Recent times have seen a remarkable growth of interest for smart software that offers advanced features built upon cutting-edge artificial intelligence (AI) and machine learning (ML) methods, such as the ability to detect and adapt to voice commands, parse natural language, classify video and images, forecast more accurately complex financial processes, and drive real-time decision making in data-driven organizations. Since AI/ML methods are widely applicable across multiple domains and can leverage Big Data to deliver the expected levels of accuracy and precision after training, many commercial opportunities arise as a result: Opportunities for software vendors to deliver new AI solutions to the market, thus creatively addressing emerging customer needs to unlocking novel revenue channels. COMPRISE helps developers to penetrate this smart software market thanks to its voice-driven applications.1
Established marketing analysts expect that in the next four years, this ongoing convergence of software with AI will grow with the global AI market revenue to an estimated $554.3 billion (including hardware, software and services).2 Within the three AI market categories, software represented 88% of AI revenues in 2020. Rather than relying on the major players in the field (e.g., IBM, Microsoft, Google, VMware, etc.), establishing a strong core of knowledge and expertise is a well needed strategic step for a true Digital Single Market (DSM) in the European economy,3 and for the European ecosystem at large.
In the context of the COMPRISE project, we focused on AI-powered voice-based smart software, which may be defined in terms of heterogeneous service-based architectures that integrate both classic (e.g., e-commerce, health, consumers, etc.) and AI services (speech-to-text, text-to-speech, machine translation, etc.), which augment traditional software capabilities and enrich their runtime features and platforms through AI/ML methods.
Harnessing this next generation smart AI software with the programmability of Software Defined Infrastructure (SDI), in either public or private environments, provides a path to meet the ever-increasing demands of the market. This demand to develop AI-powered systems creates the need to urgently evolve software technology, methodologies and tools, in order to ensure that companies are immediately able to reap the benefits of implementing high-quality smart software while reducing the learning curve and overcoming harsh developer training requirements that are associated with this new development. In particular, there is a critical need for smart software production to be resilient, secure and safe when operating in a competitive vertical market. These properties are fundamental and align to the benefits that COMPRISE brings to you. Furthermore, it is crucial to foster end user acceptance with a high level of transparency on how the AI/ML methods are leveraged, i.e., to meet customer expectations concerning the interaction quality between the end-user and the voice-based system.
From the point of view of a developer, the trustworthiness, safety, and resilience of a system may only be partially assessed during the design and initial delivery since the scale and diversity of expected end-user input to the AI service may be difficult or even impossible to faithfully replicate in a lab. The continuous update due to the online learning of some AI/ML methods makes it questionable to rely solely on classic tests (e.g., unit tests) as proxies for software quality. It has therefore become increasingly problematic to deal with smart software components in a traditional software development and quality assurance process cycle. To address this, it has become imperative that the perimeter of AI software development itself is broadened to encompass more holistically the runtime production environment, making it the main object of study and the place where it is more natural for the AI-powered product to continuously evolved. This vision is entirely in line with the DevOps paradigm, according to which applications should be continuously delivered and updated in production. The idea is that the DevOps framework is the key enabler for the effective development of high-quality AI-powered smart systems, anticipating and addressing the impending problems that classic software engineering methodologies are going to face whilst embarking in the domain of AI software.
For voice-based systems, this refers to coupling COMPRISE software tools to your current system architectures to enable voice-based interactions and reinforce your delivery methods through secure DevOps methodologies, runtime verification, runtime testing, and requirement learning directly in the operational environment, so to continuously evolve therein the code, architecture, knowledge and requirements based on what is observed from the actual end-user interaction data. This is possible with the flexibility offered by the COMPRISE tools to securely store biometric voice data, annotate and label voice data, update language components as the data increases, add new languages as your market expands and constantly provide the opportunity to improve the AI services that handle the end-user interaction with your service-based backend architecture.
Dr. Youssef RIDENE