20 Years of AI “We Are All the Stakeholders”: Microsoft Philanthropies Chief AI Architect
Dr Michael Tjalve has spent over two decades researching AI and working to demystify the technology.
Key Takeaways
- The evolution of AI from its early days to its widespread adoption today reflects decades of research and experimentation.
- Breakthrough generative AI technologies now reach over 100 million people worldwide weekly, showcasing the impact of researchers.
- While highly utilized tools like speech recognition software emerged in the 1990s, broader applications of AI continue to drive innovation and societal impact.
In the late 90s, public access to artificial intelligence was largely limited. But AI, as we interact with it today is the result of decades of research and experimentation – and the people who brought us here.
Thanks to the early innovators of AI breakthrough generative AI such as OpenAI’s ChatGPT or Anthropic’s Claude now reaches more than 100 million people across 185 countries each week.
But these widely used tools, are just the tip of the iceberg.
Early AI Frustrations
Dr. Michael Tjalve, Assistant Professor at the University of Washington and Chief AI Architect at Microsoft Philanthropies began his journey in AI over two decades ago when technology was still rudimentary, often frustrating users more than assisting them.
“When I first started working on AI, the technology looked very different compared to today,” Dr. Tjalve recalls. Initially focused on improving the accuracy of automatic speech recognition, his work included research of specific applications like voice-controlled car systems. The technology, while innovative, often required a determined user to see its practical value.
His early work focused on improving the accuracy of automatic speech recognition, a field that had roots as early as the 1950s but in 1990 was still in its relative infancy. Applications were limited to specific uses, like voice-controlled car systems, and the technology often demanded a determined user to realize its potential.
“The experience for most users at the time was one of frustration with the capabilities. You had to really want it to work in order to find practical value from it. However, when it did, we’d often see users amazed at what appeared like magic.”
Despite these challenges, several highly publicized AI milestones marked AI in the 1990s, entering public consciousness and reshaping the public view of AI which up until that point had been largely limited to sci-fi movies and dystopian novels.
Deep Blue, NASA Robots, and AI Breakthroughs
AI made headlines when Deep Blue, an artificial intelligence, defeated chess grandmaster Garry Kasparov in 1997. This decade also saw the deployment of Sojourner, NASA’s first autonomous robotic system.
A less publicized, but no less groundbreaking, event was the world’s first commercial speech recognition software built by Dragon NaturallySpeaking, that became available to consumers in the same year of Gary Kasparov’s defeat by Deep Blue.
The advancements in personal computing during the 1990s played a significant role in driving forward speech recognition technology. The availability of faster processors and increased computational power enabled software developers to create more sophisticated speech recognition algorithms and applications like Dragon Dictate.
AI In Philanthropy
Beyond the highly publicized displays of AI innovations, the application of AI in the real world is of most interest to Dr Tjalve.
“Given the maturity of the AI capabilities at the time, not just for speech but in other branches of AI as well, I started to feel that the most interesting challenges were less around how to improve the quality of the underlying capabilities and more around how AI was used in the real world.”
Traditionally, the nonprofit sector hasn’t been an early adopter of the latest technologies, however, according to Dr Tjalve nonprofits are now exploring the use of AI across a wide range of innovative scenarios.
To give an example of the wide range of applications of AI in nonprofits, Dr Tjalve shares how he is currently working with a nonprofit on a solution to automatically assess spoken language proficiency of indigenous languages in communities where the younger generation struggles with talking with their grandparents in their local tongue.
Tjalve explains that AI often serves to improve the efficiency of an existing process but it can also enable scenarios that are impossible without AI, such as leveraging geospatial images from before and after a natural disaster to train an AI model.
Addressing Global Challenges and Demystifying AI
Of course, the promise of how AI will bring new opportunities should also tempered by the reality of how it may limit these for some.
When discussing ethics and AI, many immediately think of data protection or job displacement. Dr Tjalve concedes that this is an important discussion to have “but it’s not always an easy one.”
“Part of the challenge is that you really need to have broad representation in the room across stakeholders to stress test assumptions, to ensure diversity of voices and backgrounds, to understand concerns and build trust, and to reduce the risk of blind spots when the technology is built and when it is used.”
Given his focus on bringing AI to the nonprofit sector, the relationship between AI and philanthropy is something that Dr Tjalve is committed to demystifying through his work.
Having worked on a range of AI capabilities and gradually more directly focused on the societal impact of the technology being built – “both the good and the bad”, Tjalve adds, “I joined a team called Tech for Social Impact (TSI) in Microsoft Philanthropies to move that charter from aspirational to concrete.”
He concedes that “modern AI, and particularly generative AI, does represent a tectonic shift and it has introduced or accelerated valid concerns. So I work on demystifying AI, helping nonprofits understand what AI is and what it isn’t, how it makes decisions, how it makes mistakes, and how you can put in place mitigation strategies to proactively counter those mistakes.”
“The Stakeholders are, well.. all of us.”
Today, AI technologies are integrated into various facets of daily life, from virtual assistants to advanced data analytics. The evolution has been driven by a combination of improved algorithms, increased computational power, and vast amounts of data.
AI capabilities have continued to advance and breakthroughs have led to an explosion of use cases and broad adoption, particularly with mainstream access to generative AI a couple of years ago.
Dr. Tjalve’s work at Microsoft reflects this integration, focusing on creating AI systems that are both powerful and accessible.
“When you’re working with technology that’s disruptive at a societal scale, the stakeholders are, well.. all of us.”
Technology Moves Fast, AI Even More So
With the increasing pressure on governments to regulate emerging technologies and a shortage of experts in fields like AI, cybersecurity, and blockchain, there remains a growing need to close the skills gap in the industry.
The rapid advancements in AI and the broad adoption over the past couple of years have accelerated regulation in this space. Dr Tjalve believes it is encouraging to see the EU’s AI Act and regulation in other regions taking shape but there’s still “much more needed.”
“Regulation and bureaucracy move slowly by design to avoid us moving too fast in any one direction. Technology, on the other hand, moves fast and AI specifically even more so.”
