Dr Ruoyi Zhou of IBM Research discusses tech accessibility, the growth of IoT connectivity and the challenge of AI bias.
IBM established a research lab in Dublin in 2011 as a technology centre focused on creating smarter cities – the first of its kind in Europe.
A decade later, it has become a hub for scientists and engineers working on breakthrough technology in the areas of AI, quantum, healthcare and security.
Dr Ruoyi Zhou is the director of the Dublin lab and is responsible for driving innovation at the research centre.
Before moving to Dublin in May 2019, Zhou held many roles within IBM. She previously served as the director of IBM Accessibility Research, where she oversaw the development of advanced technology to enable accessibility, the creation of AI-powered assistive technology for people with disabilities, and the exploration of IoT-based AI solutions for an ageing population.
She was also the co-director of AI for Healthy Living, a joint research centre between IBM and the University of California.
‘We don’t talk enough about how technology can help society be more inclusive’
– DR RUOYI ZHOU
“Working in accessibility was definitely one of the highlights in my 25-year career at IBM,” Zhou told SiliconRepublic.com.
“We are seeing new technologies that can provide people with assistance in daily life and work, like apps for translating speech to text, or describing what’s in an image. With navigation systems getting better and better all the time, fully autonomous vehicles will be a game changer.
“We don’t talk enough about how technology can help society be more inclusive, instead of people having to adapt to fit in with the status quo. This is especially important when we think about how many neurodivergent people there are. Technology could help us all live and work better.”
Zhou holds a PhD in materials science from Rutgers University. “In middle school, I started to learn physics and was fascinated by how things in the physical world can be explained by laws of physics. I dreamed of becoming an engineer and putting physics into practical use,” she said.
“My parents were well educated though a little old-fashioned. When it was time for me to apply for college, they wanted me to study biology or chemistry, subjects perceived to be a better fit for girls at the time, instead of engineering or physics. I chose materials science as a compromise and never regretted my decision.”
Zhou’s career in IBM spans multiple decades and began with her working as a thin-film process engineer in its hard disk drive business. Four years later, she moved to the tech giant’s storage subsystem group and quickly learned how different storage technology was compared to thin-film process.
“I was reluctant to make it and very nervous initially. However, I soon regained my confidence in a new field, which helped to make my subsequent moves much easier.”
The growth of AI and IoT
From that initial move, Zhou went on to take up several leadership roles across the business, giving her a broad view of the AI and internet of things sectors as a whole.
She said the biggest change she has seen is the massive increase in connectivity due to the advancement in technology such as sensors, AI models and 5G.
“While the concept of IoT has been around for about 16 years, it didn’t start to kick off until around 2014. The number of businesses that use the IoT technologies has increased from around 13pc in 2014 to about 25pc today. On a consumer front, the market for the IoT continues to grow,” she said.
“In 2021, IoT Analytics expects the global number of connected IoT devices to grow 9pc to 12.3bn active endpoints. By 2025, there will likely be more than 27bn IoT connections. There are already more than 200 known applications in diverse sectors as industry 4.0, smart cities, smart homes, connected care and e-health.”
AI has also seen strong growth. According to Statista, the global AI software market is forecast to grow rapidly in the coming years, reaching around $126bn by 2025. Zhou said there are two major reasons for this.
“First, thanks to the internet and the exponential growth of IoT devices, there is now more data being generated than ever before. Second, we have more computing power to boost the power of AI. With high performance computing and cloud technologies, we have the ability to properly train AI models. This will continue to increase as new technologies such as neuromorphic computing or AI hardware and quantum computing emerge,” she said.
“The convergence of bits, qubits and neurons – in other words, classic computers, quantum computers and AI – will drive new breakthroughs and accelerate new discoveries in science and technology that enable us to combat climate change, curb the pandemic and tackle many other challenges that we are facing today.”
Addressing AI bias
While AI has advanced so much, there are ongoing challenges around bias and privacy. In September 2021, the UN’s human rights chief called for a moratorium on the sale and use of AI technology until safeguards are put in place to prevent potential human rights violations.
“There is no doubt that AI is making data increasingly valuable. Today, businesses and institutions are facing conflicting imperatives,” said Zhou.
“On one side, we want more accessible data to advance AI technology. On the other side, it is extremely important to ensure the security and privacy of data, especially personal data, in order to maintain consumer trust.”
Another major ethical barrier for AI adoption is that of bias, which can be present in the data that is used to train and test AI models. In 2020, it was discovered that the much-cited ‘80 Million Tiny Images’ dataset may have contaminated AI systems with racist, misogynistic and other slurs – and that’s just one example.
“An example people often cite is AI résumé screening, which is widely used in high tech industry. Such systems, if trained using profiles of current employees in high tech industry, will most likely be discriminatory against women who are still very much underrepresented in tech industry,” said Zhou.
While mitigating bias in AI is a long journey, Zhou said there are many techniques and best practices that companies can implement, including building diverse, multidisciplinary teams, checking for incomplete or biased data when training AI models and gathering input from all stakeholders, especially minority-serving organisations.
In 2018, IBM launched a tool that can scan for bias in AI algorithms and recommend adjustments in real time. This has since been expanded to become the AI 360 Toolkit, which offers three open-source tools to help mitigate AI bias.
“To effectively scale the implementation of AI and its adoption, the technology must be created and used in accordance with our ethical values,” said Zhou.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.