What Lies Ahead for AI and Machine Learning

Image courtesy of Microsoft Azure

AI and Machine Learning are already transforming the way people interact with technology, making it easier and more efficient to perform tasks, such as data analysis and predictive modelling. This technology is also being used to automate mundane tasks, enabling companies to save time and resources. With the global AI and Machine Learning market anticipated to grow exponentially, the market is expected to expand at a compound annual growth rate (CAGR) of over 20% between 2023 and 2033, according to International Data Corporation (IDC). This growth is being driven by advancements in the technology, including more sophisticated algorithms and better data processing capabilities, IDC reported.

Today, most of us encounter AI every single day. From the moment we wake up to check our smartphone to asking Siri to help find our AirPods or telling Amazon Alexa to turn on the lights and music. AI has quickly made its way into our everyday lives. In fact, AI and machine learning have been around for a long time. The earliest successful AI program was written in 1951 by Christopher Strachey, later director of the Programming Research Group at the University of Oxford. Strachey’s checkers (draughts) program ran on the Ferranti Mark I computer at the University of Manchester, England. By the summer of 1952 this program could play a complete game of checkers at a reasonable speed.

So, What’s The Setback?
With the existing state of things, some people argue that AI will displace occupations and increase unemployment. One application of AI is a robot. Therefore, they claim that there is always a chance of unemployment as a result of chatbots and robots replacing humans.  For example, robots are often utilized to replace human resources in manufacturing businesses in some more technologically advanced nations like Japan. AI offers both the promise to improve the safety and health of workers, and the possibility of placing workers at risk in both traditional and non-traditional ways.

In a world still dealing with the impact of an unprecedented pandemic, combined with an increasing amount of industrial automation, the “new normal” has generated significant demand for applications capable of ensuring worker safety. These applications include physical identification and compliance for sensitive environments like cleanrooms used in silicon production, monitoring of industrial settings with heavy machinery in use, keeping track of security zones with restricted personnel access, and more.

Enforcing compliance with policies for the appropriate use of personal protective equipment (PPE) and limiting personnel to the locations for which they are equipped results in reduced work-related accidents, thus improving employee safety while also supporting steady productivity. These practices also enable organizations to avoid expensive and reputation-damaging citations from government agencies monitoring industrial environments.

Worker safety is not to be taken lightly, as safety concerns have become an issue of widespread concern around the world. Many are turning to AI-powered machine vision-based applications to ensure compliance and maintain enforcement using a centralized architecture. By doing so, these organizations have successfully automated their systems via a combination of smart cameras, sophisticated machine vision applications, and industrial computing solutions that do not require human resources to ensure compliance. These systems are better equipped than humans to complete tasks of this kind. For example, smart monitoring solutions ensure employees are wearing the appropriate PPE when required, ensure the right people are allowed in the correct zones, and enable real-time implementation of corporate policy where it is needed.

However, there are also concerns regarding the use of AI in the workplace. This is especially true if the data used are incomplete, inappropriate, or insecure; the methods are not easily explained or understood; or if the systems operate without the oversight of a human agent. Integration of this technology, using systems which are predictable and reliable, has been shown to improve performance and acceptance. The opposite also appears to be true, demonstrated by failures associated with the Maneuvering Characteristic Augmentation System (MCAS), an AI system designed to activate and assist the pilot under circumstances, which resulted in the two crashes of Boeing 737 Max airplanes.

Trends in AI and Machine Learning
Although research gaps exist regarding the use and impact of AI and machine learning, the use of this technology is already very much part of modern living. Rather than worrying about a future AI and machine learning takeover, let us focus on some of the AI-specific trends that will help us in the next couple of years:

Adoption of cloud computing for machine learning applications: The growing adoption of cloud computing for machine learning applications is also set to be one of the major trends in AI in 2023. With cloud computing, businesses can benefit from improved scalability and flexibility when it comes to their machine learning projects.  By leveraging cloud-based services such as Amazon Web Services (AWS) or Microsoft Azure, companies can quickly scale up or down depending on their needs while also taking advantage of advanced technologies like artificial intelligence (AI) and deep learning.  As a result, more businesses are turning to cloud computing for their machine learning needs in order to take advantage of its many benefits.

Improved language modelling: ChatGPT demonstrated a new way to think about engaging with AI in an interactive experience that is good enough for a wide range of use cases in many fields, including marketing, automated customer support and user experiences. In 2023, expect to see a growing demand for quality control aspects of these improved AI language models. There has already been a backlash against inaccurate results in coding. Over the next year, companies will face pushback on inaccurate product descriptions and dangerous advice, for example. This will drive interest in finding better ways to explain how and when these tools generate errors.
 

Chiplets: AI acceleration has been one of the main drivers for heterogeneous compute in the last few years, and this trend will certainly continue as Moore’s Law slows. Heterogeneous compute refers to the system design technique whereby accelerators for specific workloads are added to more general computer hardware like CPUs, whether as separate chips or as blocks on an SoC. This trend is obvious in the data center, but endpoint SoCs for everything from household appliances to mobile phones now have specific blocks dedicated to AI acceleration. For large-scale chips, such as those used in the data center, chiplets are a significant enabling technology. Chiplets allow huge chips to be built by connecting multiple similar reticle-sized die through a silicon interposer, but they also enable heterogeneous compute by enabling the connection of CPU, memory, and accelerator die at high bandwidth. Chiplet technologies are maturing and will continue to do so in the next two years as we see chips like Intel’s Ponte Vecchio hit the market.

Human and machine collaboration: In the coming years, we can expect to see an increasing trend towards human and machine collaboration in the field of AI. As AI technologies become more advanced and sophisticated, they will be able to work with humans in order to achieve greater results. This could take many forms, from machines providing assistance with tasks or decisions that are too complex for humans alone, to humans teaching machines new skills and knowledge. By leveraging the combined strengths of both human intelligence and AI technology, businesses will be able to unlock a wealth of possibilities for improved efficiency and productivity.

Transformer networks have been widely used for natural-language processing for some time, with impressive results. Today’s large language models (LLMs) can be used to power chatbots, answer questions, and write essays, and the text they produce is often indistinguishable from what a human might have written. These networks use a technique called attention to explore relationships between words in a sentence or paragraph. The trouble is that to really understand language, LLMs need to consider relationships between words that are further apart in the text. The result is that transformer models are rapidly growing, and so are the computers needed to train them. Training of GPT-3 was estimated to cost millions of dollars — that is to train one model, one time. Despite huge costs, the demand for accelerated compute from transformer networks is not slowing down. Economic or practical limits to transformer size — if they exist — have not yet been sighted, let alone reached.

Generative AI has already been used in creative applications such as music composition and video game design. In 2023, we can expect this trend to continue as generative models become even more sophisticated and powerful. This could lead to new possibilities for creative expression and innovation across a variety of industries. Generative AI models use neural networks to generate new content from existing data. This type of AI can be used for a wide range of applications – from creating images or video effects to generating natural language text and music.  Generative models offer an efficient way to generate large amounts of artificial data, allowing organizations to save time and resources in the creative process. For example, generative AI can help developers create virtual environments for video games by automatically generating textures and patterns based on pre-existing data.

Generative AI is also helping businesses better understand their customers’ preferences and tailor products accordingly. Companies such as Amazon are increasingly using generative AI technologies to recommend products that customers may like based on past behaviour – resulting in improved customer experiences and better sales.

AI industry regulation: As the use of AI becomes more widespread, governments worldwide are beginning to take steps toward regulating its use to protect consumer privacy and ensure ethical practices are followed. We can expect this trend toward regulation to continue in 2023 as governments look for ways to ensure that companies using AI technology do so responsibly. Heightened AI industry regulation is already being implemented across the world. In Europe, for example, the General Data Protection Regulation (or GDPR) sets strict standards for companies that collect and use consumer data. 

AI-driven technologies such as facial recognition are also increasingly falling under governmental scrutiny – with some countries passing laws to ban their use or restrict its application in certain areas.

In the coming year, we can expect even more regulations to come into effect worldwide – making sure organizations using AI technology do so ethically and responsibly. Such regulations could be critical to protecting public safety and privacy while continuing to allow businesses access to powerful AI toolsets reaping rewards without creating a risk of harm.

AI to make ethical decisions: In 2023, Explainable AI (XAI) is set to be a major trend in the world of Artificial Intelligence. XAI can help machines make ethical decisions by providing humans with an explanation of how a decision was made and why it was taken. This will enable humans to better understand the machine’s decision, thus enabling them to take corrective action if needed. XAI also allows for greater accountability, as humans can review and reject decisions that are not in line with company policies or public safety regulations. XAI makes use of techniques such as natural language processing and machine learning to analyze data and provide meaningful explanations for AI-generated decisions. By using explainable AI, companies can ensure that their AI systems are making ethical and responsible decisions without compromising on accuracy or speed. As such, XAI technology will be essential for businesses looking to deploy AI systems responsibly in 2023 and beyond. Furthermore, explainable AI can be used to promote transparency when it comes to data privacy and usage practices. By providing users with detailed explanations about how their data is being used – including what kinds of algorithms are being employed – companies can ensure that users have full control over their own personal data while still benefitting from the advantages of AI technologies. As a result, explainable AI looks set to become an important tool in helping businesses achieve greater levels of trustworthiness when it comes to protecting user data in 2023 and beyond.

Brain-inspired computing: Neuromorphic computing refers to chips that use one of several brain-inspired techniques to produce ultra-low–power devices for specific types of AI workloads. While “neuromorphic” may be applied to any chip that mixes memory and compute at a fine level of granularity and utilizes many-to-many connectivity, it is more frequently applied to chips that are designed to process and accelerate spiking neural networks (SNNs). SNNs, which are distinct from mainstream AI (deep learning), copy the brain’s method of processing data and communicating between neurons.

These networks are extremely sparse and can enable extremely low-power chips. Our current understanding of neuroscience suggests that voltage spikes travel from one neuron to the next, whereby the neuron performs some form of integration of the data (roughly analogous to applying neural network weights) before firing a spike to the next neuron in the circuit.

Approaches to replicate this may encode data in spike amplitudes and use digital electronics (BrainChip) or encode data in timing of spikes and use asynchronous digital (Intel Loihi) or analog electronics (Innatera). As these technologies (and our understanding of neuroscience) continue to mature, we will see more brain-inspired chip companies, as well as further integration between neuromorphic computing and neuromorphic sensing, where there are certainly synergies to be exploited. SynSense, for example, is already working with Inivation and Prophesee on combining its neuromorphic chip with event-based image sensors.

Voice and language-driven intelligence: is set to be a major trend in AI in 2023, as the technology gets better at understanding natural language. This will enable machines to interact with humans more naturally, allowing for more efficient conversations and greater accuracy when it comes to interpreting user commands.  By utilizing voice recognition and natural language processing technologies, businesses can provide their customers with an enhanced level of customer service that is both faster and more accurate than ever before. Furthermore, this type of AI technology can be used for a wide range of applications such as automated customer support systems, virtual assistants, and even conversational agents that can help users’ complete tasks or make decisions quickly. As such, voice and language-driven intelligence looks set to revolutionize the way we interact with computers in 2023.

Conclusion
2023 is set to be a year of major advancements in the world of AI and machine learning. From AI-driven text, speech and vision to explainable AI, these trends will shape how businesses interact with their customers. It is prepared to function at scale. Its full application cannot be predicted at this point, but it ensures real-time actionable insights and offers newfound responsiveness in an uncertain world. We can expect increased investment in quantum computing research and development as companies look for ways to take advantage of its powerful capabilities. Additionally, edge computing is ready to become more important for low power devices, such as wearables or IoT sensors due to faster response times and lower energy consumption. Meanwhile, cloud computing is on the rise for machine learning applications because it provides improved scalability and flexibility when compared to traditional methods.

Although innovation has unintended side effects and too much of a good thing can still be damaging, it is important to criticize with basis. Yes, history is scattered with bad technology calculations, but early adopters of AI should involve judgements with an open mind. The opportunities for AI and machine learning are infinite. While moral, ethical, and legal questions are important for society to answer, the technology needs principled direction and smart engineers to support the implementation. Integrating AI and machine learning into core functions will help companies become more resilient and recession-proofed.

When leveraged correctly, this technology can help save time, streamline processes, efficiently access organizational knowledge, and provide masterful insights. But it is about more than just disregarding mistakes and inefficiencies—it is customizing our preferences and work out complex problems as they happen. In ways we cannot yet imagine, AI will help make us more human and help humans interact in more natural ways. It just needs our help getting started. AI will help individuals, companies, and government to navigate through the emerging risks and uncertainties of the future if applied smartly.

With all these developments taking place over the next few years, one thing is certain: AI and machine learning has an exciting future ahead!