Icebergs & Industry (Is AI the Real Deal?)

All Things Venture #091

I have been thinking a lot lately about the newspaper industry and how it may be emblematic of where we are in our current cycle of technology innovation. In capitalism there are winners and losers, and over the past twenty years the newspaper industry has undeniably been one of the biggest losers in recent history. A few examples:

• Advertising revenue is down from a peak of $49.3bn in 2005 down to $9.6bn in 2020, accounting for an 81% decline in revenue

• Employment in the industry has shrunk by more than 75% from 2000 to 2022 (source)

• On average, the industry has lost $2.5 billion dollars in advertising revenue a year, for the past 15 years 

• When responding to “how much trust and confidence do you have in mass media, such as newspapers, TV and radio” 38% of Americans responded as having “none at all” - the highest percentage ever recorded

• Even Warren Buffet threw in the towel on newspaper ownership

So what happened?

Well I can show you better than I can tell you. But I’ll tell you a little bit about it anyway. Basically Google happened. Sergey and Larry and the technological wave of the internet, beginning in the 1990s, absolutely drop kicked the newspaper industry into oblivion. The below graph tells you all you need to know. Google EVISCERATED newspapers. Why? Because they created a superior product (i.e search based ads) that was better, faster, and cheaper than the leading analog (i.e print based ads). 

And just to reinforce a point. The above graph is just Google’s revenue. One company. The below graph is the revenue of Google, Facebook, Twitter, Yelp, Pinterest, and Snapchat. All of which primarily make money from selling ads. 

By 2020, six companies that for the most part were less than a decade old, generated nearly 30x the amount of comparable revenue as an entire industry that has been around for centuries. What’s even more astonishing is that, from 2017 to 2020, these six companies generated more revenue than the entire newspaper industry had, in the past twenty years combined. Newspapers aren’t defunct, but they no longer capture public attention and therefore no longer capture the same economic value they did, a quarter century ago. They were fundamentally disrupted by the internet and looking backward, we can learn a few things:

• Significant TAM creation occurs when a product is superior (i.e digital vs print ads)

• Verticalized platforms can accrue value, but the power law persists (i.e Google outperforms Yelp, Facebook outperforms Twitter)

• Outsiders become insiders, none of the founders of the above companies were repeat Silicon Valley style founders

• Technology leads to clear productivity gains. Advertising revenue per employee at the above companies is 16x that of the newspaper industry in 2020

• Multi-decade megatrends compound to create enduring value; CAGR for the combined 6 companies has been  > 60% since 2000

So knowing the above, and knowing that the technology world is in the midst of another innovation cycle that some believe to be as or more important than the internet,  the question to ask ourselves is, what is AI’s version of the newspaper industry? What will the advent and continuous improvement of natural language processing, neural networks, and large language models, disrupt? Said more bluntly, who’s going to get drop kicked? Who, unbeknownst to them, is currently having significant parts of their job automated away by some engineer who furnishes their apartment like this?

Right now, my instincts are telling me that the professions that are most at risk, are primarily white collar jobs that follow a framework that I think of as ODC: “Open-ended, dynamic, consultative.” These are jobs where the mandate of what to do isn’t predefined, the answers to the questions that are asked are rarely known ahead of time, and they largely require some level of critical thinking and/or expertise. Whether we like it or not - this is happening. Companies like Tome, Harvey AI, and Tabnine are each focused on leveraging LLMs to automate work primarily done by sales reps, lawyers, and developers respectively. Work that in each instance is open-ended, dynamic, and consultative.

I think about it this way. AI is a lot of things, but if we boil it down to its core components AI is a type of technology that acts as a predictive machine. And I know this sounds weird, but a lot of modern work is exactly that. If you are in any form of professional services today, a large part of your job is:

1. Receiving an open ended assignment

2. Collecting data related to the open ended assignment

3. Running analysis to determine the best path forward 

4. Executing against your determined best path forward

5. Gaining feedback on your performance of the overall assignment

6. Starting again from step 1

There is absolutely room for office work to be done better, faster, and cheaper by software. That’s no secret, but if the industry continues to progress, it’s worth asking yourself, what happens if everything the technologists are trying to achieve goes right? 

What happens if we are truly in the earliest innings of the application era for AI? 

To share my own thoughts, here’s what I generally think happens if we are in the earliest innings of the application era for AI:

• The VC industry kicks into overdrive to begin making speculative bets on the application layer across a range of industries (this is already occurring, see here)

• Applications that achieve early traction have a demonstrable 10x improvement over the current analog, and are concentrated in “low stakes” processes (i.e Synthesia for enterprise grade internal L&D content)

• Entrepreneurs begin to realize the key to unlocking “better” results from training data is by placing domain experts into the feedback loop of improving models; as a result, winning companies deliberately hire internal “centers of excellence” that are staffed by these operators

• In the immediate term (6 - 12 months), demand for GPUs continues to grow bolstering NVIDIA and associated cloud services providers both generalized (i.e Azure, GCP, AWS) and AI specific (i.e CoreWeave, Lambda)

• Success of early application layer breakouts inspires additional entrepreneurs with domain expertise in niche verticals. Early employees at early winners leave to attack service areas with similar complexity. A repeat of vertical SaaS maturation occurs (i.e Salesforce to Veeva). The Vertical SaaS category becomes the Vertical AI category. 

• Professional services become a real time, on demand global product. Increasing supply of professional services dramatically decreases costs, unlocks latent demand for professional advisory style work (i.e personal tutors, dieticians, financial advisors, etc)

• Lower cost of professional services work leads to greater access. In the long term (i.e 20+ years), it becomes commonplace for all working adults to have a personalized financial advisor, doctor, lawyer, dietician, tutor etc. 

I think what is important to acknowledge as well is what is potentially at stake. If, and again this is a big if, the arc of technological innovation continues to improve - then profound change is going to occur within society. If high quality legal services are available on a global basis, will the 60% of immigrants facing deportation no longer navigate the justice system without counsel? If the cost of a financial advisor decreased 10x, would the 64% of Americans who can’t afford a $400 emergency be in a better position to manage their finances? If every American student had access to on-demand, personalized tutors would we reverse the trend of lowest student achievement that we’ve seen in two decades?

Finally, there’s an adage in silicon valley and VC circles that is attributed to the science fiction author William Gibson that, “the future is already here. It’s just not evenly distributed.” But in addition to the uneven distribution of the future, I believe that there’s also an uneven observation of the future. This is painstakingly obvious from a BBC interview with the legendary musician David Bowie in 1999. When prompted to share his perspective on the burgeoning technology that was the internet this is what he had to say:

“I don’t think we’ve even seen the tip of the iceberg. I think the potential of what the internet is going to do to society, both good and bad, is unimaginable. I think we’re actually on the cusp of something exhilarating and terrifying… It’s an alien life form… The actual context and state of content is going to be so different to anything we can envisage at the moment, where the interplay between the user and the provider will be so in sympatico it’s going to crush our ideas of what mediums are all about”

If you believe that AI has had it’s Mosaic browser moment with ChatGPT, if you believe that the recent past can serve as a guide for the present, and if you believe in the enduring power of technological innovation, than to me, the underlying conclusion is that just like for the newspaper industry in 2005, there is an iceberg in the water today, and lots of people are asleep at the wheel. Chart your paths accordingly, place your bets, and let’s hold on, the world is about to look a lot different over the next 20 years.

---

That’s it everyone, hopefully you all enjoyed this post! If you did, I would definitely appreciate if you shared it with others or left your feedback in a comment below. Cheers - Dez