29 December 2022

The Future Of Search in 2023 & Beyond (Hint: it’s not Google)

0Frameworks

8Insights

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 1

Myth Buster06:30

Better Personalized Answers Can Create a Worse Information Bubble

The hosts caution that selecting only preferred sources can reinforce existing beliefs rather than broaden understanding. AI search must balance relevance with discovery so users still encounter unfamiliar perspectives and information outside their established worldview.

Source selection can intensify ideological filter bubbles
Unexpected results can help users escape preconceived views
Discovery remains as important as direct answer retrieval
Relevance should not eliminate exposure to unfamiliar information

“if I'm allowed to pick and choose my sources, well, I can really just like live in my little bubble.”
Kieran Flanagan · 06:30

“Discovery is just as important as search.”
Kipp Bodnar · 07:30

#filter bubbles#discovery#personalization#media literacy

Hot Take· 2

Hot Take03:00

AI Will Fragment Search Into Purpose-Built Engines

Kipp argues that users do not need one universally personalized search engine. Instead, AI will enable many specialized search experiences optimized for particular subjects, media formats, and user intentions.

One dominant search engine may give way to many specialized engines
Different tasks benefit from different search experiences
AI is the catalyst that makes search fragmentation practical

“we don't need search to be personalized, we need different types of search engines for different things that we're trying to do.”
Kipp Bodnar · 03:00

“search is going to fragment dramatically and AI is the technology catalyst to enable that fragmentation.”
Kipp Bodnar · 03:30

#search#artificial intelligence#personalization#fragmentation

Hot Take08:00

Google's Caution Could Turn AI Search Into an Existential Crisis

The hosts criticize Google's slow release of AI products, attributing the delay to fear of reputational damage if something goes wrong. They argue that protecting the incumbent business and public image may leave Google exposed while Microsoft-backed OpenAI moves faster.

Large companies can stagnate when reputational caution dominates
Google possesses strong AI technology but has been slow to release it
Microsoft's investment in OpenAI increases competitive pressure
AI search presents a rare existential threat to Google's model

“This is why big companies stagnate.”
Kipp Bodnar · 08:00

“this is the first real like existential crisis for them that I can remember in the last decade at least.”
Kipp Bodnar · 09:00

#google#innovation dilemma#openai#microsoft#competition

Explainer· 2

Explainer03:40

Why Embedding the Internet Could Cost Just $50 Million

The hosts discuss an OpenAI engineer's estimate that the entire internet could be embedded for approximately $50 million. They interpret the estimate as evidence that advances in AI infrastructure could make credible alternatives to Google economically attainable.

Embedding costs have fallen dramatically
Internet-scale semantic retrieval may no longer require billions of dollars
Lower infrastructure costs weaken assumptions about Google's invulnerability

“$50 million to embed the entire.”
Kipp Bodnar · 04:00

“We thought Google was insurmountable.”
Kipp Bodnar · 04:30

#embeddings#openai#search economics#google

Explainer07:20

AI Is Merging Search With Social-Style Discovery

Kipp contrasts Google's historical role as the place for intentional searches with social platforms' role in discovering unexpected material. He predicts that AI will bring these formerly separate modes together, changing how companies and brands earn attention.

Google historically dominated intentional information retrieval
Social platforms became the primary discovery mechanisms
AI is converging direct search and serendipitous discovery
Marketing through search will change substantially over five years

“for a long time Google was search and then social apps, like Facebook, like Insta, like TikTok, were our discovery mechanisms.”
Kipp Bodnar · 07:30

“Those things are coming together in the world of AI”
Kipp Bodnar · 07:30

#discovery#social media#search marketing#ai

Q&A· 1

Q&A05:50

Who Funds Publishing When AI Search Sends No Traffic?

Kieran identifies an unresolved economic problem for AI-mediated search: models depend on publishers, but conversational answers may neither credit those publishers nor send them visitors. The existing exchange between Google and content creators could collapse without a replacement incentive.

AI systems still require publishers to produce source material
Chat interfaces may answer questions without sending referral traffic
Publishers receive little credit for content used to train models
AI search needs a sustainable value exchange with content creators

“In a chat world that relationship does not exist.”
Kieran Flanagan · 06:00

“There's not a clear motivation for why I would still publish content because I get no credit for that helpin' to train your model.”
Kieran Flanagan · 06:00

#publishers#content economics#ai search#referral traffic

Tool· 2

Tool01:17

Turn an Entire Podcast Archive Into a Specialized Search Engine

Kieran describes using AI to search and question a curated podcast archive instead of relying on a general-purpose search engine. A Huberman Lab chatbot demonstrates how a narrowly trained interface can retrieve relevant answers from audio content that was previously difficult to search.

AI can make podcast archives conversationally searchable
Specialized chatbots can answer questions from a single trusted corpus
Audio and video search stand to improve substantially

“I can have a search engine for podcasts.”
Kieran Flanagan · 01:30

“our ability to use AI to make audio video search way better is gonna be huge.”
Kipp Bodnar · 03:00

#ai search#podcasts#chatbots#content discovery

Tool04:50

A Search Subscription Built From Sources You Already Trust

Kieran imagines a subscription service offering multiple search experiences built around trusted sources for travel, business, health, and other domains. Kipp extends the idea by suggesting that consumption history could establish the initial source set, with a browser extension adding new sources over time.

Users could maintain different trusted corpora for different subjects
Existing consumption habits could seed a personalized source list
A browser extension could continually add trusted publishers
Curated search could filter noise from broad internet results

“The service I would build is basically a customizable subscription service to a multitude of different search experiences.”
Kieran Flanagan · 05:00

“We could take all the consumption habits you already have and ladder them through and build a baseline of trusted sites.”
Kipp Bodnar · 05:30

#trusted sources#personalization#search product#curation