20VC The Value Chain of Machine Learning, Is There Really An Incumbency Advantage in ML & Will The Rise In Cyber Remain For the Long Term with Jake Flomenberg, Partner @ Accel

In this episode of 20 Minutes VC, host Harry Stebbings interviews Jake Flomenberg, a partner at Accel, a leading venture firm with investments in Facebook, Dropbox, Slack, and Deliveroo. Jake, who comes with a rich background in big data, having worked at Splunk and Cloudera, discusses the importance of workflow and data in the AI stack, emphasizing that a great workflow must precede data capture and model building. He also touches on the challenges startups face in collecting unique data sets and the overvaluation of large data sets, unless they are defensible and difficult to replicate. Additionally, Jake highlights the evolving landscape of cybersecurity, stressing the need for tools that enhance efficiency against sophisticated threats and the significance of data in improving security orchestration. He concludes by noting the importance of recruiting in VC and shares his recent investment in Radar Lab, motivated by the team's unique insights and the company's promising traction.

• Harry Stebbings hosts the podcast "20 minutes VC" and announces a special feature week with Excel Partners.
• Jake Flomenberg, a partner at Excel, is introduced as a specialist in big data with a background in product management at Splunk and Cloudera.
• The episode is made possible thanks to Brian O'Malley's introduction.
• Harry also mentions sponsors namely and Eero, emphasizing their impact on HR and Wi-Fi spaces respectively.

"And we are back for part two of this very special feature week with Excel partners." "Jake is a partner at Excel. As mentioned before, one of the world's leading venture firms with the likes of Facebook, Dropbox, Slack and Deliveroo all in their portfolio." "Again, all credit to Brian O'Malley for the intro to Jake's day, without which this episode would not have been possible."

• Harry Stebbings is the host and is emphasizing the importance of the special week with Excel Partners.
• Jake Flomenberg is introduced as a notable figure in venture capital, particularly in big data.
• The introduction by Brian O'Malley is acknowledged as a crucial facilitator for the episode.

• Jake studied electrical engineering, computer science, and economics at Duke University.
• He participated in a rotational program at Lockheed Martin, gaining insights into large company operations.
• Jake pursued an MBA, became interested in cloud and data trends, and joined Cloudera in its early stages.
• After Cloudera, Jake worked at Splunk and focused on product management and big data strategy.
• He was invited by his now partner Ping to join Excel and helped moderate their big data conference.

"Undergrad, I went to Duke, and I did electrical engineering, computer science, and economics." "I made a list of 20 companies cold called the ceos and wound up at Cloudera in the very early days." "I got an email from my now partner ping, and he know, hey, Jake, let's catch up. I want to tell you about this big data conference we're organizing. Maybe you can help."

• Jake's academic background laid the foundation for his career in technology and data.
• His proactive approach of cold calling CEOs led to his involvement with Cloudera, showcasing his initiative and networking skills.
• The invitation from his future partner Ping highlights the significance of networking and industry events in career progression.

• Jake notes the inefficiencies and challenges of scaling within large organizations like Lockheed Martin.
• He emphasizes the importance of agility and the ability to disrupt oneself, referencing Clay Christensen's theory of disruption.
• His observations from Lockheed Martin inform his approach to advising startups on organizational growth and adaptability.

"The incumbent nature of your business can hold you back." "If we need to go back to the drawing board, if we need to give up this line of revenue stream for something that's going to be more exciting in the future, I think that gets harder and harder for you to do as the company grows."

• Jake's experience at Lockheed Martin highlighted the constraints that come with large organizational structures.
• He stresses the need for companies to be willing to pivot and sacrifice current revenue streams for future opportunities.
• The insights gained from large companies are applied to help startups maintain the capability to innovate and disrupt.

• Jake discusses the concept of making AI practical and not just hype.
• He introduces the three-legged stool analogy for the AI stack, which includes algorithms, data, and domain expertise.
• The discussion focuses on the importance of each factor and how they interact to create a successful AI strategy.

"A large element of your role today is how to make AI, not BS." "If you would think of a triangle, if you will, maybe with algorithms or AI or ML, any sort of fancy analytic word that you would want, up at the top and maybe in the bottom left you..."

• Jake's role involves demystifying AI and ensuring it has practical applications.

• He uses the triangle analogy to describe the interdependence of key components in AI: algorithms, data, and domain expertise.

• The quote sets the stage for a deeper discussion on how these elements contribute to the effectiveness of AI solutions.## Algorithm Differentiation and Business Models

• Companies often aim for algorithm development but durable differentiation is challenging.

• The competitive landscape includes tech giants like Google, Baidu, Facebook, and Apple.

• These companies not only invest heavily in algorithms but also open source their work.

• It is difficult to envision a successful business model based solely on algorithmic innovation.

"It's very hard to imagine an algorithm that's durably differentiable in some sense."

This quote emphasizes the skepticism around the long-term viability of business models that rely solely on proprietary algorithms, given the rapid advancements and sharing of such technologies by larger companies.

• Workflow is considered the most critical factor in building a business.
• Companies need to focus on solving actual problems through workflow.
• Data should be used to improve the end user experience over time.
• Google is an example where simple workflow and data led to a reinforcing cycle.

"For me, the most important factor is workflow."

Workflow is highlighted as the primary element for a company's focus, indicating its significance over other factors like algorithms or data in isolation.

• Slack began by addressing a communication problem.
• Over time, Slack accumulated a valuable data set.
• Slack developed a strong AI and ML team to enhance customer experience.
• This demonstrates that significant businesses can be built without initial focus on data and algorithms.

"A great example of this is Slack."

Slack is used as a case study to illustrate how a company can evolve into a data-driven business by initially focusing on workflow to solve a key problem.

• Understanding past and present is crucial before predicting the future.
• The approach involves data capture, model building, and iteration.
• Demisto is an example of a company that built workflow tools for security incident responders.
• They captured unique data on incident responder actions, which is valuable despite not being large in volume.

"Let's sort of understand the past and the present before we start thinking about the future."

This quote underlines the incremental approach to data strategy, emphasizing the need to thoroughly understand existing conditions before attempting to leverage data for predictive purposes.

• Startups face the cold start problem with data collection.
• Bootstrapping can be achieved through partnerships or unique data acquisition.
• Workflow can provide immediate value, and any additional data collected is a bonus.

"If you have some way to bootstrap that problem by partnering or somehow acquiring a unique or defensible, maybe with a legal agreement set of data that no one else can get their hands on, that's a great strategy for bootstrapping."

This quote discusses strategies for startups to overcome initial data collection hurdles, suggesting partnerships or exclusive access to data as potential solutions.

• The value of large data sets is debated and context-dependent.
• Data defensibility and the value of incremental data are key considerations.
• In certain industries, such as medical imaging, exclusive data access can be highly valuable.
• The significance of large data sets varies, and feedback loops and supervised learning can enhance data utility.

"I think there's certain situations in which the value of large data is not overplayed."

This quote acknowledges that the value attributed to large data sets can be justified in specific contexts, particularly when data is both defensible and incrementally valuable.## Supervised Learning and Derivative Data

• Supervised learning involves using feedback loops to improve algorithmic performance.
• Derivative data is created from the original data through interactions and feedback.
• Derivative data can refine the algorithm's recommendations based on user corrections.
• Google's PageRank algorithm incorporates user click feedback to improve search results.

"The derivative data that you can perform supervised learning on is really this feedback loop."

This quote explains that derivative data is the result of a feedback loop that is integral to supervised learning, which helps improve the performance of algorithms over time.

• Reinforcement learning is a type of AI that relates to automated data creation.
• Simulated environments, such as those used for self-driving cars, can train algorithms effectively.
• The challenge is ensuring the simulated data fully reflects real-world scenarios and edge cases.
• There is a risk involved when the data does not account for all possible situations, especially when human lives are at stake.

"But the question is, how fully reflective is the underlying data set of the real world?"

This quote highlights the concern regarding the fidelity of simulated data to real-world conditions, which is crucial for the reliability and safety of AI systems trained in such environments.

• Startups face challenges when competing with large companies that have exclusive data sets.
• The intent of large companies to use and monetize their data can threaten startup viability.
• Startups may find opportunities in areas not central to the mission of large incumbents.
• The risk for startups increases if large companies have both the data and the intent to enter the same market.

"It's something I think about a lot."

The speaker is expressing their concern over the competitive advantage that large companies hold due to their exclusive access to extensive data sets and their potential to use it against startups.

• The distinction between a feature and a product is not always clear-cut.
• Features may represent immature or nonexistent markets, while products can address significant market opportunities.
• The perception of whether something is a feature or a product is subjective and varies.
• A product has the potential to evolve into a platform, whereas a feature might be an add-on.

"I sort of describe features as immature to nonexistent markets."

This quote indicates that the speaker views features as components that cater to markets that are either undeveloped or do not yet exist, distinguishing them from products that serve established markets.

• The increasing public awareness of cybersecurity is likely to sustain interest in cyber investing.
• Nation-state cyber actors pose new and persistent threats to enterprises.
• Such adversaries are well-funded and relentless, making it hard for companies to defend themselves.

"And so it's tremendously difficult for the average enterprise to arm and secure themselves against such a well capitalized and persistent adversary."

The speaker is emphasizing the challenges that businesses face in protecting themselves against sophisticated and determined nation-state cyber threats, which is likely to drive continued investment in cybersecurity solutions.

• Some areas of security technology, such as machine learning anomaly detection, may be overhyped.
• Effective security measures are often only practical for the most sophisticated security teams.
• There is a mismatch between the capabilities of advanced security analytics and the needs of most organizations.

"One area that I think is unfortunately slightly overplayed is this machine learning anomaly detection to rule them all off by itself."

This quote suggests skepticism towards the idea that machine learning anomaly detection alone is sufficient for organizational security, indicating that this approach may be exaggerated in terms of its effectiveness and applicability.## Explainability in Security

• The importance of explainability in security solutions is highlighted.
• Security anomalies need to be understandable to less experienced team members.
• CSOs may ignore issues they can't address due to a lack of actionable insights.

"If you don't have this degree of explainability, if you can't tell them, hey, here's the problem, and in fact, here's the recipe or playbook, how to resolve the problem that you can give to the 21 year old that just joined my security team two weeks ago."

The quote emphasizes the necessity for security problems to be explained in a way that even the newest and least experienced team members can understand and act upon.

• Security orchestration automates redundant parts of security tasks.
• Validation tools ensure security settings are correct and guide investment decisions.
• The goal is to improve efficiency in dealing with sophisticated security threats.

"With security orchestration, with companies like Demisto, we ask this question of how can we help automate not the most complicated 10% of your job, but the 20% of the most redundant portion of your job?"

This quote discusses the use of tools like those from Demisto to automate repetitive aspects of security work, freeing up staff to focus on more complex tasks.

• Excitement about various emerging areas in the security sector.
• Cloud security, data security and encryption, and container security are key focus areas.
• There's a lack of common language in cloud security, but it remains an interesting field.

"I think cloud security, although we don't really share a common language or parlance around what that actually means, unfortunately, in the industry today is very, very interesting."

The quote points out the current ambiguity and interest in cloud security, despite the industry lacking a standardized definition or terminology.

• Discusses personal reading preferences and the impact of fiction.
• "Ready Player One" by Ernest Cline is highlighted for its dystopian themes and relevance.

"My favorite book of the past year is a book called Ready Player One by Ernest Klein, and it's part of my year of reading, mostly fiction only."

This quote reveals the speaker's favorite book and suggests a personal challenge to read mostly fiction, highlighting the value found in the themes of "Ready Player One."

• Critiques the advice of raising as much capital as the market allows.
• Warns against the dangers of overspending based on large funding rounds.

"Raise what the market will bear."

The quote summarizes the commonly heard advice in the industry, which the speaker believes can lead to future financial problems for startups.

• The speaker shares surprise about the extent of recruiting responsibilities in venture capital.

"I always thought that recruiting would be an important part of my job. I didn't realize exactly how much of my time would be taken up by recruiting."

This quote reflects on the unexpected amount of time the speaker spends on recruiting, indicating its critical role in venture capital.

• Endorsement of "The Morning Paper" blog for translating complex computer science papers into understandable language.

"So there's this little blog called the Morning Paper by someone named Adrian Coiler... he takes a computer science paper and does what I'd call a la tech to english translation."

The quote describes the value found in "The Morning Paper" blog, which simplifies complex technical information for a broader audience.

• Interest in understanding the decision-making process of Amazon's CEO, Jeff Bezos.
• Bezos's ability to manage diverse and successful business ventures is admired.

"I think I'd have to be Jeff Bezos right now... It would be really interesting to get in his head and understand how he thinks about his incremental investments of time."

The quote expresses a desire to learn from Jeff Bezos's approach to managing and investing time across various successful initiatives.

• The rationale behind investing in Radar Lab, a location platform for mobile apps.
• The decision was based on market timing, traction, and the team's expertise and insights.

"My most recently announced investment is a little seed I did in a company called Radar Lab."

This quote introduces Radar Lab as the speaker's most recent investment and sets the stage for explaining the reasons behind the decision.

• Promotion of the 20 VC app and request for user feedback.
• Mention of the HR platform Namely and the Wi-Fi system Eero.
• Shoutout to the upcoming episode with Jyoti Bansal, former founder and CEO at App Dynamics.

"If you like the more technical episode, I'd love to hear your thoughts and you can provide feedback on our new 20 vc app."

The quote promotes the 20 VC app and encourages listeners to provide feedback, demonstrating engagement with the audience.