2023.Aug.01

REPORTS

Report on Women in AI APAC 2023 Summit “AI for Good”

B’AI Global Forum’s Director, Professor Yuko Itatsu, participated in Women in AI APAC summit held in Sydney, Australia on 14 and 15 June 2023. Women in AI, where Professor Itatsu serves as a Japan Ambassador, was founded in Paris, France in 2016, and currently has more than 9000 members from more than 140 countries of the world. Its APAC division has representatives in Malaysia, Singapore, Japan, India, and Australia. This year’s summit was organised around two issues that guide Women in AI’s activities – development of responsible AI and improving of gender representation in the AI field.

Many of the 40-odd speakers of the event paid attention to recent developments in the field of Large Language Models (LLMs), or AI models trained on large textual corpora and used for language generation. As Professor Toby Walsh from the University of New South Wales pointed out, with improved quality of the output of these models and the rise to fame of ChatGPT and similar applications, AI has come to be understood as a part of our everyday lives. This adds urgency to the necessary development of responsible AI by industry actors and to the regulation of AI by governments.

Summit was held around the time of EU’s adoption of its first legislation aimed at regulating AI (known as EU AI Act), so European regulations were featured prominently in the summit’s discussions. This legislation requires companies to disclose information about data used for training of ML models, as well as clear marking of images, text and other content that had been generated by AI. These regulations and their implications were the focus of the panel discussion featuring Professor Natalie Stoianoff who also leads Asian Pacific Copyright Association, Dr. Rita Matulionyte, a senior lecturer at Macquarie University and Jane Rawlings, a lawyer specializing in Australian intellectual property law.

The three speakers explored the example of the lawsuit by several artists challenging Stable Diffusion AI image generator to demonstrate the complexity of regulating AI. Most of the image data used to train AI models such as Stable Diffusion is copyrighted content, and even creating a database of such data constitutes copyright infringement in Australia. However, the algorithm used in Stable Diffusion modifies the original images by adding noise and other elements to them before storing them as training data, which makes it difficult to enforce copyright law in its current form.

Dr. Alexander Kriebitz and Ms. Auxane Boch of the Technical University of Munich also spoke about the complexity of AI regulation. They attributed it to the nature of AI as a composite of multiple technologies and the complexity of those technologies as two important factors. Dr. Kriebitz and Ms. Boch suggested that the basis for creating a common framework of AI should be human rights, which are already agreed upon globally through international law including the Universal Declaration of Human Rights and the Charter of Fundamental Rights of the European Union. However, despite high consumer demand for responsible AI, there are few examples of its successful implementation and there is still a strong perception of low return on investment (ROI) for these efforts among private companies.

How to resolve this contradiction and balance innovation and responsible AI is the challenge which was addressed by many of the speakers at the Women in AI Summit, particularly those from the field of wellness and medicine. Associate Professor Yang Song, Associate Professor Beena Ahmed, and Dr. Gelareh Mohammadi from the University of New South Wales are working on developing algorithms for image recognition and speech recognition and reliable data analysis in the medical field. The three researchers focus on using AI in their research on child development, speech-language pathology, and biomarkers – molecules that signify presence or progression of a disease. These pathways for application of AI are not considered a worthy investment by private companies, which reinforces the need for researchers to take the lead in using AI in these areas.

For Professor Ahmed, working in the field of voice recognition means constant efforts to collect data which is very easily obtained in large volumes by companies that provide services like voice assistants. This anecdote hints at a power imbalance between researchers and private companies working in the field of AI. Further examples of power imbalances leading to competing priorities emerged during a panel discussion and Q&A featuring Ms. Xiaomei Wang, CEO of PathoAI, an AI diagnostic imaging software and hardware company operating in China and several other countries, and Dr. Fatemeh Vafaee, Associate Professor at the University of New South Wales. The two spoke about the importance and ethical challenges of implementing AI-powered image analysis for diagnostic in the developing world.

In this way, despite the shared understanding of the need for responsible AI, there are still tensions between actors in industry, academia, and the private sector. As Professor Kim Weatherall, who specialises in intellectual property at the University of Sydney, pointed out, to resolve these tensions and achieve this common goal, government regulations and ethical guidelines need to be reconceptualised as drivers of innovation. The National AI Centre coordinated by the Commonwealth Scientific and Industrial Research Organisation (CSIRO) and led by the summit’s keynote speaker Stella Solar, has been using this approach in helping small and medium-sized Australian companies to use AI in transforming their business. According to Ms. Solar, implementing AI in a responsible way as soon as it begins to be introduced as a part of a company’s activities is the key to making a reality the motto of the summit – “AI for Good.”

Reflecting on her own career in IT and the experiences of women in her immediate circle, Ms. Solar also emphasized the importance of wider support for increasing the diversity of the field. This includes such seemingly unrelated to work aspects as more equitable sharing of childcare and household duties with male partners. Many of the other speakers at the Women in AI Summit reflected on their own experiences and the challenges faced by women working in the AI industry, and their strategies to overcome these challenges. Ms. Jeanne Lim, one of the creators of Sophia, a robot famous for being granted Saudi Arabian citizenship, reminisced that when she attended robotics conferences, often the only two women in the room were herself and Sofya.  Ms. Lim’s experience is a testament to the need to narrow the gender gap in AI and related fields. Women in AI’s Ambassadors from Malaysia, the Philippines, Australia, and Japan, who were involved in organizing the event, spoke of the gender gap in their respective countries, as well as their efforts to ensure the diversity of the field, including creation of networking and mentoring opportunities, including for students of primary and secondary schools. All countries shared issues including a lack of female role models and the added burden of emotional labour for the women in the industry who may be pushed into being such role models. In his closing remarks, Dr. Mahendra Samarawickrama who serves as a director of the Center for Sustainable AI, linked the need for male allies to work to eliminate gender disparities to creation of responsible AI that takes into account gender and other bias. For the many students from computer science and other fields who attended the Summit, the event was a valuable opportunity to get to know both female role models and allies.