Season 6, Episode 17: Heidegger's caution

The discussion explores the monetization strategies for chatbots, particularly focusing on the limitations of affiliate models and the potential of personalized advertising. It argues that personalized advertising offers a more sustainable and scalable path for monetization, as it can leverage user data beyond just chat context and is not confined to commercial queries. The conversation highlights the risks of affiliate models, including their inefficiency in dynamic pricing and potential to undermine the content ecosystem they rely on. The broader implications of AI's role in the economy and society are also considered, drawing on philosophical perspectives from Heidegger and Dewey.

• Personalized advertising is argued to be a more sustainable business model for chatbots compared to the affiliate model adopted by OpenAI.
• The discussion explores the potential of personalized advertising to ensure the commercial viability of AI tools like ChatGPT.
• There is a need to explore the tension between different monetization strategies for consumer-facing AI.

"I've written extensively about why I believe personalized advertising is the optimal and inevitable path to durable and viable monetization for ChatGPT."

• This quote highlights the author's belief in personalized advertising as the most effective business model for AI tools like ChatGPT.

• OpenAI's deals with major chip manufacturers like Broadcom, Nvidia, and AMD underline its significant economic influence.
• The failure of OpenAI could impact global enthusiasm and investment in artificial intelligence as a transformative technology.
• OpenAI's role as a leading figure in AI development makes its success or failure pivotal to the perception of AI's potential.

"I think OpenAI presents a structural risk to the global economy because it has anointed itself the avatar of artificial intelligence as a transformative economic and social force."

• This quote emphasizes the potential global economic implications if OpenAI were to fail, impacting the perception of AI's potential benefits.

• OpenAI excels in marketing and generating excitement about AI but lacks a clear business model for sustainable growth.
• Sam Altman, OpenAI's CEO, has softened his stance on advertising, suggesting a potential shift in business strategy.
• The comparison with Netflix's eventual adoption of advertising underscores the possibility of OpenAI moving towards an ad-based model.

"Sam Altman appeared on every major technology podcast to talk about the prospects for AGI and the infrastructure costs it will incur. But he had very little to say about the business model that will enable the company to navigate that course."

• This quote points out the disparity between OpenAI's promotional efforts and its unclear business strategy.

• Sam Altman's comments on Instagram ads suggest a reconsideration of advertising's role in OpenAI's business model.
• The comparison with Netflix's earlier denial and eventual adoption of ads indicates a potential similar trajectory for OpenAI.
• The narrative around advertising in technology companies often shifts from initial resistance to eventual acceptance.

"I don't think Sam Altman would draw such a sharp distinction between Instagram advertising and the more generalized advertising model if OpenAI wasn't already exploring, if not building, a conversion optimized advertising platform."

• This quote suggests that OpenAI is likely considering or developing an advertising model despite previous resistance.

• AI has shown measurable positive economic impacts, particularly in digital advertising and marketplaces, as evidenced by major tech companies' earnings reports.
• The improvements in ad retrieval relevancy and click-through rates demonstrate AI's contribution to commercial success.
• The hidden nature of these improvements results in gradual but significant enhancements in economic performance.

"AI has produced demonstrable, measurable, significant, positive economic impact through its application to digital advertising and digital marketplaces."

• This quote asserts the tangible economic benefits that AI has delivered, particularly in enhancing digital advertising efficiency.

• There is widespread skepticism about AI's profitability, particularly concerning OpenAI's business model.
• The media often questions when AI investments will yield financial returns.
• OpenAI is seen as a representation of both AI's potential and the skepticism surrounding it.

"If OpenAI can't embrace a scalable business model, that chorus of disbelief will only grow louder again."

• OpenAI's ability to develop a sustainable business model is crucial to counter media skepticism and validate AI investments.

• Martin Heidegger views technology as a force that transforms everything into resources to be optimized, potentially reducing all experiences to efficiency.
• John Dewey sees technology as an experimental tool that extends human inquiry and intelligence.

"Martin Heidegger... describes technology not as a collection of tools, but as a means of revealing the world... He warned that as we advance technologically, we risk reducing all experience to efficiency."

• Heidegger cautions that technology could lead to a loss of freedom by imposing a mindset focused solely on optimization and efficiency.

"John Dewey... viewed technology as an experiment, as a continuous series of iterations through which humans extend inquiry into the world."

• Dewey's perspective highlights technology as a tool for enhancing human understanding and addressing human needs through experimentation.

• The debate on AI's role in digital advertising revolves around whether precision in optimization is beneficial or harmful.
• The focus shifts from the suitability of business models for chatbots to the ethical implications of digital advertising.

"The larger question... is whether ever more precision applied to optimizing an economy is a perverse or nefarious path... or if it's a course of action that thoughtful technologists should deliberately engineer."

• The ethical considerations of digital advertising's precision are more complex than choosing between business models for chatbots.

• Three flawed assumptions about chatbot monetization are identified: context as the sole monetizable signal, chatbots remaining text-based, and reliance on review data.

"The first assumption is that context is the only monetizable signal for a chatbot... chatbots differ fundamentally from search."

• Chatbots involve continuous conversations, making context alone insufficient for monetization compared to search-based advertising.

"The second flawed assumption is that chatbots will remain primarily text-based... Multimodality is the natural endpoint of interface evolution."

• The evolution of AI interfaces towards multimedia formats opens new monetization opportunities beyond text-based models.

"The third assumption is related to the knock on effects of chatbots... that a large corpus of review data will always be available for chatbots in order to serve as research agents."

• The assumption that chatbots can always access extensive review data for research purposes is questioned, impacting their potential as research tools.

• Contextual targeting is often seen as less effective than behavioral targeting, which uses historical data to target ads more accurately.
• Chatbots have the potential to leverage behavioral data for more effective advertising.

"Empirically, behavioral targeting outperforms contextual targeting when a sufficient volume of high quality intent data is available to be targeted against."

• Behavioral targeting's effectiveness relies on access to high-quality data, suggesting chatbots could benefit from similar strategies.

• The shift towards multimedia in AI interfaces suggests new monetization models akin to those used in platforms like YouTube.
• Video and other multimedia formats offer diverse ad opportunities beyond text-based affiliate links.

"If the best response to a query is a video tutorial, is that output better monetized with E commerce affiliate cards or with a pre roll or mid roll video ad?"

• The transition to multimedia formats in chatbots could align more with video ad models, enhancing monetization potential.

• Chatbots trained on internet-scale data can provide direct answers to queries, reducing the need for users to visit original content sources.
• The monetization of chatbots through affiliate links could undermine the content ecosystem by diverting traffic and commissions away from original content creators.
• This dynamic could lead to a decline in the production of original content, as creators lose the incentive to produce reviews and information.

"Chatbots train on Internet scale data and they can access websites in situ while answering questions in ways that obviate the need for consumers to go directly to the sources of that data."

• Chatbots can provide information directly, bypassing the need for users to visit the original sources of data.

"In success, instant checkout will annihilate the sources of the content that it uses for research."

• The monetization model of chatbots, if successful, could destroy the very content sources they rely on for data.

"The monetization success of the model destroys the information supply that made the model valuable in the first place."

• The affiliate monetization model could lead to a collapse of the information ecosystem that chatbots depend on.

• Affiliate monetization by chatbots could lead to a self-cannibalizing loop, whereas advertising sustains diverse information sources.
• Advertising models, particularly those using auctions, allow advertisers to bid for impressions based on true value, unlike affiliate models that use fixed percentage fees.
• The auction mechanism in advertising aligns incentives and maximizes total welfare by allowing advertisers to express their private information.

"Advertising, however, doesn't create this self cannibalizing loop. An advertiser pays for exposure, not appropriation."

• Advertising models sustain the diversity of information sources without undermining them.

"Digital advertising is mediated by an auction mechanism that allows advertisers to bid for impressions in ways that incentivize those bids to reflect the true value of the impression to them."

• Digital advertising auctions enable advertisers to bid based on the true value of impressions, unlike fixed-fee affiliate models.

"The auction mechanism accomplishes two important things. First, it lets advertisers express their private information. The platform doesn't need to know each merchant's margin, the bid reveals it."

• The auction mechanism allows advertisers to reveal their true valuations, leading to efficient value discovery and aligned incentives.

• Affiliate systems don't allow retailers to bid on conversions, potentially pricing some out of participation.
• The fixed percentage fee in affiliate systems can disadvantage retailers with lower margins or higher price points.
• Affiliate systems lack flexibility, preventing advertisers from maximizing their commercial outcomes.

"Without allowing retailers to bid on conversions, it simply prices some retailers out of participating altogether."

• Retailers with low margins may be unable to participate in affiliate systems due to fixed fees.

"Affiliate systems might favor lower cost products."

• Affiliate systems may inherently favor cheaper products due to their fixed commission structure.

"A platform is better served by giving retailers the opportunity to bid for impressions at whatever price those impressions are worth to them."

• Allowing retailers to bid on impressions can maximize platform revenue and advertiser participation.

• The goal of advertising platforms is to allocate attention to advertisers who value it most, maximizing the sum of advertiser surplus, consumer utility, and platform revenue.
• When advertisers, consumers, and platforms have aligned objectives, total welfare is maximized.
• The affiliate model's inability to differentiate value among advertisers can lead to a race to the bottom, reducing overall surplus and engagement.

"When advertisers bid truthfully, the equilibrium allocation maximizes the sum of advertiser surplus, consumer utility, and platform revenue."

• Truthful bidding in auctions maximizes total welfare by aligning the interests of advertisers, consumers, and platforms.

"The result is a race to the bottom lower prices, lower commissions, and less total surplus."

• The affiliate model's lack of differentiation can lead to decreased prices and commissions, reducing overall surplus.

• The digital affiliate model, established in the mid-90s, is critiqued for being less efficient compared to personalized advertising.
• Affiliate discovery favors low-value goods that struggle in digital advertising auctions.
• Chatbots with affiliate models may undermine their value as research tools by promoting low-margin, low-cost goods.

"The digital affiliate model is neither new nor novel. As I point out in the piece, it was pioneered in the mid-90s."

• The affiliate model is outdated and less effective in the current digital advertising landscape.

"Affiliate discovery is not allocatively efficient and it tends to favor low value goods that can't get a foothold in digital advertising auctions."

• The inefficiency of affiliate discovery leads to the promotion of low-value goods, reducing the effectiveness of chatbots as research agents.

• Personalized advertising can monetize more queries than affiliate models by not relying on the commercial context of queries.
• It allows chatbots to generate more ads and increase per ad monetization.

"Like search, the affiliate model can only monetize queries with a commercial locus."

• The affiliate model's reliance on commercial queries limits its monetization potential.

"Personalized advertising would allow chatbots to monetize more queries, and per the previous argument, it's likely that it could monetize them better than affiliate links can."

• Personalized advertising offers a broader and potentially more profitable monetization strategy for chatbots.

• Amazon opposes agentic commerce due to its significant role in digital advertising and its proprietary systems like Rufus.
• Agentic commerce threatens Amazon's direct relationship with users and its advertising revenue.

"Amazon captures roughly 40% of US E commerce revenue, so its position here is important."

• Amazon's dominance in e-commerce influences its resistance to agentic commerce models.

"Amazon is not only a retailer, but a closed loop ad marketplace layered over logistics and cloud infrastructure."

• The integration of advertising with its retail operations is crucial to Amazon's business model.

• Walmart's e-commerce operation is smaller than Amazon's, influencing its openness to third-party commerce agents.
• Walmart views third-party integrations as strategically valuable, unlike Amazon.

"Walmart's E commerce operation, while growing quickly, is a relatively small portion of overall revenue."

• Walmart's smaller e-commerce footprint makes it more receptive to agentic commerce.

"While Walmart is an omnichannel retailer, it generates a minority share of its revenue from E commerce, although that share is growing."

• Walmart's growing e-commerce segment contrasts with Amazon's established dominance, affecting their strategies towards agentic commerce.

• Chatbots are more useful for researching high-intent goods, which are expensive and require thorough comparison.
• High-intent purchases involve more scrutiny, and chatbots may not effectively capture these transactions.

"If a user is relying on a chatbot for product research, it's almost certainly for a high intent good."

• Chatbots are primarily used for researching significant purchases, where users seek comprehensive information.

"Chatbot commerce collapses at the moment of highest commercial value. It can assist in research, but it may not capture high intent transactions."

• Chatbots struggle to convert high-intent research into transactions, highlighting their limitations in monetizing expensive goods.

• Chatbots exist at an evolving intersection of design, economic incentives, and cultural trust.
• Historical precedents offer limited guidance for predicting the future of chatbot monetization.

"Chatbots are still a relatively recent phenomenon, and they're navigating an unstable and unpredictable intersection of interface design, economic incentives, and cultural trust."

• The development of chatbots is ongoing, with many uncertainties in their monetization strategies.

"This is an experiment. We can take some guidance from historical precedent, but its predictive power is limited."

• The experimental nature of chatbots means that past models may not accurately predict future outcomes.