Subscription Fatigue

The allure of subscriptions for businesses is undeniable. Predictable revenue? Check. Steady customer base? Check. An excuse to regularly engage with customers? Triple check. In theory, it’s a win-win: businesses get a constant cash flow, while customers enjoy ongoing access to products or services. But that model doesn’t factor in inflation, changing consumer attitudes, and mass AI adoption, as noted in Scott Worden’s timely analysis.Realistically, I don’t want to read an article or research paper; instead, I want my custom GPT to have access to the text to summarize it for me!

Beneath this surface lies a complex puzzle of value versus volume. Consider this: a recent study showed that the average consumer spends hundreds of dollars on subscriptions annually, often for services seldom used. Many people (42%) have forgotten that they’re still paying for a subscription they no longer use. This isn’t just about cash flow anymore; it’s about redefining value in the digital age.

Right now, my monthly bank statement looks like a laundry list of subscriptions — from streaming services and software to meal kits and meditation apps. Each subscription is a tiny financial commitment, seemingly inconsequential in isolation. Do I really need a $4.99 subscription to read a friend’s article on Medium?

The Paradox of Plenty

Businesses need to rethink not just pricing, but value. What does a subscription offer that can’t be found elsewhere? Is it indispensable, or just nice to have? The key is to shift from a mindset of “recurring revenue” to “recurring value.” The subscription fatigue is real. The problem isn’t just the cost; it’s the commitment. Subscriptions lock us into ongoing payments, often for services we use sporadically. The solution? Micropayments.

Micropayments: Breaking the Buffet into Bite-Sized Pieces

Micropayments are the embodiment of practicality. You read an article, you pay a few cents. You watch a video, you pay a bit more. It’s as simple as that. This isn’t just a pivot from subscription fatigue; it’s a reflection of a growing consumer demand for fairness and value. The 2023 ECB report on micropayments echoes this sentiment, highlighting a shift towards transparent and equitable pricing. This change resonates with our evolving attitudes towards consumption — we’re becoming mindful spenders, not just passive consumers.

Why Haven’t Micropayments Taken Over Yet?

So, what’s holding back this seemingly perfect solution? The road to micropayment adoption is paved with technological and psychological hurdles. Traditional payment methods buckle under the weight of tiny transactions due to high processing costs. However, with Apple Pay, Google Pay, and the ever-evolving world of cryptocurrency, these barriers are starting to crumble. Forrester’s prediction? By 2024, we’ll see a global content provider embracing micropayments to counteract subscription cancellations, potentially revolutionizing digital content consumption.

Speaking of cryptocurrency, let’s not forget its significant role in this paradigm shift. The 2022 CoinDesk article on Bitcoin’s unfinished business: Why Micropayments Still Matter highlights why we have yet to see a mainstream micropayment solution. In a nutshell, despite numerous advancements, the cost of executing a single transaction on Bitcoin continues to be relatively expensive.

Yet, the story doesn’t end there. The many well-documented adoption hurdles have not prevented the development of cryptopayments and feeless DLTs. Platforms like SatoshiPay showcase how blockchain can facilitate small transactions efficiently and economically. They’re proving that the dream of practical, widespread micropayments isn’t just a fantasy but a work-in-progress.

Micropayments: An Essential Puzzle Piece of The AI Future

The implications of micropayments extend well beyond buying articles or streaming videos. Picture a world where machines transact with each other — where your AI assistant pays for the electricity it uses or for each file it shares over a decentralized network. The potential market here is massive and largely untapped. One of the most compelling benefits of micropayments lies in their ability to drive the growth of cheaply available intellectual property and allow users to monetize their data.

In an era where businesses distribute free content as bait for our attention, micropayments are not just an alternative but a respite for our depleted attention spans. They incentivize creators to reserve their best work for those who truly value it. It’s about ushering in a stream of quality content that doesn’t just aim to catch our attention but truly deserves it.

“Micropayments will play a key role in fueling this growth of cheaply available intellectual property.” -Scott Worden

It’s a move from the all-you-can-eat buffet of digital content to a more curated, à la carte experience that’s in sync with our individual tastes and wallets.

In the long run, this would lead to the creation of a new business model where consumer data is not just harvested but valued and traded transparently and ethically in a mutually beneficial manner.

So, would you rather stick to the buffet of subscriptions, or are you ready to embrace the tailored experience of paying just for the what you truly want?

Afterthoughts:

Not all content is created equal, so I wonder how I feel if I paid for a video or article that didn’t deliver? I’ll probably consume less?

🤖Disclaimer: This article was generated with the assistance of ChatGPT4 and Perplexity to enhance research and content creation, reflecting a collaborative effort between human input and artificial intelligence.

I attended a birthday party recently, where I profoundly experienced what the Danish philosopher, Søren Kierkegaard, describes as Aesthetic Existence, or the Either/Or. This wasn’t an isolated incident or a first, but it was the first party where…I had the out-of-body experience of watching myself “be” at a party.

Søren Kierkegaard, “the father of existentialism”, is best known for his introspective journals and critiques about individual authenticity in the context of mass media, society, and religion — and that paradoxical feeling of isolation amidst social gatherings, of “being alone in a crowd.”

The Party as a Microcosm of Society

In one poignant journal entry, Kierkegaard speaks of attending a party where he was the “life and soul.”

In his words:

I have just come back from a party where I was the life and soul. Witticisms flowed from my lips. Everyone laughed and admired me — but, I left, yes, that dash should be as long as the radii of the earth’s orbit — — — and wanted to shoot myself (I A 161).

This duality — of being celebrated externally while grappling with internal despair is not an experience reserved for 19th-century philosophers. It is more common nowadays than ever before. Let me explain:

Despairing Kierkegaard in Three Acts

For Kierkegaard, despair is a spiritual sickness, and it’s the result of not being in a right relationship with oneself and ultimately one’s chosen concept of “God”. The Crowd is a mask for despair. Kierkegaard believed that “the crowd” offered individuals an escape from the responsibility and anxiety of individual existence. When submerged in the crowd, the individual can hide from their true self, thereby falling into a state of despair without even recognizing it.

There are types (and arguably phases) of despair according to Kierkegaard:

• Despair of Ignorance: ignorant despair is when an individual is in despair without realizing it, often mistaking it for contentment or indifference.

• Despair of Weakness: This arises when an individual recognizes their despair but feels powerless to change it. They’re aware of their unhappiness or lack of fulfillment but feel trapped in their situation.

• Despair of Defiance: This is a more intense form of despair, where the individual recognizes their despair and chooses to remain in it, often in defiance of “God” or their true self. It’s a willful turning away from one’s essence or potential self.

As someone who has attended many parties and felt out of place, oblivious to my own suffering — I have intimately experienced Ignorant Despair only to have it morph into Despair of Defiance at this weekend’s birthday party.

Kierkegaard was wary of ‘the crowd,’ an impersonal, faceless entity that often stifles individuality. Parties, with their fleeting conversations and ephemeral connections, epitomize this. While attendees laugh and converse, how often do they truly connect, transcending the superficial?

Kierkegaard’s post-party desolation ties into his broader exploration of despair. This isn’t mere sadness but an existential crisis — a dissonance between one’s external actions and internal convictions. The party becomes a metaphor: even surrounded by people, one can feel profoundly alone.

Jean-Paul Sartre, another prominent existentialist, illustrates the eerie nature of societal performances in his “waiter” analogy which he used to describe the concept of “bad faith” (“Mauvaise foi”) in his work Being and Nothingness.

The Really Good Waiter

Sartre describes a waiter in a café who is overly engrossed in his role. The waiter moves too quickly, too precisely, showcasing an exaggerated performance of “being a waiter.” For Sartre, this performance is an example of “bad faith.” The waiter, in over-identifying with his role, denies his own freedom and reduces himself to a mere object, a “waiter,” rather than recognizing the depth and freedom of his existence. At a party, one is expected to be joyful and embrace the role of a partygoer, letting all other aspects of ‘being’ fade away

By clinging to a prescribed role are we denying ourselves the experience of “being at a party?” And by choosing to be a “party-goer” putting on a brave face, pretending to be joyful at at time when the world feels like it’s falling apart, are we becoming objects? Wilfully sinking deeper into a state of Defiant Despair?

Both Sartre’s “bad faith” and Kierkegaard’s concept of “despair” involve a kind of self-deception. For Sartre, it’s the denial of one’s inherent freedom, while for Kierkegaard, it’s the disconnect between one’s external actions and internal convictions. Kierkegaard’s concerns about the “crowd” can be likened to Sartre’s waiter’s over-identification with his role. In both cases, the individual is subsumed by an external identity or expectation, leading to a loss of genuine selfhood, of losing oneself in roles prescribed by society. “Mauvaise foi” is enacted by individuals who are deceiving themselves into thinking they do not have the freedom to make choices, effectively avoiding the anxiety that comes with freedom and responsibility.

And then there’s #TubeGirl, I envy you if you haven’t come across this latest manifestation of surveillance capitalism.

Social media is the contemporary embodiment of “the crowd.” It offers a space where individuals can lose their authentic selves in the mass of likes, shares, trends, and viral content. Surveillance capitalism thrives on predictability. This creates a feedback loop on social media: algorithms show content that reinforces existing beliefs, preferences, and behaviours, which in turn influences users to conform to these patterns, to unknowingly merge with a crowd.

Similarly, Sartre’s concept of “the gaze”, how the very awareness of being watched can alter one’s behaviour, is on display in every post, reel and story on instagram. The omnipresent surveillance mechanisms of social media platforms is the digital extension of Sartre’s “gaze.” Users modify their behavior, curating their online selves, due to the conscious (or even subconscious) awareness that they are being watched and judged. The “self” presented online is frequently a performance tailored for digital consumption, influenced by the invisible pressures of capitalist surveillance mechanisms.

TikTok is redefining culture

TubeGirl performs her commute on TikTok, like many other vloggers, creators, and social media users before her. She relinquished her privacy in exchange for “views” and eventually brand deals. When users post, like, share, or even scroll on social media platforms, they generate a constant stream of behavior data. While it may seem like an act of free expression or individuality, it is in fact surplus behaviour data — taking aspects of your life not tracked by the surveillance capitalism machine and freely uploading them to be harvested and commodified by tech companies. So, what appears to be an individual’s genuine self-expression becomes raw material for profit in the capitalist surveillance state. #TubeGirl is simply a monetization of the mundane, of the individual internalizing the Gaze and the Crowd and living in defiant despair.

In a world shaped by the invisible strings of surveillance capitalism and the ever-present societal gaze, I find myself oscillating between Kierkegaard’s notions of despair and Sartre’s portrayal of “bad faith.” Both philosophers, separated by time and tradition, converge on a shared existential crisis: the challenge of preserving one’s authentic self amidst societal pressures and self-imposed roles. Social media, in all its allure, not only exemplifies this crisis but amplifies it. It’s the grand stage where I, you, we, willingly or unwittingly, don the mask of the crowd, where our every click and scroll fuels the very mechanisms that entice us into conformity.

#TubeGirl’s daily commute, a seemingly bold act, is but a performance for the perpetual gaze, an enactment of identity dictated more by the invisible algorithms and societal expectations than genuine self-expression. An exploitation of free labor by tech companies, the ultimate modern day pyramid scheme.

Kierkegaard’s party, in all its paradoxical loneliness, reflects our modern gatherings — both digital and physical. We’re amidst a sea of faces, voices, and notifications, yet find ourselves seeking a genuine connection, a momentary escape from the crowd — the algorithm. Perhaps, the power lies in recognizing this tension, in understanding that while the platforms of expression have evolved, the quest for authenticity remains a timeless challenge, one that can be found in any group setting.

As I navigate this digitized landscape, I have to remind myself that behind every post, every like, and every share, there lies a choice — a choice to conform or to express, to hide in the crowd or to seek the self amidst its noise, even in an age where the line between watcher and watched is blurred. I have to recognize that every time I pull out my phone, I am willingly entering the surveillance capitalism state…despairing weakly.

🤖Disclaimer: This article was generated with the assistance of ChatGPT to enhance research and content creation, reflecting a collaborative effort between human input and artificial intelligence.

If you would like to learn more about Existentialism, feel free to use my ChatGPT prompt to jumpstart your journey:

I would like to learn as much as possible about Existentialism, especially the works of Søren Kierkegaard in order to write a thoughtful article about how his philosophy influences my life. As a tenured philosophy professor, explain his main ideas, and help me identify thoughtful passages from his journals to include in my piece. Assume a conversational tone and keep replies short. Always let me know when you’re directly quoting the philosopher and when you are summarizing his ideas to avoid confusion and misquotation. Provide reputable sources whenever possible.

The convergence of blockchain and AI is what organic practices were to industrial farming practices. However, blockchain doesn’t just stop at decentralizing the farming process; It revolutionizes the very soil and tools used to cultivate AI.

The dream of an AI blockchain-powered farm is one where every tool is designed for optimal energy efficiency, every seed is traceable, and every yield contributes back to the health of the digital ecosystem. It’s about cultivating AI in a way that’s not just technologically advanced but also environmentally conscious and sustainable for future generations. (Please refer to this great source for a fantastic deep dive into all the technical ways AI can benefit from blockchain).

Decentralized Cultivation: just as organic farming emphasizes local, community-driven cultivation, blockchain brings decentralization to the AI landscape. It breaks down the large, industrial-scale farms — monopolized by tech giants — into smaller, community-managed plots. This approach ensures a diverse array of developers, from individual enthusiasts to small tech firms, can contribute to and benefit from AI’s growth.

Resource Efficiency: In traditional farming, large-scale operations can deplete the soil, requiring heavy use of resources. Similarly, conventional AI models often strain computational resources. Blockchain introduces methods akin to crop rotation and soil conservation in organic farming, efficiently managing computational tasks across a decentralized network to keep the digital ‘soil’ fertile and productive.

Transparent, Traceable Practices: Organic farming’s value lies in its transparency, allowing consumers to trace produce from farm to table. Blockchain brings a similar level of transparency to AI. Each data input, model adjustment, and output in a blockchain-backed AI system can be tracked and verified (ZK), ensuring the integrity of AI systems just as with organic produce.

Economic Inclusivity: Blockchain democratizes the economic landscape of AI. In the organic farming world, small-scale farmers find new markets and fair prices for their produce. Similarly, blockchain enables smaller players in AI to find their footing. By facilitating fair compensation for data and computational contributions, it supports a more inclusive AI market.

Sustainable Growth: Ultimately, the integration of blockchain in AI nurtures a more sustainable, equitable growth path — much like how organic practices promise a healthier, more sustainable future in agriculture. It’s about planting seeds for an AI that grows efficiently and robustly, enriching the digital ecosystem for generations.

This is how I idealize the benefits of combining two of the planet’s buzziest technologies. It’s easy to get carried away with win-win scenarios and the democratization of technology. Treating foundation models as a common good rather than pure intellectual property will require significant mindset and economic shifts. But as AI laws are still being written, maybe it’s okay to indulge in hopeful ideals?

🤖Usual AI disclaimers apply. Who even publishes anything nowadays without running it through ChatGPT first?

First, A Quick Lesson on AI Mechanics

(Skip ahead if you already know about tokens, weights, model architecture, and the difference between training and inference):

Before an AI like ChatGPT can process a message, it breaks down the text into ‘tokens,’ much like a reader parsing words and sentences. These tokens are the basic units the model uses to understand and generate language.

Each model has a unique architecture or a neural network, formed much like a human brain through learning (training). It includes layers of neurons (mathematical functions), each layer learning different aspects of language. More complex architectures can grasp subtleties in language but require more energy to run. The challenge for developers is designing architectures that strike a balance between sophistication and energy efficiency. Just like how a gardener decides which plants to grow and where to place them.

Before this garden can thrive, it needs to be nurtured — this is where ‘training’ the AI model comes in. Think of it as the intensive phase of tending to the farm: preparing the soil, planting seeds (feeding data) into the soil (neural network) and ensuring they receive the right amount of water and nutrients (computational resources) to grow.

Once the AI model, or crop, is mature, we enter the ‘inference’ phase — the harvest time. Here, the AI utilizes its training to make predictions or decisions, similar to reaping the produce from a mature, well-tended farm. This phase is like routine maintenance of a flourishing garden, less intensive than the initial growth period but still needs water and tending.

The size of an LLM is often directly related to the number of weights it has. Weights (how many decimal places a response is accurate to) can also affect energy consumption. Larger models (bigger farms) with more weights can capture more nuanced patterns in data but also require significantly more computational resources to train and run. Using lower-precision, or to stick with the garden analogy, “ less resource-intensive farming methods” will reduce the amount of memory and computation needed.

The Energy Debate: Not Just a Crypto Problem

Remember the uproar over Bitcoin’s energy consumption? Similar to the cryptocurrency and blockchain industry, which has faced scrutiny for its environmental impact, there’s a similar storyline emerging for AI. Both industries demand a high energy input for complex calculations — crypto for securing transactions and AI for processing vast amounts of data.

Understanding the Scale

To put it in perspective, training a single AI model can consume as much electricity as several hundred homes do in a year.

If every search on Google used AI similar to ChatGPT, it might burn through as much electricity annually as the country of Ireland. Why? Adding generative AI to Google Search increases its energy use more than tenfold. (-The Verge)

The reason? AI, especially deep learning langauge models, require a vast amount of computational power. These models learn from huge datasets, and crunching these numbers is like trying to solve a jigsaw puzzle the size of a football field, where each piece constantly changes shape.

Consider self-driving cars, a pinnacle of AI application. These vehicles don’t just ‘see’ the road; they process a constant stream of data from sensors, cameras, and GPS systems. This processing happens almost in real-time and demands a continuous, high-powered computational effort. Multiply this by thousands of AI applications, from voice assistants to automated trading systems, and you start to grasp the scale of the energy challenge, or as this April 2023 Research Paper on AI trend points out:

As more and more devices use AI (locally or remotely) the energy consumption can escalate just by means of increased penetration, in the same way that cars have become more efficient in the past two decades but there are many more cars in the world today.

Three Pioneering Solutions: Gensyn, Box, and Sakana

Innovative solutions are emerging, addressing the energy challenges of AI with creativity and technological ingenuity.

Box AI: On the business model front, Box AI is redefining how AI services are priced with a unique credit system. In an industry where running foundational models like OpenAI’s can be prohibitively expensive, Box AI’s approach is refreshingly practical. Users receive a set number of credits for AI interactions, putting a cap on usage. This model indirectly limits the energy consumed by each user, making them more conscious of their AI usage. For power users, additional credits are available, ensuring scalability while still promoting mindful consumption. This innovative pricing strategy not only addresses cost concerns but also subtly encourages users to consider the energy impact of their AI usage. This model reminds of water rationing practices in farming–an essential practice for the common good!

Sakana AI: Sakana AI is exploring an entirely different yet “natural” frontier. Based in Tokyo and led by former Google AI luminaries, this startup is drawing lessons from nature to build the next generation of AI. Inspired by the fluidity and adaptability of natural systems, like schools of fish (derived from the word for fish さかな in Japanese), they are working on AI models that are more flexible and responsive to their environment. This nature-inspired approach could lead to AI systems that are not only more efficient in terms of energy consumption but also more robust and adaptable in their functionality.

Gensyn: By creating a blockchain-based marketplace, Gensyn democratizes access to computational power. This platform taps into a global network of underutilized resources, from smaller data centers to personal gaming computers, harnessing their latent power for AI development. It’s a game-changer, potentially increasing available compute power for machine learning by an order of magnitude. The integration of blockchain in AI goes beyond just a supporting infrastructure. It can actively offer new ways to handle data, ensure security, and create economic incentives for data sharing and model training. I like to think of this approach as permaculture!

These diverse approaches by Gensyn, Box AI, and Sakana AI reflect a growing awareness in the tech community of the need to balance AI’s benefits with its environmental costs. Each solution brings a unique perspective to the table, collectively pushing the boundaries of what’s possible in sustainable AI development and business.

The Blockchain-AI Intersection:

A note on blockchain:

It’s no secret that so many crypto developers flocked to “AI” this year. Yes, it could be about dwindling investments in web3. But I’d like to think that it’s because of a shared ethos and bright-eyed idealism of all open source movements, dating back to the early days of the INTERNET itself.

Transparent information sharing, creating viable alternatives to traditional methods of conducting business, individual transactions, owning one’s assets, data, and identity. All these ideas are central to the ethos of the blockchain movement, and are also applicable in the AI space. The convergence of blockchain and AI is what organic practices were to farming practices, But blockchain doesn’t just stop at decentralizing the “farming industry”. It revolutionizes the very soil and tools used to cultivate AI. The dream of an AI blockchain-powered “farm” is one where every tool is designed for optimal energy efficiency, every seed is traceable, and every yield contributes back to the health of the digital ecosystem. It’s about cultivating AI in a way that’s not just technologically advanced but also environmentally conscious and sustainable for future generations— more on this in a future article by popular demand!

I’d like to think that it’s because of a shared ethos and bright-eyed idealism of all open source movements, dating back to the early days of the INTERNET itself.

Beyond Blockchain: Technological Advances and Strategies

As we all grapple with the energy demands of AI, several energy-efficient strategies are emerging, each playing a crucial role in reducing the computational footprint of AI systems. Here’s a quick breakdown of some key strategies that are making AI more sustainable:

Model Quantization: Think of quantization as a way of making AI models ‘lighter’ in terms of data, selecting the most essential nutrients for growth — just like, not everything in the soil is needed for a plant to thrive. In technical terms, it involves reducing the precision of the numbers used in the model’s calculations. For example, instead of using numbers with a long decimal tail, quantization might round these numbers off. This rounding-off significantly reduces the computational load and memory usage, leading to less energy consumption. It’s like compressing a large photo file into a smaller size so it’s easier to send via email, but still clear enough to view.

Distillation: Model distillation is like extracting the essence of a plant, similar to creating a potent herbal extract from a larger mass of raw material. This method ensures that the smaller model retains the effectiveness of the larger one but with a fraction of the energy cost. It’s similar to distilling the key insights from a thick textbook into a concise study guide that’s quicker and easier to use.

Sparsification: Sparsification involves removing redundant or non-essential information from AI models similar to pruning unnecessary branches from a tree to promote healthier growth.. This ‘thinning’ helps the model to process information more quickly and efficiently.

Custom Hardware for Energy-Efficient Inference and Training: Developing custom hardware for AI is like crafting specialized farming tools designed for specific tasks. These tools — tailored for AI’s unique requirements — handle computational processes more efficiently than general-purpose tools.

More Good News

A recent MIT study highlights how simple hardware interventions, like GPU power capping, are being deployed with a big impact on energy costs. So, it really is about refining our techniques and also marginally increasing our patience.

Limiting GPU power to 150 watts saw a two-hour increase in training time (from 80 to 82 hours) but saved the equivalent of a U.S. household’s week of energy. (MIT News)

The invisible energy cost of AI is a critical issue we cannot afford to ignore in the name of progress. By supporting sustainable AI research, adopting energy-efficient technologies, and educating ourselves on the environmental impact, we can make more conscious choices in our AI usage. Just as buying organic supports agricultural practices that are better for the environment, opting for sustainable AI promotes technology that is environmentally responsible. Our choices significantly influence the sustainability and ethical implications of consumption, whether it’s the food we eat or the technology we use.

🤖Disclaimer: This article was generated with the assistance of ChatGPT4 and Perplexity to enhance research and content creation, reflecting a collaborative effort between human input and artificial intelligence.