Why 2023 Is the New 1993
Hear from a Veteran Entrepreneur and AI Expert: the AI Revolution, Founder Lessons, and What the Best VC Investments Have in Common
Recently,
, an AI veteran, entrepreneur, and educator well respected in Silicon Valley, gave a talk at the Wharton School of the University of Pennsylvania. Howie shared his thoughts on the ongoing AI revolution, as well as lessons learned as a founder and investor. This story recaps the highlights.
Why 2023 Is the New 1993
Howie: We will look back to 2023 as the year where a new pradigm shift started. We are effectively reliving 1993 in case anyone is “complaining” they missed the beginning of the Internet era.
What happened in 1993? It marked the beginning of the Internet Age with the introduction of Mosaic, the first widely available web browsers that popularized the internet. To be clear, the internet wasn’t born in 1993; its development can be traced back to the 1970s as a tool primarily used by governments and researchers. However, the internet wasn’t meaningful at all for everyday people and businesses until user-friendly web browsers like Mosaic came along. Since then, internet content and applications have exploded, ushering in a new era.
Similarly, AI technologies have been advancing continuously for decades, but the introduction of user-friendly generative AI models like ChatGPT has transformed the field. The widespread availability of AI power will trigger exponential growth in AI applications. If any moment in history will be remembered as the beginning of the AI Age, it is now. It was once said that "there are decades where nothing happens; and there are weeks where decades happen". The release of ChatGPT in late 2022 has ushered in a new era of rapid and profound changes. There have been many decades of progress in 2023 already, and we are only half way through the year.
That said, we need to set realistic expectations for what AI is capable of in the near future. In most use cases, AI will act as a "copilot" - a smart assistant to humans who will have to remain in the loop. The technology is not nearly advanced enough to completely replace humans in complicated tasks.
Lessons Learnt as An Entrepreneur: Fear No More, Disrespect No More, and Never Take No for an Answer
Howie: My experience as a founder has taught me to “fear no more”: no incumbent is too big to take on. For example, by 2000, Microsoft held a dominant position in the x86-based server OS business that was regarded as almost impossible to challenge. However, in the 2000s, VMware upended Microsoft in this area with its edge-cutting virtualization technology, which laid the early foundation for cloud computing. By the time I left VMware in 2011, the company was generating ~$4 billion in annual revenue and had a market cap of ~$40 billion. And most importantly, by then more Windows server workloads were running on VMware’s virtualized platform than on physical servers directly, so in a way, VMware had become the most dominant Server OS and surpassed Microsoft.
Meanwhile, working with startups constantly reminds one to “disrespect no more” — do not discount anyone’s potential because of their youth, tenure, or lack of experience. For example, the CEO of Figma, Dylan Field, was a product design summer intern at LinkedIn just two years before he co-founded Figma. Over the following decade, he and his team grew Figma into a $20 billion company by the time Adobe announced its acquisition of Figma in 2022.
While people often talk about essential founder qualities, I have seen examples of success that contradict most of the so-called "must-haves." However, all successful founders have at least one thing in common: they must be tremendously persistent and resilient, given the nature of the founding journey.
As a founder, you can experience several emotional roller-coaster rides in a week, or sometimes even a day. At one moment, you may be everyone's hero for closing an important customer, but a few hours later, you could hear from one of your critical employees that he or she decided to part ways. These frequent ups and downs are not for the faint of heart for sure.
As a founder, you will have to face countless rejections, but it's important to never take "No" for an answer. In the world of startups, "Yes" often follows multiple "No"s. When we were fundraising at TrustPath, the company I co-founded and led as CEO, one of our investors decided to back out. It was disappointing, and we could have let them walk away. Instead, I requested a meeting with them that night (on a Sunday) to discuss why they decided to pass. After a deep conversation, we managed to change their mind and convince them to back us after all.
However, persistence does not mean repeating the same request over and over until you get a different response. It requires understanding and often involves compromise: you must figure out the other party's concerns and explore ways to address them. You may need to make concessions on certain matters in exchange for what is most important to you.
Lessons from Greylock: The Best Investments Are Always Controversial
Howie: Working at Greylock with some of the best minds in venture capital has helped shape how I think about venture investing. My biggest takeaway, which is also an institutional belief at Greylock, is that the best investments are always controversial. As Greylock's Managing Partner, David Sze, once famously put it in public: "The best investments are non-obvious enough that they result in a mixed vote by our partnership."
(Below is a summary of selected Q&As between Howie and the audience)
Q&A
Q: What non-consensus views do you hold on AI?
Howie: Many people argue that AI applications can't build a moat since they don't possess proprietary technology as advanced as the foundational models they rely on. However, I don't agree with this view. In the early days of the internet, people said the same things about Yahoo and eBay - that they had nothing unique, were too easy to copy, etc. After all, weren't they just aggregating snippets of other people's web pages? How could that provide lasting value? And yet, we have all seen how this turned out.
Another ongoing debate concerns whether generative AI startups might find more opportunities in B2B enterprise or B2C consumer applications? Instead of generally recommending one over the other, I believe the founder should choose a market for which they have the greatest passion, conviction, and understanding. When Elon Musk set his sights on electric vehicles while still a college student, the field was fraught with technological and commercial challenges, and few people believed in EV startups. However, with his exceptional talent and conviction, he made Tesla what it is today.
At the end of the day, regardless of my personal views, I strongly advise everyone to think from first principles and develop their own perspectives on any topic. It is dangerous to blindly accept what “experts” say, particularly for a field as dynamic as AI.
Q: 1993 was followed by the dot-com bubble a few years later. As we compare the landscape today to 1993, should we be concerned about a potential AI bubble on the horizon?
Howie: Companies that build products relevant to their time need not worry about bubbles. In fact, bubbles help distinguish those with lasting power from their weaker competitors. For instance, if you look at what happened to Amazon during the dot-com bubble, initially its stock price sank big time along with everyone else’s, but eventually, it benefited greatly from the elimination of numerous e-commerce players that lacked solid business models or operations.
Q: What advice do you have for founders when choosing what AI product to build?
Howie: Here is my suggested framework for ideation at least in the enterprise space: first, choose a sub-domain that you know well and where you can build something valuable (pain-killers, not vitamins) quickly. You do not need a clear thesis on Day One for how this product can build a moat - it is often unclear initially anyway, and a moat is generally created by excellent execution, rather than careful strategizing. Second, find a problem that was a “mission impossible” for incumbents to solve in practice, but might be addressed with the unique strength of generative AI. To identify such problems, you have to dive deep into a domain.
Q: Would the high cost per query for large language models limit their use cases? Would this cause value to predominantly accrue to hardware and AI infrastructure companies like Nvidia, as opposed to AI applications, in the near-term?
Howie: The cost per query should not be evaluated in isolation, but rather relative to the value created by the query, and the cost of an alternative solution. A cost for a query to suggest takeout restaurant choices may not be justified. Yet the same cost can be negligible if the language model solves a consumer’s puzzle that otherwise would require him or her to find a more expensive solution. The key is to find use cases with the right economics for the language model.
About Howie
Howie Xu is a 20+ year Silicon Valley veteran entrepreneur and educator. He is currently the Senior Vice President of Engineering AI/ML at Palo Alto Networks, a global leader in cybersecurity.
Previously, as the founder of VMware Network, Howie started and managed VMWare's networking organization for a decade. He played a key role in growing the flagship product vSphere from zero to approximately $4 billion in annual revenue.
In 2015-16, Howie served as an Entrepreneur in Residence (EIR) at Greylock Partners, a prestigious VC firm. In 2017, Howie co-founded TrustPath, a machine learning based security startup. He served as its CEO until it was acquired by Zscaler. After the acquisition, Howie remained with the company for four and a half years while Zscaler’s revenue grew from $150 million to $2 billion.
Howie is also a guest lecturer at Stanford University.
Like this post? Why not share it on social media or forward to a friend (and if you’re new to Entrepreneurship of Life, subscribe to get future updates)? I also interview distinguished US immigrant tech founders, builders, and investors. See you again soon :)