By Nicholas Larsen, International Banker
Largely driven by advances in machine learning (ML) and deep learning (DL) training capacities in recent years, artificial intelligence (AI) has come in leaps and bounds in terms of what it can accomplish and how it can be applied, further developed and how the world can benefit from it. By ingesting and interpreting data, AI algorithms can generate more accurate insights and predictions that continue to support smarter decision-making. With adaptive AI systems proving adept at continuously reacting, learning, and modifying their results from ingesting new data, the capabilities of this technology should improve dramatically.
While traditional ML consumes raw data to be applied in the real world, the process is largely static, where the learning process is complete and remains unchanged as the system is used. As long as world conditions remain the same, the model should be able to produce successful results. But of course, our world is constantly changing, which means that the less the model reflects this changing environment, the less accurate it becomes.
Adaptive AI dynamically integrates new data from your operating environment to gain more accurate insights in real time. It is increasingly regarded as the next evolutionary stage of artificial intelligence. By incorporating a more responsive learning methodology, such as Using techniques such as agent-based modeling (ABM) and reinforcement learning (RL), adaptive AI systems are more responsive to the changing world around them and can therefore more seamlessly adapt to new environments and circumstances in the early stages of AI development -System not available. This can be achieved by working on new data in runtime and development environments that allow models to adapt and update their own codes, allowing the AI system to dynamically retrain, learn, and improve upon these changes to the environment .
Adaptive AI systems support “a decision-making framework that focuses on making faster decisions while remaining flexible to adapt to problems as they arise,” noted Gartner, while acknowledging that such systems aim to continually evolve based on new ones Learning data in run-time environments in order to adapt more quickly to changes in real-world circumstances. This data can represent changes in consumer or business behavior in real-time, meaning adaptive AI can continuously maintain its accuracy. “Flexibility and adaptability are now critical, as many organizations have learned during the recent health and climate crises,” noted Erick Brethenoux, Gartner Research’s Distinguished VP Analyst, in October 2022. “Adaptive AI systems aim to continuously retrain models or apply other mechanisms to adapt and learn within runtime and development environments – making them more adaptable and resilient to change.”
This adaptability is sure to prove crucial in the coming years, when the Internet of Things (IoT) and autonomous vehicles are expected to rise sharply in popularity. Such applications must continually consume vast amounts of data to reflect ongoing changes in the external environment in real time. Simply put, current static machine learning models are unable to quench this thirst for new and continuous data and therefore cannot be used effectively for such use cases. However, adaptive AI models are capable of ingesting and responding to an endless stream of data.
Adaptive AI’s applications could even prove life-saving, especially given the potential it has to improve healthcare industry performance. The ability to consistently analyze data related to thousands, if not millions, of patient symptoms and vital signs can enable adaptive AI systems to optimize the clinical recommendations they generate.
Adaptive AI can even adapt to distinguish between a mix of patients in different regions of the country. “A Minneapolis hospital may see a very different mix of patients than one in Baton Rouge, 1,200 miles down the Mississippi, in terms of age, comorbidities like obesity or diabetes, and other factors,” Sam Surette, director of regulatory affairs and quality assurance at the AI-focused medical company Caption Health and a former reviewer for the US Food and Drug Administration (FDA), wrote in an October 2020 article for the health publication Stat. “Because clinically appropriate performance depends in part on factors such as disease prevalence, access to local data can help tailor performance to the precise needs of each facility. Adaptive AI could even learn subtle differences between institutions, such as how frequently they perform certain blood tests, that are otherwise difficult to factor into calculations.”
So, over the long term, adaptive AI will deliver faster and more accurate results, which should mean companies can uncover more meaningful insights to further optimize decision-making. This also implies that adaptive AI systems can exercise more autonomy in providing such outcomes — having the ability to independently adjust their own learning capacities in response to changes in the real world that were not previously present when each system was designed improve. Therefore, companies that rely on AI don’t have to put as much human capital into the process as it might have once required.
However, adaptive learning models rely on solid training in machine learning skills. “It works best when trained on millions or even billions of customer interactions across geographies, verticals and use cases,” Mike Gozzo, chief product officer at Ada, a provider of AI-powered automated brand platforms, told Digital Customer Experience publication CMSWire in October 2022. “This creates a rich dataset that drives personalized and proactive experiences for every customer at every interaction.”
It is also important to recognize that simply having access to new data does not necessarily mean that an adaptive AI system will improve its performance, especially if it draws the “wrong” lessons from the new data. Algorithms are designed to make predictions based on the information they consume. However, if this information is consistently biased in terms of obtaining it, the system may be biased in operation by the same.
In 2018, for example, Amazon was forced to scrap its recruitment engine, which used AI to determine which resumes made the best applicants, because the models displayed clear biases against women. This was largely because machine learning models were trained on patterns in resumes submitted to the tech giant over the past 10 years. With men submitting the overwhelming majority of resumes – reflecting male dominance in the tech industry – the AI system “learned” that male candidates were preferable to females.
And using healthcare as an example, a 2019 research paper found clear evidence of racial bias in a commonly used algorithm, such that black patients assigned the same risk level as white patients by the algorithm were actually sicker than white patients. “The authors estimated that this racial bias reduced the number of Black patients identified for additional treatment by more than half,” the study said. “Bias occurs because the algorithm uses healthcare costs as a proxy for healthcare needs. Less money is spent on black patients with the same needs, and the algorithm therefore incorrectly concludes that black patients are healthier than equally ill white patients.”
However, as the world increasingly relies on real-time data processing and analytics powered by IoT, and businesses become more data-driven, adaptive AI systems will be built more frequently as the need to spontaneously adapt to environmental changes becomes more pressing. Gartner predicts that by 2026, companies that have adopted AI engineering practices to build and manage adaptive AI systems will outperform their peers in terms of the time and number of processes required to operationalize AI models be exceeded by at least 25 percent.
In the meantime, companies may need to recalibrate their business models to streamline their decision-making processes and take full advantage of adaptive AI. “First lay the foundations of adaptive AI systems by complementing current AI implementations with continuous intelligence design patterns and event stream capabilities — and eventually move to agent-based methods to give more autonomy to system components,” Gartner’s Brethenoux said. “To make it easier for business users to adopt AI and to help manage adaptive AI systems, incorporate explicit and measurable business indicators through operationalized systems and build trust into the decision-making framework.”
Businesses will almost certainly also have ethical considerations about the appropriate and compliant use of AI when engaging in this reengineering, while regulators will inevitably have a say on how adaptive AI should be deployed in a controlled manner.