AI expert Ieva Martinkenaite, an executive with Telenor Research, reflects on the AI promises and moonshots that once filled the newspaper pages, finally answering the question: What happened?
Over the past several years, we have experienced incredible hype around all things “AI.” This has been fueled by staggering VC investments into research and breakthrough innovations in machine learning (ML) from the internet giants.
If we were to believe what we were once promised, we should be working alongside intelligent robots, sitting in self-driving cars, and engaging with human-like chat-bots by now. However, due to real-world complexities, making those technologies work and be accepted as safe is a much more difficult task than we first anticipated.
What are the key AI challenges we still face?
First, making fully autonomous ML agents work in real-world settings requires feeding them with large amounts of high-quality data, which is still a problem in many industrial installations; data is embedded in complex systems; it’s noisy and highly distributed or impacted by various external factors.
Second, the business community is struggling to productify and scale AI, primarily because it is hard to do it right, and in the end, it requires significant investments into technology, skills, and most importantly, changes in leadership mindset and redesigning how organization works.
Third, ethical challenges associated with data privacy and security, biased training data, human manipulation through fake content, lack of explainability of complex ML models top media charts and policy debates around the world; those create additional concerns and lower excitement in the uptake of this technology among business and the public.
Does this mean we are facing yet another “AI winter”, as evidenced by unmet promises of AI, sluggish business uptake and ever increasing ethical and environmental concerns in society?
My answer to that question is resolute ‘no’. In fact, it’s the opposite. I would argue that what we see as AI hype diminishing is a move from exuberance to pragmatism. VCs are shifting their interest from moonshot projects to more mundane and “boring” AI applications. Companies continue to invest in data and AI capabilities to transform their own businesses, which is less visible in the public and now considered business as usual.
For many of us, “doing AI” is about prioritising problems to solve, getting data pipelines in place, experimenting with models, organising workflows, deploying the right software and skill base. This shows that AI tech exuberance is followed by bubbles bursting and then refocusing on down to earth, practical AI applications, which is a common pattern for many transformational technologies (such as electricity, mobile telephony, or cloud computing).