AI and the Next Wave of Digital Transformation
I started my career in Silicon Valley back when large enterprises were frantically trying to navigate the rise of social media as their customers started sharing intimate details of their daily lives on public websites and employees were tweeting out company secrets. Fast-forward to today, we are witnessing a new wave of digital transformation led by a mix of (newish) AI vendors, cloud computing, and social media companies.
Recently, I was invited by Siemens to attend their “AI with Purpose” summit in Munich in my triple roles as blogger/media, diversity in AI advocate, and responsible AI leader. The Summit programming reflected the wide-spread belief in the potential of AI but also signaled there is no appetite for billionaire thought experiments and a growing demand for solutions grounded in today’s reality rather than an imaginary future.
TL;DR
Here is a summary of my insights from the 4 conference tracks — data, scaling, regulation, and the future of industrial AI — and conversations with business leaders and attendees.
✅AI is fueling the next wave of digital transformation in enterprises.
✅Companies are trying to strike the right balance between AI innovation, results, and risks.
✅Return on AI investments is still elusive.
✅Data accuracy and security is critical for building customer trust and staving off legal issues.
✅Data scarcity and growing compute costs are spurring more interest in SLMs (Small Language Models).
✅AI regulation in Europe is real and businesses need to get ready to comply.
✅Translating AI safety from theory to action in an industrial setting is challenging.
✅Industrial AI requires interoperability with and integration into existing industrial and enterprise solutions.
✅Market leaders like Siemens can play a key role in their customers digital transformation journeys and also influence development of trustworthy AI systems.
✅Industrial AI is still dominated by men, while it was great to see many amazing women leaders and new AI talent, there’s room for more!
Balancing innovation, results, and risks.
The delicate balancing act between innovation and results while addressing risks was a persistent theme at all AI summits and conferences I attended across Europe over the past few weeks.
Tech vendors like to claim that AI is magical or different from all previous technology waves, and this has led to unrealistic expectations according to Parmy Olson at Bloomberg, who opines that AI is in the “trough of disillusionment.” A survey of over 2,500 businesses by Lucid Works reveals significant slowdown in spending, “with only 63% of companies planning to increase AI investments in the next 12 months compared to 93% in 2023.” You can only awe business leaders with robotic dogs and glitzy proof of concepts (POCs) for so long before they demand ROAI (Return on AI Investment) as Joe Kendrick pointed out in his recent story.
In the industrial AI realm, technology companies like Siemens don’t view AI as a standalone product but as an enhancement to their existing industrial and enterprise solutions. Their approach is more pragmatic and they view digital transformation as a convergence of Information Technology (IT) and Operational Technology (OT), which necessitates interoperability and integration between AI and other digital technologies.
Given the massive environmental impact and rising compute costs, there is a gradual industry-wide shift away from Large Language Models (LLMs) to more efficient Small Language Models (SLMs). Data efficiency is key to addressing data scarcity. Adding semantic context to improve accuracy to reduce hallucinations and improve accuracy is essential for building trust with customers and also to stave off liability issues. However, AI benchmarks weren’t available for public sharing at this summit. Data security and reliability is also high on Siemens priority list for all the above reasons. Siemens’ approach to data privacy wasn’t clear, which was a bit odd as Europe typically leads on this crucial issue while the US struggles with it.
At the summit, Siemens announced their industrial AI Copilot partnership with Microsoft with acknowledgement that Microsoft is still working through some (privacy) issues, which I suspect may affect the pace and scale of adoption. On a related note, perhaps due to the growing number of legal controversies, Microsoft seems to be moving away from consumer applications and doubling down on enterprise and industrial use cases.
As a market leader with over 175 years of history and 320,000 employees, Siemens is well-placed to influence the digital transformation journeys of their industrial customer base and also steer AI vendors towards development of trustworthy solutions. So it seems logical that they would launch the Siemens Xcelerator Marketplace to connect sellers and developers that provide hardware and software for digital transformation.
Move thoughtfully and don’t break things.
Whatever else you may think of Sam Altman, he and OpenAI have been hugely successful in fueling the Generative AI hype cycle. It’s one thing to get consumers using free apps to make bad art but to convince enterprise customers to use any software or technology that’s the target of a growing number of lawsuits and struggles with hallucinations is quite an impressive feat.
Many enterprises are building their AI planes as they are flying it (metaphorically speaking). Siemens, despite its extensive technology credentials with over 50 AI patents, still needs to show thought leadership in an overhyped Generative AI space, while also trying to coax positive results from these largely untested technologies. Despite a mountain of unanswered questions and unaddressed risks, businesses are still marching forward with adoption of AI, spurred on by a combination of FOMO (Fear of Missing Out), and likely relentless pressure from stakeholders.
However, there is one key development in Europe that bodes well for future of responsible and trustworthy AI. Prof. Sonja Zillner, the Lead of Trustworthy AI at Siemens outlined the current state of AI regulations stating that, “AI regulation (in Europe) is now real. Industry must be prepared to provide compliance with these requirements for AI products that fall within the above categories.”
The timeline for AI Act will come into force in July 2024, different parts of the regulation will be applied at different times. Some use cases like social scoring will no longer be allowed in Europe and will be subject to heavy fines.
US is behind on regulating AI but the AI act has kicked off the process for its agencies to develop their roadmaps. Similarly, UK currently only provides guidance and additional mitigation measures are needed for trustworthy AI. Europe has the first and only legal binding approach with heavy penalties involved.
When it comes to development of trustworthy AI, in their comprehensive presentation on AI Safety, Mark Zeller and Andreas Hapfelmeier from Siemens reiterated a key issue that many AI practitioners are familiar with - translating talk to action in an industrial setting is not as easy as it sounds.
The key issues are summarized as follows: First, you are operating in a dynamic open environment where there is a great deal of uncertainty. Second, AI/ML is a black box. Third, academic studies conflate benchmark performance and safety, but these are not the same. Lastly, there are emerging guidelines but no established standards for AI safety. The key is to manage and address safety concerns throughout the entire AI lifecycle and not just as an afterthought. Also, worthwhile to keep in mind while designing for safety is that AI is “only” a small part of the entire industrial system and requires mitigations at different steps of the AI lifecycle.
(Re) Surge of AI research labs.
During a brief tour of the newly launched Siemens Technology Center (STC) at Garching Research Campus in the north of Munich, I had a chance to view demos for some of their new AI Proof of Concepts (POCs).
Similar to many technology companies in Silicon Valley that started in a garage, Siemens too started in a workshop in a back courtyard in Berlin. Over the years, the garages and workshops have been replaced with Research and Development (R&D) labs, which are now a mainstay at leading technology companies around the globe. Some like Google have recently scaled back their innovation labs, due to lack of commercial success for many of their ambitious ideas, as part of broader cost cutting efforts, and are doubling down on their AI projects.
According to their press release, the Center is co-located with the Technical University of Munich (TUM), Max Planck Institute and SAP, with around 28,000 people working there, making it one of the largest centers for science, research and teaching in all of Europe. Universities play a key strategic role in AI in many European countries due to their cutting-edge research and ability to attract global talent, while companies like Siemens give students an opportunity to train on real-world applications.
One of the main use cases for Generative AI shared by the Siemens team was automating repetitive industrial tasks such as generating custom code for manufacturing systems. It seems to be a promising way for manufacturers to upgrade their automation systems without requiring custom hardware and providing more flexibility with minimal downtime. Other demos included complementary technologies like augmented reality and digital twins, which enhance existing industrial operations through simulation and remote management of inaccessible or volatile systems.
Commercialization and scaling these POCs while mitigating any related risks will be the key to successful deployment and adoption of AI.
I wrote about my experience with lack of diversity at tech conferences across Europe and believe that it is a shared responsibility between organizers, sponsors, and speakers. Event organizers must be intentional about inviting women, the sponsors should be thoughtful about the constraints (childcare, financial, other) that women face, and women must show up when the opportunity presents itself or pass it on to another deserving woman.
Overall, it was an excellent opportunity to get an insider view into the leading German technology powerhouse, have meaningful discussions with women leaders, meet the next generation of diverse AI talent, and get a good read on what’s happening in the industrial AI space.
At future summits, I would like to see more discussions and developments on critical responsible AI topics like sustainability, privacy, and diversity.
PS: I also participated in a panel discussion on Women in AI leadership at the Summit and will post the key takeaways from that separately.
Stay tuned!
If you need help scaling your responsible AI programs, contact me via https://lighthouse3.com/ and to support more diversity in AI, sign up at https://womeninaiethics.org/
