What does the nearest future, shaped by artificial intelligence and machine learning, look like? On trends in AI and ML

It is worth considering the use of these modern solutions as a part of a company's strategy. When implemented effectively and wisely, they bring many benefits. The field of AI/ML itself is, of course, very broad, developing on many levels, and today I would like to present two topics that in my opinion are worth paying attention to in the near future.

Deep learning as support for data analytics

Companies are drowning in data. The volume of information archived on a daily basis is growing all the time, and traditional analysis measures are no longer sufficient to process it and form a solid foundation for good decision-making. This is where deep learning algorithms can help. They are a form of imitation of neural networks, which is how the human brain operates. For an experienced employee, a glance at a single graph is sometimes enough to spot an anomaly. Neural networks work in a similar way, but until recently their requirements for computing power determined their low cost-effectiveness in many applications.

Thanks to cloud infrastructure, we now have sufficient IT facilities to quickly create efficient neural models (algorithms) that can successfully process huge amounts of numerical, textual or binary data (images, video or audio files).

AI-supported healthcare

An example of an industry that is successfully implementing this type of solution is healthcare. Hospitals, clinics or laboratories have a large amount of highly unstructured data, such as medical images, patient records, data streams from diagnostic equipment - and it has to be analysed quickly so the doctors can make the right decisions for the benefit of the patient. By 2021, 90% of US hospitals had already identified a strategy to implement AI solutions to support daily operations. In 2019, 47% of these facilities had no strategy related to AI solutions (2).

In addition to the obvious benefits, the reason for the increasing popularity of this technology is also the democratisation of AI solutions, which manifests, for example, in the availability of off-the-shelf algorithms that allow such solutions to be implemented even faster. It is worth mentioning that Microsoft offers a set of cloud-based tools and services, as well as an extensive knowledge base dedicated to solutions implemented in healthcare (3).

TinyML: machine learning in a nutshell

There is a popular perception that using AI/ML solutions is expensive and requires strong cloud computing facilities. Furthermore, in recent years, the huge growth in data volumes has made the performance requirements of ML algorithms on the IT infrastructure too great to be used on local (so-called edge) devices.

However, a very large part of the data we capture comes from small IoT (internet of things) devices, which, although already present almost everywhere, do not have much computing power themselves.

But if some of this data could be analysed right on the device and only the result sent instead of the source data, this would reduce the need to store much of this data and, therefore, significantly reduce the need for computing power in the cloud.

This is the idea behind the tinyML concept: to integrate ML with the IoT to analyse data as close to its point of origin as possible. To achive this, special ML algorithms are being developed that have low computing power and RAM requirements, but generate results with sufficiently high accuracy. These types of sensors, which are essentially microcomputers, can be used to raise an alert when anomalies are found in readings from a production machine or (when equipped with a camera) to signal a defect in a product just released.

The key issue here is, of course, the cost - as the price of an IoT device that can process data can be as little as a dozen or sometimes even a few dollars. Comparing this with the cost of processing huge masses of data in the cloud, it will come as no surprise that the market for edge devices (so-called edge computing) is estimated to reach, depending on the forecast, $40-60 billion even before 2030 (4).

Generative AI - ChatGPT the internet sensation

It is difficult not to mention the latest AI service of widespread interest - ChatGPT - launched in November 2022. It explores an area of artificial intelligence called NLP, or natural language processing, present for years in e.g. virtual assistants. Here, however, we have the possibility to ask complex questions that go beyond one specific subject domain, as is the case with 'standard' chatbots.

The popularity of ChatGPT is due, among other things, to the possibility to test it for free on the tool's website (5). However, this service provided by OpenAI will be payable. The challenge of the coming months will therefore be to commercialise it and find the right business use cases, but it already appears to be a great alternative to the leading search engine, which has long since turned from a content provider into an advertising platform.

Artificial intelligence driving growth

We are living in times of technological breakthroughs, among which AI is leading the way. Organisations around the world in the fields of healthcare, industry, commerce, etc. are already introducing groundbreaking innovations in artificial intelligence and machine learning into their daily processes. AI is not only shaping the future of almost every industry, but is also acting as a driver for technologies such as big data and IoT. Given the rate of growth of AI and machine learning, their impact cannot be underestimated, but it is important to look at how these technologies can support our business.

‍

Sources:

1. PwC’s Global Artificial IntelligenceStudy: Sizing the prize, https://www.pwc.com/gx/en/issues/data-and-analytics/publications/artificial-intelligence-study.html‍
2. 90% of Hospitals Have Artificial Intelligence Strategies in Place, https://healthitanalytics.com/news/90-of-hospitals-have-artificial-intelligence-strategies-in-place‍
3. Health - Microsoft Industry Blogs, https://www.microsoft.com/en-us/industry/blog/health/‍
4. Machine learning at the edge: TinyMLis getting big, https://www.zdnet.com/article/machine-learning-at-the-edge-tinyml-is-getting-big/‍
5. ChatGPT, https://chatgptonline.net/