Mastering Machine Learning & Security: Insights from a Seasoned Founder

Table of Contents

• Introduction: Navigating the Startup and Research Landscape
• Disruption: The Key to Successful Startups
• Startup Perspectives: US vs. Europe
• Nicolas Colon's Analysis: Insights on European Startups
• Lessons Learned from Cognitive Security
• Optimizing Research Investments in Large Corporations
• Strategies for Startups: Navigating Uncertainties and Validations
• Adapting Research Practices for Startups
• Building a Competitive Advantage through Research
• Conclusion: Embracing the Startup and Research Journey

Introduction: Navigating the Startup and Research Landscape

In today's fast-paced and ever-evolving business landscape, startups and research play a pivotal role in shaping the future. As an experienced entrepreneur and researcher with over 15 years of expertise in machine learning and security, I have had the privilege of navigating both realms. In this article, I will share my insights and lessons learned, providing valuable perspectives on the intersection between startups and research.

At the heart of my journey lies Cognitive Security, a startup co-founded with a team of like-minded individuals. With a strong background in academia and experience working on U.S. Department of Defense projects, we had a choice to make: optimize missile delivery for maximum impact or focus on network security. We chose the latter, recognizing the growing importance of cybersecurity, and spun off from the university to establish Cognitive Security. With the support of venture capital funding from Credo Ventures as their first investment, we embarked on an incredible journey that culminated in our acquisition by Cisco in 2013.

The Cognitive Security Journey

Cognitive Security's success was built on a foundation of transparency and fairness. We never promised more than we could deliver, always providing our customers with honest assessments. This approach helped us establish long-lasting relationships and earn the trust of investors like Index Ventures, Credo, and Seedcamp, paving the way for our next venture, Persistent AI. Today, Persistent AI is focused on protecting AI and machine learning from human vulnerabilities. With 28 employees and a rapidly growing team, we are dedicated to securing the financial domain against attacks on machine learning models, which are currently fragile and susceptible to exploitation.

Transparent and Fair Approach

One of the core values that has guided our journey is transparency and fairness. We believe in being upfront and honest with our customers, even if it means losing some potential business in the short term. This approach has helped us build lasting relationships and retain customers for years, as they recognize the value in our unwavering commitment to delivering what we promise. At Persistent AI, we continue to uphold this principle, fostering a culture of openness and trust. Our team consists of individuals with a strong technical background, including many co-founders with PhDs, and we prioritize publishing our research in reputable conferences and journals. This commitment to sharing knowledge and advancements in AI and security is a testament to our belief in a transparent and fair approach.

Disruption: The Key to Successful Startups

In my experience, the key to building a successful startup lies in identifying and capitalizing on disruption. There must be a significant change in the world around you, whether it's a technological shift or a market evolution, that allows you to build something new and groundbreaking in a space that didn't exist before. Disruption creates instability, and it's within this unstable environment that startups can thrive.

Successful startups understand this principle and position themselves to take advantage of emerging opportunities. They recognize that the only constant in the startup world is change, and they embrace it as a driving force behind their innovations. By anticipating and adapting to disruptions, startups can create products and services that meet the evolving needs of their customers, staying ahead of the curve and establishing a competitive advantage.

Startup Perspectives: US vs. Europe

When it comes to startups, there is a notable difference in perspective between the United States and Europe. In the US, startups typically begin with a market-driven approach. They identify a problem within a specific market and then set out to solve it, initially focusing on the product and the customer's needs rather than the underlying technology or means of solution. It's only after exhausting other options that they invest in developing the necessary technology.

In Europe, the approach is quite different. Many startups, including our own experience at Cognitive Security, originate from universities. Researchers and academics develop a specific technology or area of expertise and then gradually work towards turning that technology into a viable product. The process often involves showcasing early prototypes to potential users, gauging their interest and willingness to pay, and then iterating based on feedback to refine the product and identify a viable market.

Nicolas Colon's Analysis: Insights on European Startups

Nicolas Colon, a French investor and entrepreneur, offers valuable insights into the challenges faced by European startups. In one of his blog posts, Colon highlights the importance of distinguishing between technology and innovation. He emphasizes that having the best algorithm or technological solution does not necessarily translate into meaningful impact or success in the real world. Improving an existing solution by a few percentage points is rarely noticed or appreciated by the market.

Colon also argues that startups should not invest significant resources into long-term research and development efforts. He believes that startups often pivot or change their focus before they can complete a strategic R&D activity, rendering such investments futile. He suggests that R&D is more relevant for large tech companies that have the resources and infrastructure to turn research into monetizable products and services.

While I don't always agree with Colon's views, his perspectives challenge conventional thinking and encourage startups to reevaluate their approach to research and innovation. His analysis highlights the importance of focusing on monetization and real-world impact, rather than solely on technological advancements that may not resonate with the market.

Lessons Learned from Cognitive Security

Looking back at our experience with Cognitive Security, we can draw valuable lessons about the role of research in startups. Despite substantial investments in R&D, including funding from both U.S. and Czech government sources, we realized that approximately 60% of our research efforts before the Cisco acquisition were ultimately ineffective and never made it into production or generated revenue.

One example of this was our extensive work on building a peer-to-peer optimized system for battlefield use. While the technology was highly advanced and well-suited for that specific scenario, the rise of cloud computing rendered much of our self-configuration research irrelevant, as managing everything centrally in the cloud became a more cost-effective and practical approach.

Additionally, political events like the Snowden leaks led to a dramatic shift in internet traffic encryption, forcing us to pivot and redirect our research efforts towards analyzing and breaking encryption rather than pursuing our initial research directions.

Optimizing Research Investments in Large Corporations

My experience at Cisco after the acquisition of Cognitive Security provided valuable insights into optimizing research investments within large corporations. In these environments, research serves a specific purpose: to generate ideas and explore multiple avenues, with the understanding that only a fraction of those efforts will ultimately be deployed and monetized.

Large corporations are machines built to make money from technology and research. They have the resources and infrastructure to fund numerous research projects simultaneously, allowing them to let those ideas compete and ultimately select the most promising ones for further development and commercialization.

While it may seem wasteful to discard the majority of research efforts, this approach is economically sound for large corporations. If they can take a technology developed by a small team and deploy it at scale across millions of users, the returns can be substantial, even if only 20% of the original research efforts prove successful. The scale and reach of these companies allow them to absorb the costs of research that doesn't make it to market, as long as the remaining successful projects generate significant value.

Strategies for Startups: Navigating Uncertainties and Validations

For startups, navigating uncertainties and validating their approach is critical to success. One key lesson is to avoid taking on more than one significant uncertainty at a time. Building a new product for an untested market using unproven technology is a recipe for disaster, as it compounds the risks and makes it exceedingly difficult to secure funding and resources.

Startups should strive to validate their approach through successful exits or initial public offerings (IPOs). While IPOs are rare, getting acquired by another company can provide significant financial returns, especially if the startup has a working business model and a slightly less advanced technology. Investors tend to value a proven business more than perfect technology without a clear path to commercialization.

During my time at Cisco, I witnessed this dynamic firsthand as we acquired other startups. The valuation process heavily favored companies with a working business and paying customers, even if their technology was slightly less advanced. Large corporations understand that they can invest in research to bridge those gaps, but they cannot build the entire machinery required to turn an innovation into a profitable business from scratch.

Adapting Research Practices for Startups

In the startup world, research practices must adapt to the unique challenges and time constraints faced by these agile organizations. One crucial piece of advice is to forget many of the conventions learned in academia, where research is conducted with different time horizons and objectives in mind.

In startups, iteration and agility are key. Instead of planning and executing research projects with the goal of publishing a single result in a conference or journal, startups must embrace a more dynamic approach. Expect to do something completely different every three months, as market demands and customer feedback shape the direction of the research.

Additionally, startups should carefully consider the value of publishing their research findings. While publishing is often a priority in academia, in the startup world, it may be more beneficial to focus on building a competitive advantage through proprietary research and development. Sharing all your findings without proper protection can erode your competitive edge and fail to deliver a meaningful return on investment.

Building a Competitive Advantage through Research

For startups, one of the primary reasons to invest in research is to build a competitive advantage over others in the market. However, this advantage is not necessarily derived from disruptive or groundbreaking innovations that prevent others from entering the market altogether. Instead, startups should focus on efficiency improvements and incremental advancements that can make their products or services 10% better than their competitors'.

In a competitive marketplace, even a modest improvement in efficiency or performance can translate into a significant advantage, allowing startups to win more business and generate higher revenue. This approach is much more pragmatic and defensible than relying solely on patents to protect disruptive ideas, which can be challenging to enforce and may not always provide a sustainable competitive edge.

By investing in research that delivers tangible improvements in efficiency, scalability, or user experience, startups can differentiate themselves and build a loyal customer base that values the measurable benefits their products or services offer.

Conclusion: Embracing the Startup and Research Journey

In conclusion, navigating the intersection between startups and research requires a nuanced understanding of the unique challenges and opportunities presented by both realms. By embracing disruption, adapting research practices to the startup environment, and focusing on building competitive advantages through efficiency improvements, startups can leverage research to drive innovation and success.

The lessons learned from our experiences at Cognitive Security and Persistent AI highlight the importance of transparency, fairness, and a willingness to pivot and adapt as market conditions evolve. By avoiding excessive risk, validating business models, and strategically investing in research that delivers tangible benefits, startups can increase their chances of thriving in an ever-changing landscape.

Ultimately, the journey of building a successful startup and conducting impactful research is one that requires resilience, creativity, and a deep understanding of the intersection between technology, market dynamics, and customer needs. By embracing these principles and continuously learning from experience, entrepreneurs and researchers alike can navigate the complexities of their respective domains and contribute to the advancement of innovation and progress.

FAQ

Q: Is a PhD required for conducting research in startups?
A: No, a PhD is not strictly required. Having a curious and inquisitive mind capable of PhD-level thinking is more important than the degree itself.

Q: How can startups benefit from collaborating with universities?
A: University collaborations can be beneficial if done right, with a clear long-term research goal aligned with the startup's competitive advantage, and with the startup acting as a bridge between the customer and the university.

Q: How can startups deal with competition from larger companies?
A: Competition is not necessarily a major concern. Startups are more likely to fail due to building obsolete solutions or solving problems that don't have a market need, rather than direct competition.

Q: How should startups approach research and development (R&D) efforts?
A: Startups should iterate quickly and be willing to change directions every three months, rather than following the traditional academic approach of publishing once on a specific dataset. The focus should be on building a competitive advantage and efficiency improvements.

Q: What role do patents play for startups?
A: Patents are more useful for protecting efficiency improvements that give a 10% advantage over competitors, rather than attempting to block others from entering the market entirely. Disruptive ideas are generally difficult to defend with patents.

Q: How can startups validate their technology and business?
A: Startups with a working business and slightly less advanced technology tend to get higher valuations when acquired, compared to those with perfect technology but no established business or paying customers.

Q: What are the key factors for success in startups?
A: Successful startups typically fit into a disruption, whether technological or market-driven, and focus on minimizing uncertainties by addressing either a new product or a new market, but not both simultaneously.

Q: How does research differ between academia and startups?
A: In academia, the focus is on publishing replicable results once on a specific dataset. In startups, the emphasis is on building robust, iterative code that can handle real-world situations and active adversaries, with less emphasis on publishing.

Q: What challenges did Cognitive Security face during its journey?
A: Cognitive Security faced challenges such as shifts in technology (like the emergence of cloud computing) and political influences (like the Snowden leaks), which rendered some of their research efforts irrelevant or less effective.

Q: How can startups monetize their research efforts?
A: Startups can monetize their research by either being acquired by larger companies that can leverage the technology at scale or by going public (IPO). However, having a working business with slightly less advanced technology often leads to higher valuations during acquisitions.