09 May 2023

A.I. Data Wars: Why Elon Is So Scared Of OpenAI (#116)

1Frameworks

10Insights

Frameworks in this episode

Strategy6 steps

Three-Front AI Data Audit

Audit data ownership, public exposure, and regulatory risk before AI reshapes your market.

Read →

Insights & moments

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

Hot Take· 3

Hot Take04:30

Meta's Closed Data Strategy Could Become an AI Advantage

Kipp argues that Meta's longstanding restrictions on access to Facebook data now give it a differentiated asset for AI development. The discussion connects that advantage to a broader return of closed platform ecosystems.

Meta retained control over a large proprietary social dataset.
Controlled data can differentiate models as model technology commoditizes.
Platforms increasingly suppress outbound links and keep engagement inside their ecosystems.
The open-web distribution benefits once offered by social platforms are weakening.

“Zuckerberg was right.”
Kipp Bodnar · 04:30

“everyone's gonna build this kind of closed, walled experience.”
Kieran Flanagan · 04:30

#meta#walled gardens#social media#ai

Hot Take08:00

Why Reddit May Be More AI-Resilient Than Stack Overflow

Kieran distinguishes community conversation from transactional question answering. He expects Stack Overflow's core utility to be easier for ChatGPT-like systems to replicate, while Reddit retains more value from human-to-human interaction.

Both platforms contain user-generated data they can restrict or license.
Stack Overflow primarily helps users obtain discrete answers.
AI assistants can reproduce transactional question-answering experiences.
Reddit's conversational and community dimensions may be harder to replace.

“Stack Overflows really is just a question and answer site.”
Kieran Flanagan · 08:30

“Reddit it's much more like a human to human interaction, right?”
Kieran Flanagan · 08:30

#reddit#stack overflow#communities#chatgpt

Hot Take19:00

AI Tax Tools Could Make Previously Exclusive Loopholes Widely Accessible

Kipp predicts that conversational access to complex tax codes could let ordinary taxpayers identify strategies previously surfaced by expensive professionals. Widespread use could alter tax revenues and force governments to revise legislation.

Complex tax knowledge has historically favored people able to hire specialists.
AI can make statutory information easier for non-experts to interrogate.
Broader exploitation of legal tax provisions could affect government revenue.
Governments may respond by simplifying or rewriting rules.

“it's going to remove a lot of loopholes.”
Kipp Bodnar · 19:00

“the average person who can't spend tens of thousands of dollars a year on like a tax attorney and a CPA, can now go and…”
Kipp Bodnar · 19:30

#tax#regulation#accessibility#government

Explainer· 5

Explainer02:30

Why Google's Value Exchange Softened the Original Data Conflict

Kieran contrasts Google's rise with the emerging conflict around large language models. Publishers tolerated Google's aggregation partly because search returned traffic and because building a competing search engine was difficult.

Google organized public web content into a defensible search product.
Publishers received referral traffic in exchange for being indexed.
Large language models may compete more directly with the sources supplying their data.
Organizations such as Reddit can build specialized models from their own content.

“Hey, we are gonna organize all this together, and show your content when it's applicable, and give you traffic.”
Kieran Flanagan · 03:00

#google#search#publishers#data

Explainer05:30

Large Language Models May Become Each Platform's New Algorithm

The hosts compare proprietary language models with the recommendation algorithms that differentiate major platforms. They expect many models rather than one universal winner because unique platform data will shape each model's behavior and value.

Model technology is likely to become increasingly commoditized.
Proprietary datasets can remain a meaningful source of differentiation.
Early distribution gives OpenAI a major advantage but not guaranteed permanence.
The market may resemble fragmented cable television rather than a single consolidated service.

“We're still gonna have this very fragmented, cable TV-esque experience,”
Kipp Bodnar · 06:00

“No lead is insurmountable, except maybe Microsoft and Apple.”
Kipp Bodnar · 07:00

#llms#platforms#algorithms#competition

Explainer09:30

AI Assistants Threaten Booking and Data Middle Layers

The hosts argue that conversational systems can connect users more directly to answers and transactions. Businesses that mainly aggregate information or sit between a platform and an end service may lose their role when an AI assistant performs discovery and completion in one interface.

Aggregation often produces disintermediation elsewhere in the value chain.
Browsing and transactional AI reduce the need to visit middleware services.
Booking and data aggregation businesses face similar structural pressure.
Intermediaries need value beyond assembling information.

“what always comes with aggregation Kieran, disintermediation,”
Kipp Bodnar · 10:30

“a lot of the middleware that exists in data, is just gonna go away.”
Kipp Bodnar · 11:00

#disintermediation#aggregators#travel#business models

Explainer12:00

Vertical AI Assistants Are Emerging Wherever Training Data Is Available

The episode highlights legal, tax, finance, and home-repair assistants built around domain-specific information. FinChat is presented as a useful example because it lets investors query filings, earnings calls, and portfolio disclosures conversationally.

Harvey.ai applies AI to legal workflows.
TaxGPT packages public tax information for accounting use cases.
FinChat covers public-company filings, transcripts, and investor portfolios.
Consumer verticals such as home repair can also support specialized assistants.
Product quality varies even when the data opportunity is strong.

“anything where you have freemium content to train on, you're gonna have vertical AI assistance.”
Kieran Flanagan · 14:30

“It is ChatGPT for finance.”
Kipp Bodnar · 15:00

#vertical ai#finance#legal#tax

Explainer17:00

National AI Models Turn Data Into Geopolitical Power

Russia's GigaChat launch after restricting ChatGPT illustrates how major countries may demand domestically controlled AI systems. The hosts expect model access and capabilities to vary by region as governments seek influence over what AI tells their populations.

Italy's ChatGPT ban highlighted conflict between model development and privacy rules.
Russia launched GigaChat after banning ChatGPT.
Large countries may prefer models they can govern or inspect.
AI capability could become a strategic national asset comparable to an arms-race advantage.

“Russia launched GigaChat after it banned ChatGPT.”
Kipp Bodnar · 18:00

“countries are gonna wanna have the most sophisticated, the most intelligent AI.”
Kieran Flanagan · 21:00

#geopolitics#gigachat#regulation#national ai

Tool· 1

Tool09:00

ZoomInfo's Free Product Turns Users Into Data Contributors

The hosts describe a freemium mechanism in which customers receive free product access by contributing contact information. That exchange helps ZoomInfo expand its dataset instead of relying exclusively on externally aggregated data.

Free access provides an incentive for users to contribute data.
User contributions expand and refresh the product's underlying dataset.
The mechanism can reduce dependence on purely external aggregation.
Data provenance still matters when evaluating the defensibility of the resulting product.

“to give away a free version, you actually have to give them some of your contact lists, or some of your data.”
Kieran Flanagan · 09:00

#zoominfo#freemium#data flywheel#b2b

Takeaway· 1

Takeaway21:30

Data Is the Scarce Resource Behind the Next Automation Wave

Kipp closes by comparing data competition with battles over cobalt, lithium, and other strategic materials. AI enables automation and productivity, but those gains depend on access to sufficiently valuable information.

AI acts as a bridge to higher levels of automation.
Automation can increase productivity across industries and societies.
Advanced systems cannot deliver those gains without suitable data.
Scarce, differentiated data will attract expense, regulation, and strategic competition.
Existing laws may need to be rewritten around AI rather than simply applied unchanged.

“AI is our bridge to the next level of automation, right?”
Kipp Bodnar · 21:30

“I don't think AI will bend to existing frameworks, and laws, and the way things are done.”
Kieran Flanagan · 22:30

#automation#productivity#data economy#ai policy