27 Comments

A true example of disruptive search is TikTok: looks like a toy, serves very different and underserved segments, operates on a social dimension that Google would be hard-pressed to replicate.

The key point here is "search vs. answers." ChatGPT is not disruptive, because you can't make the mistake of conflating the two ontologies; most commenters aren't familiar with a lot of very basic definitions in e.g. library science. To simplify, you might describe the two in terms of intent. "Search" shines, and makes its money, in high-intent queries. Local plumber, cancer treatments, etc. --I started my career around 2005 working for the interactive Yellow Pages, which was still making truly insane amounts of money: if someone looked you up in the YP, they were going to call you, and businesses clocked this really well. (Bigger the ad space--higher chance you'd get the call--more money to YP.) Google took note. This kind of ad is what throws off that $60B in free cash flow.

Chat GPT shines with low-intent and/or low-certainty queries, which typically involve synthesizing information. At Goldman-Sachs, the bright but inexperienced analyst will produce an overview of the oil industry in Pakistan. That BBIA may miss critical information, include inaccurate information (a facilities survey from 1961), and/or present that information in an inaccurate, misleading, or incomplete way. It still saves the i-banker time, though; he'll apply his expertise to the BBIA's work, hopefully seeing the gaps and accounting for them.

As someone who taught the humanities at college level, a lot of Chat GPT's output has been an eerie blast from the past for me: a lot of it *exactly* resembles the output I used to get from my undergraduates. It draws from a wider range of sources (I believe, although it has the undergraduate habit of not citing any, heh) but is still quite wobbly with the actual interpretation. Like an undergraduate, it consistently misprices the value of information--shorthand for "accuracy"--something Google just doesn't do.

Google is quite capable of taking advantage here, as the piece points out. I don't think Chat-GPT is a threat to its core model of high-intent, high-precision searches, and would in fact allow monetization of under-monetized "messier" searches. Few realize that one reason Google did so well on the scaling of search is that they aggressively hired, for years, anyone who knew even a little bit about it. That deep bench will pay off.

Expand full comment

Great insight on AI and ChatGPT! At this stage all it can do is pull information and spit out what appears to be a “narrative” answer. What AI lacks is the ability to perform at Maslow’s higher levels of thinking: synthesis, analysis, and evaluation. It also lacks the ability to find the relationship and common threads between seemingly disparate observations and pieces of information. That is where great insights and breakthroughs occur in almost any field. In this sense, AI is only as smart of the programmer, but cannot “reason” in the ways people can.

Expand full comment

100%. Just give it some riddles or logic puzzles and see how it does. Unless it's similar enough to something already published on the internet, it kind of bombs. Really puts all the AGI/singularity fear-mongering in perspective

Expand full comment

Agreed that differentiating between high-intent, commercial queries and informational ones is key. Google knows this; informational searches often generate GPT-esque answers (though they're often ripped clumsily straight from search results), while higher-intent queries yield ads.

Google doesn't stand to lose ad revenue by replacing their current informational result format with sexier Bard results. In fact, they can afford to offer this for free, undercutting OpenAI's $20/mo revenue stream.

The high-intent side of Search will be more difficult for OpenAI to counter-position. Google would be wise to test the plugin strategy before OpenAI gets any more of a head start, but having worked with Google for years, I don't trust them to move quickly or decisively enough.

Excellent article, Packy. Feels redundant to say with the quality of all of your work, but this was a favorite.

Expand full comment

Great article about the limits of disruption! But the real solution is to come up with a catchy, Tweetable name to replace “The Innovator’s Dilemma” :-)

Some options:

- Incumbent’s Inertia

- Positioner’s Curse

- Product Paralysis

- Monopolists Myopia

Surely you or ChatGPT-4 can create an ear worm to fill that conceptual hole!

Expand full comment

Awesome piece, Packy. Love how you’ve overlayed the deep/shallow and platform/product spectrums in regards to positioning. Also love your derivation of ‘positioning’ from Hamilton’s coined ‘counter-positioning’ - I’ll be leveraging that in the future!

The only critique I have on the piece is that I think too much time/space is spent at the forefront on labeling what does or does not technically fall under the ‘innovator’s dilemma’. Don’t get me wrong - I think it’s important to identify that people often mischaracterize such business positionings, but as a relatively new subscriber who’s hungry for juicy business insights and perspectives, the piece felt bogged down at the top with such semantics.

All-in-all, I’m thankful for the piece and am coming away more mindful having read it - I just think some of the most meaningful insights deserved more of a spotlight.

Expand full comment

Good call! I actually cut a tonnnn of that, but I think you're right -- could have cut more.

Expand full comment

No biggie - no need for us lazy audience members to have ChatGPT summarize your great work yet!

Expand full comment

Awesome writeup. While Google as a SE still a long way from being replaced, with a big bang in data availability and collection, a specific query answering is the immediate need of this decade. IMHO, LLM has miles to go and need refinement in understanding the cultural nuances when a query is sought. So Google can and still address this by location basis. Is my understanding very rudimentary?

Expand full comment

A great reminder that business models are true disrupters. Thanks for the depth of analysis on this piece.

Expand full comment

Fantastic piece. You deliver with storytelling every time, but posts like these really hit home. It's Stratechery x Pop Culture/Packy's Voice indeed.

Expand full comment

Thanks, Louis! I really appreciate that.

Expand full comment

Great piece! Well articulated and reasoned! Your taxonomy for being positioned is refreshing and thinking about this from a business setting (or the military setting...loves the bombers over medieval camp) is one that should be taught in strategy, philosophy, and business.

As I see it, AI is far from a competitor to google or search engines in general. I see it as a complement that adds to the search by providing a narratives to go along with additional information.

Expand full comment

Packy, I am so old that I can remember somebody telling me about two great new search engines; one is called Dogpile and the other is Google. I guess that says something about predicting the future.

I'm not worried about positioning for Chat GPT or its children ---- they are going to resurface the whole of Western civilization, not unlike the advent of computers. I'm thinking the first effect will be a full employment act for copyright attorneys.

Expand full comment

Fascinating piece and perspective. Thank you.

Expand full comment

This was an excellent way to start the morning. Your discussion of positioning vs disruption made sense, and the diversity/depth analysis followed nicely. I feel like I learned a lot about relevant industry behavior, thank you!

Expand full comment

That was an excellent piece. Thank you.

Expand full comment

Thanks, David!

Expand full comment

I disagree with the positioning of Google vs. LLM.

What an LLM is really, is a new intermediary between the user and search. This is the precise model which Siri and Alexa were trying to push.

If the user chooses the intermediary instead of doing the search themselves via the Google search bar, then the provider of search can be switched out invisibly and cost free to the user thus transferring market power from Google's market leading search to the intermediary.

This is nothing more than simple disintermediation and is not the same as being invested in something which no longer matters.

Nor is the view that Google is helpless in the face of LLMs, necessarily valid.

LLMs by definition are dependent on search results. Google is the largest search, one of the 2 largest mobile OS's, very large in email, cloud, etc etc.

Who is to say that Google cannot identify LLM data pulls and, subtly or unsubtly, sabotage the results? Maybe using their own LLM as the counter...

Nor is your medieval army analogy a good one. Bombers require an enormous industrial overhead to build and support; they need planes, pilots, bombs, fuel and airfields - even decent cartography. The notion that they could appear out of nowhere is pure fantasy. This isn't Game of Thrones with magical dragons that appear out of nowhere...

A better example would be the introduction of the stirrup into mounted warfare. Small, easily introduced but enables all manner of things including mounted archery, lances, vastly improved horse to horse as well as mounted vs foot combat. But the stirrup itself doesn't change much...

Are LLMs like the stirrup? Far from clear to me.

Expand full comment

basically, being deeply positioned is not ergodic enough?

Expand full comment

Really enjoyed this article and appreciated revisiting the actual definitions of Innovator's Dilemma, what you've identified with Positioning is much more fitting.

I find myself slightly disagreeing with two broad points from your article:

- Is it really the case that LLMs are strictly better than search? For a lot of queries out there - timely lookups, location based searches, fact checking, etc - seem better suited for a "traditional" Google search. I think LLMs will thrive in highly iterative or generative contexts and will open up altogether new product categories, I'm not sure how much I see them as an outright search killer.

- I think Google's investment and productization of AI is a bit understated. They did the original research on Transformers/Attention and then productized it with BERT and iterated on it with MUM. Google gets a lot of (probably deserved) flack as a risk-averse bureaucracy, but from an outsider's perspective it seems like they've been constantly improving search to product their 90% market share versus just letting the cash machine run.

I wrote up my take on these issues here: https://joshs-blog-33531c.webflow.io/article/the-search-ai-wars

In any case, great read, really got me thinking!

Expand full comment

What Google lacks is a strategy for how to integrate AI with their various businesses, and perhaps they're already on the move.

Expand full comment

Excellent write up. This reminds me of https://www.ft.com/content/3c1ab748-b09b-11e8-8d14-6f049d06439c by Tim Harford on "Why Big Companies Squander Good Ideas" - the organizational structure of the incumbent prevents it from competing with an idea that needs a totally different business model and org structure. This is Rebecca Henderson's Architectural Innovation where the relationships between the pieces of the problem changes leading to incumbents being unable to compete.

Expand full comment