa limits the effectiveness of “real-time decision-making based on weather data”.
Yet such information is more critical where vast tracts of land rely on rain. He says: “Agriculture is a weather-driven sector. Having weather (information) as a driver for making appropriate decisions to extend productivity, but additionally to administer risk and time market entrance, becomes very powerful.”
Always take the weather with you
Tomorrownow.org is developing and expanding weather-based advisories in Kenya with the backing of the Gates Foundation. Its aim is to enable African farmers to develop into more climate resilient by providing access to machine learning-based weather information from Tomorrow.io via ways resembling SMS. This access is starting to make a difference in Kenya, where alerts help farmers make informed decisions.
The pressure on resources, particularly water, may change the way in which we operate our food system, Bergvinson says, moving it away from productivity towards “nutrition and profitability per drop (of water)”.
The excellent news is that AI can optimise how we use water for food production. It will enable us to plan a plan of action to mitigate and adapt to climate change. Increasingly accurate modelling helps us to discover previously missed aspects resembling the effect of cloud phase. It gives us a greater grasp on the complexity and pace of change.
Broad access to AI implies that innovation is not any longer the preserve of the expert. Dryland agriculture in semi-arid areas, from India to Africa to California, can all be enhanced by predictive technology.
Environmental conservation
Connected to the demand for each farmland and resources, global deforestation is extensive, with most occurring in Brazil and Indonesia.
Logging and land clearance for subsistence and profit exacerbates floods, landslides and drought and is alleged to be liable for as much as 20 per cent of greenhouse gas emissions. Habitat destruction and poaching put the survival of various species unsure, threatening the biodiversity that defines a healthy ecosystem.
Conservation is the flip side of the exploitation coin — and AI may be used to shape strategy.
According to a 2023 Nigerian study published within the World Journal of Advanced Research and Reviews, AI can monitor ecosystems and discover rare or endangered species through sound and image evaluation.
It will even warn about environmental problems as they emerge.
Satellite images, that are trained on models, help to sustain the health of habitats by identifying areas where there is illegitimate logging in addition to those in need of restoration. The scale and accuracy are greater than may be achieved by manpower alone. Through evaluation of migratory patterns and human activity, AI algorithms can predict where poaching might occur, helping conservation teams to react with greater effect.
The International Union for the Conservation of Nature highlights the usage of video evaluation to pick rare species from hours of footage and to preserve biodiversity by flagging environmental issues. In China, satellite imaging and AI have detected forest fires as they start and helped to cut back their spread. In Wisconsin, AI-driven camera systems can reduce the speed of wind turbine blades when rare birds approach.
Machine learning algorithms can sift and learn from data so speedily that they’ll forecast which threats plants face, in addition to predict patterns of bird migration.
Environmental DNA, a force for the longer term
To these conservation methods, the World Economic Forum adds the study of environmental DNA which provides detailed insights into ecosystems. Via eDNA, sloughed-off cells or faeces may be used to discover the species and organisms that contribute to a healthy environment.
Some technology may even quantify the professionals and cons of taking specific actions to attempt to preserve biodiversity. Programmes resembling Conservation Area Prioritisation Through Artificial Intelligence will discover areas where biodiversity can most effectively be preserved and calculate the advantages of doing so, offering more compelling arguments than a random number of conservation targets. Budgets, human pressures and climate change are all inputs into the system’s simulations.
Medicine and healthcare
At a basic level, the opportunities to deploy AI in healthcare are just like those in other industries. AI can tackle administrative and repetitive tasks and so boost productivity. To give one example: a clinician can use an AI app to hearken to a consultation and add a summary to a patient’s notes.
One impediment to AI taking over sophisticated tasks is the varied nature of healthcare data in addition to poor data hygiene.
Caroline Chung, vice-president and chief data officer on the MD Anderson Cancer Center in Houston, Texas, says that the healthcare sector has struggled for years to enhance transparency and establish consistent data.
She points out that the sector’s provision of services is predicated on a “human to human connection” which is at all times prioritised over standardised paperwork. To make the most of AI, the healthcare system might want to explore higher ways to record and manage data.
One approach, she says, could be to require data standardisation, although this might not be easy. Curating and cleansing data retrospectively is simply a partial fix. While it is possible that data so processed may be used to coach models, the latter are likely to fail as recent non-standardised data is generated at the purpose of care.
“Metadata could help us cross-calibrate and understand how these data sets are different — the information that you simply’re applying whenever you’re implementing and using the model versus the information set used to coach or test the model.” This might help to pick out higher models or ascertain why some fail, but even this degree of knowledge just isn’t captured constantly or consistently. “Everyone recognises it’s an issue but they don’t know the precise ‘why?’ to have an approach to troubleshoot these challenges.”
There are some areas of healthcare which were digitised for a while and where AI could make a difference. Diagnostic imaging, resembling in radiology and cardiology, is a field where data has been generated in a consistent fashion and is “machine consumable”.
Quantum sensors and brain imagery
The homogeneity of imaging data allows models to be trained on historical images. The sheer volume of knowledge they eat can enable them not only to see existing issues but to predict problems. In January 2024 the US Food and Drug Administration approved the primary AI-powered skin cancer diagnostic tool; in 2021 Paige. AI was approved to be used in prostate diagnostics, and two years later it was applied to the detection of breast cancer.
Brain imagery will profit too. Quantum sensors are used to observe brain activity and, resulting from their small size, may be worn while patients move around. This adds a brand new dimension to how an activity is observed and measured.
Where data may be consistently gathered, other applications for AI are emerging. In the treatment of diabetes, AI models read data from the “closed loop” of glucose monitoring and insulin delivery. The calibrated and continuous measurements are reliable and trustworthy enough to direct the required treatment.
Elsewhere AI can improve evaluation of DNA, helping to predict patient susceptibilities to disease or complications and catch these at an early stage, sometimes before they develop into apparent in other ways. Predictive analytics, which examines historical data to discover trends and patterns, can discover high-risk aspects and mark a patient for intervention.
Pattern matching will find recent uses for old drugs
Companies resembling Deep6AI, Saama and Medidata all use AI to assist design studies and choose clinical trial populations which can be optimised for those patients almost certainly to answer treatment. Such selection can lower costs by finding eligible candidates faster, in addition to pulling in candidates from wider pools.
EClinical Solutions, the information and analytics platform and biometrics services provider, says generative AI could facilitate recent drug creation by predicting molecular behaviours or simulating patient populations to check treatments in a virtual environment.
Chung also says that AI pattern matching in patient data will help find recent applications for old drugs. She says the passion across the technology is pushing providers to search out ways to homogenise data to permit more extensive uses. Up so far the efforts have been more organic but there have been “some early movements around generating more structured reports”, with the American College of Surgeons requiring more structured operating room reports, for example.
There are hurdles however the health sector is rapidly expanding its use of AI. According to Snowflake, the cloud data warehouse, the expansion in data tagging and the usage of Python coding indicates a leap in the applying of AI — or not less than a jump in exploratory programming.
A clutch of universities, resembling Cornell in New York, now offer courses in data management and programming for healthcare. They teach students to make use of technology to assist them predict whether a patient could develop sepsis, for example.
Raise the ground and lift the ceiling
Future patient care may very well be enhanced by AI, with practitioners having rapid access to the particular and detailed data for a person. Patients could even have around the clock access to advice based on their very own data.
Chung says “many individuals are going after the sexy blue sky solutions” with AI. She says more deal with “raising the ground” of medical healthcare is as essential as elevating the ceiling, especially in terms of expanding and improving access to healthcare and reducing disparities. “If individuals are unable to travel, are you able to find higher solutions for delivery?”
Drawbacks
From a conservation and climate perspective, there’s tension between deploying AI and the land and energy required by data centres. As was identified at a recent FT Climate Capital Council roundtable, AI requires plenty of energy. A ChatGPT search can take 10 times as much energy as a daily Google search. Some experts consider that increased demand from AI could increase energy consumption by 250 per cent inside five years.
While AI is a strong tool, it continues to be only pretty much as good as the information that feeds it. Consequently, plenty of effort needs to enter sourcing, cleansing and maintaining inputs.
Chung says that when looking for correlations in data, for example in cancer imaging, time can also be an element. An image of a cancerous cell in any given moment will look different soon afterwards, so timing of the information getting used is very relevant.
Ethical considerations are also paramount, particularly in healthcare. Not all patients shall be keen to have their data shared even anonymously, and having apps hearken to consultations may very well be fraught with problems.
The importance of mindset
While there are many benefits to using AI, it’s critical to discover the issue that should be solved before buying right into a technology.
Happily, AI is approaching the purpose where it could actually help here, too. The next stage, Bergvinson says, is where AI is given the context to grasp an issue for which it could actually then “backward integrate these technologies to deliver the answer”.
For many social projects, technology could have to be blended with humans’ soft skills for a partnership to be effective.
Similarly, recommendations based on technology must account for the human element. Even with the delivery of weather forecasts, any messaging must account for the constraints on each farmer and their decision-making process to make it relevant.
Finally, AI needs safeguards to be sure that “nefarious intent” just isn’t amplified. Bergvinson says: “It’s in everyone’s interest to administer this fastidiously. If we burn the bridges of trust it’ll be very hard for society to be willing to have a second go.”
Next steps
Early adopters of AI are convinced that while chat large language models are most prevalent now, the following phase will involve small language models designed for specific uses. These could have area of interest applications that give access to specialised knowledge in response to a spoken query.
Bergvinson says: “While there’s an enormous investment in LLMs costing several million dollars, the SLMs which can be focused on specific verticals inside a sector will play an increasingly essential role for smaller enterprises to deliver value to society within the medical, education and agriculture sectors.”
Personalisation is gaining ground in all these areas, including nutrition, which he describes as “the nexus between health and agriculture”.
This shift is already under way. SMLs resembling Med-PaLM, a language model trained on medical data, offers answers to patients and healthcare professionals, while Nvidia has specialised medical computing tools that users can customise.
But while there’s hope that AI language models with high-level medical expertise could eventually offer skilled diagnoses, the technology continues to be young and mustn’t be relied upon exclusively.
