ConnectAI brought together AI experts, practitioners, disruptive companies, global organizations, and leading research institutions inartificial intelligence. This three day event, organized by the Center of Business Excellence of the School of Economics and Business at the University of Ljubljana (CBE SEB LU) and the UNESCO International Artificial Intelligence Research Centre (IRCAI), included 50+ speakers, 20+ masterclasses, 10+ workshops and 20+ business cases in AI.
Intact Digital CEO, Natasa Milic-Frayling, contributed two masterclasses, focused on AI readiness and AI quality assurance.
INTACT Software Library platform and services provide stable computing environments and a principled approach to documenting AI systems. It enables reliable benchmarking, validation and reproducibility of AI processing and reconstruction of past activities for the purpose of regulatory compliance, IP protection, project continuity, and knowledge transfer.
“AI or Not AI: Is there a choice?” masterclass held on 25 Jan 2022, 16:30-17:15 BST
Abstract. Innovation comes in waves that propagate throughout the ecosystem and transform businesses, reshape practices and rattle regulatory frameworks. Initially a new technology appears to be a differentiator but in time, with a broader adoption, it turns into a requirement. Managing an innovation trend, like incorporation of AI into products, is a challenge with a daunting realization: “The trend is unstoppable. If we don’t take it up, our competitors will.”
While one may not have a choice of whether or not to play, one can still choose how to play. In this session we reflect on the nature of innovation and pre-requisites for taking it from R&D to business development. With AI, the process is complex, without established standards, and requires in-depth expertise to assess quality at specific stages. We reflect on the dilemma between early adoption to guarantee market share and risks of moving into uncharted territories that may lead to suboptimal performance, moral responsibility and possibly a liability.
“Working with a black box: Reproducibility and quality assurance in AI” masterclass held on 26 Jan 2022, 12:00-12:45 BST
Abstract. As with any technology innovation, dependability of AI based systems is key to adoption and impact. In highly regulated sectors there are legal requirements for organizations to validate systems before deployment and to reproduce system outputs decades later. Many AI systems are non-transparent or too complex to reason about and, from the operational perspective, de-facto black-boxes. Therefore, the producers and users of AI must work jointly to define quality standards for such systems and to develop best practices for their quality assessment and quality assurance.
In this session we discuss the importance of AI reproducibility in order to support comprehension, reasoning and reliable application of AI for specific purposes. We reflect on the past technology trends and effects of fast technology obsolescence to motivate actions towards reliable and long-term reproducibility of AI results. With increased use of SaaS models to incorporate AI technologies into solutions, it is particularly important to support verification and validation of AI beyond the market-lifetime of the SaaS platforms.
