#122 Marginal Gains - Part 3
“It’s not what you look at that matters, it’s what you see.”
Henry David Thoreau
A Few Thoughts about O1 - by Delip Rao (substack.com)
In his article "A Few Thoughts about O1," Delip Rao discusses the recent release of OpenAI's o1 models, which have been hyped as their Artificial Super Intelligence (ASI) model under the codename "strawberry." Rao critiques the performance claims, pricing strategy, and context handling of the o1 models and predicts future developments in the AI industry. He argues that the o1 models are agentic rather than traditional models and that their performance gains are overstated. Additionally, Rao criticizes OpenAI's pricing strategy and context handling and predicts that the o1 models will surpass future models from competitors.
Key Takeaways
Performance Claims
Explanation: Rao argues that the impressive performance claims of the o1 models are misleading. They are agentic systems that take multiple "turns" to reason, unlike traditional models that have only one turn to respond.
Key Quote: "The o1 'models' (which OpenAI insists on calling them as models) are agentic systems."
Why It Matters: Understanding the nature of the o1 models is important for accurately evaluating and comparing their performance to other models.
Pricing Strategy
Explanation: Rao criticizes OpenAI's pricing strategy for the o1 models, which includes hidden fees for thought tokens that users cannot see.
Key Quote: "OpenAI will be the first AI company to introduce 'hidden fees' for model/system access."
Why It Matters: The pricing strategy of AI models is important for developers and users, as it affects the technology's cost and accessibility.
Context Handling
Explanation: Rao points out that the o1 models have brittle context handling, requiring developers to carefully select relevant information to include in the context to prevent overcomplicated responses.
Key Quote: "OpenAI cautions: Limit additional context in retrieval-augmented generation (RAG): When providing additional context or documents, include only the most relevant information to prevent the model from overcomplicating its response."
Why It Matters: Effective context handling is important for the usability and performance of AI models, as it affects the quality and relevance of the responses.
Future Developments
Explanation: Rao predicts that future competitors like Meta, Google, or Mistral will surpass the o1 models and that OpenAI will eventually add function calling to them.
Key Quote: "My Prediction: It will not take more than a couple of months for the o1 results to be bested by future publicly available models from Meta, Google, or Mistral."
Why It Matters: The competitive landscape of the AI industry is important for driving innovation and improving the performance and accessibility of AI models.
Data Collection
Explanation: Rao argues that releasing the o1 models is primarily a strategy to gather data and collect complex reasoning prompts and their traces.
Key Quote: "If I were at Google or elsewhere, I would encourage a similar strategy to collect challenging problems."
Why It Matters: Data collection is important for training and improving AI models, and understanding the motivations behind releasing new models is important for evaluating their impact on the industry.
Conclusion
Delip Rao's article critically analyzes OpenAI's o1 models, highlighting the misleading performance claims, problematic pricing strategy, and brittle context handling. He also predicts future developments in the AI industry and argues that releasing the o1 models is primarily a data collection strategy. The insights provided are important for developers, users, and competitors in the AI industry, as they affect AI models' evaluation, cost, and usability.
How AI Disrupts Tech Investing - by Tomas Pueyo
Tomas Pueyo's article, "How AI Disrupts Tech Investing," analyzes the evolving landscape of tech investment in the age of artificial intelligence. He argues that AI is fundamentally changing the game, creating both exciting opportunities and significant risks for venture capitalists and other investors.Here's a breakdown of the key points, supporting explanations, and relevant quotes:
1. The Shift from Physical to Digital Investments:
Key Point: The tech investment landscape has shifted dramatically from physical industries to software and internet-based companies due to lower barriers to entry and the potential for rapid growth.
Supporting Explanation: Pueyo explains that traditional tech investments in areas like manufacturing and telecom required substantial upfront capital and faced increasing regulation. The rise of software and the internet offered a more attractive alternative with lower initial investment and greater scalability.
Key Quote: "So when the software industry exploded in the 1990s, off the back of the Internet, tech investors poured in. A testament to this is the fact that 'tech' is now a synonym for 'Internet software,' which is crazy if you think about it." This quote highlights the dramatic shift in the definition of "tech" itself.
2. The Four Types of Internet Companies and VC Strategies:
Key Point: The internet era gave rise to four main types of tech companies: hyper-competitive (e.g., video games), B2B SaaS, aggregators (e.g., Google, Amazon), and everything else. This led to two primary VC strategies: early-stage "spray and pray" and later-stage growth capital.
Supporting Explanation: Pueyo categorizes internet companies and explains the different competitive dynamics in each category. He then connects this to the evolution of VC strategies, describing the high-risk, high-reward approach of early-stage investing and the focus on established growth in later-stage investments.
Key Quote: "For investors, that meant two broad strategies: 1. Early-Stage Spray and Pray […] 2. Later Stage Growth Capital." This quote succinctly summarizes the two dominant VC approaches.
3. The Impact of COVID and the Rise of AI Startups:
Key Point: The COVID-19 pandemic, followed by rising interest rates, created a challenging environment for tech companies reliant on continuous funding. Simultaneously, AI emerged as a disruptive force, enabling new services and accelerating development.
Supporting Explanation: Pueyo explains how the pandemic initially boosted tech usage but subsequent economic changes made capital more expensive, leading to the downfall of many unprofitable startups. He then introduces AI as a game-changer, attracting renewed VC interest in early-stage AI-focused companies.
Key Quote: "In 2024, two thirds of Y Combinator’s startups are focused on AI—which usually means the first type, new services made possible thanks to AI." This quote demonstrates the significant shift towards AI within the startup ecosystem.
4. The Rise of the Solopreneur and the Creator:
Key Point: AI empowers individuals (solopreneurs) to develop and launch products with minimal capital, while content creators become crucial distribution channels, challenging traditional VC models.
Supporting Explanation: Pueyo argues that AI tools are democratizing software development, enabling individuals to build and test products independently. He highlights the growing importance of content creators in reaching audiences, bypassing traditional marketing and distribution channels.
Key Quote: "The first billionaire solopreneur is probably in his underwear right now talking with ChatGPT." This provocative statement emphasizes the potential of AI to empower individuals.
5. How AI Disrupts Tech Investment:
Key Point: AI is disrupting tech investment by increasing competition, reducing the need for capital for successful startups, and potentially lowering the success rate of VC-funded ventures.
Supporting Explanation: Pueyo argues that the ease of building AI-powered products will lead to intense competition and shorter lifecycles for new ideas. Successful startups may require less external funding, limiting VC opportunities. The rise of solopreneurs further challenges the traditional VC model.
Key Quote: "So not only might VCs (and stock market investors) be locked out of good AI investments and have their existing investments disrupted by AI startups, but their investments in startups are becoming much more precarious." This quote encapsulates the multifaceted challenges facing VCs in the age of AI.
The article concludes by acknowledging the opportunities AI presents for tech investors while emphasizing the significant risks they must navigate. It sets the stage for further discussion on how VCs, retail investors, and workers can adapt to this rapidly changing landscape.
Scaling Laws for Economic Productivity: Experimental Evidence in LLM-Assisted Translation
This paper explores the empirical relationships between the amount of training compute used for Large Language Models (LLMs) and their performance, focusing on economic outcomes. The study involved 300 professional translators completing 1800 tasks with access to LLMs of varying compute sizes. The results show that increased model compute significantly enhances productivity, with a 12.3% speed improvement, 0.18 standard deviation higher grades, and a 16.1% increase in earnings per minute for every 10x increase in compute. Notably, lower-skilled workers benefited more from model scaling.
Key Takeaways
1. Empirical Relationships Between LLM Training Compute and Performance
Key Point: The study derives empirical relationships (scaling laws) that correlate the amount of training compute used for LLMs with their performance in economic tasks.
Supporting Explanation: The paper builds on existing literature that shows a consistent relationship between model training compute and model perplexity. It extends this understanding to economic outcomes by conducting a randomized controlled trial (RCT) with 300 professional translators completing 1800 tasks.
Key Quote: "This paper aims to offer the first experimental evidence on this question by conducting a randomized controlled trial (RCT) involving 300 professional translators conducting 1800 tasks of varying complexities." This quote highlights the experimental design and the focus on real-world economic outcomes.
2. Impact of Model Compute on Productivity and Quality
Key Point: Increased model compute significantly enhances productivity, with a 12.3% speed improvement and a 0.18 standard deviation higher grade for every 10x increase in compute.
Supporting Explanation: The results show that for every 10x increase in model training compute, translators completed tasks 12.3% quicker and received higher grades, leading to a 16.1% increase in earnings per minute. This demonstrates the direct economic benefits of model scaling.
Key Quote: "For every 10x increase in model training compute, translator’s managed to complete a task 12.3% quicker […] and a 16.1% increase in earnings per minute." This quote encapsulates the significant improvements in productivity and economic outcomes.
3. Disparate Impact on Low-Skilled Workers
Key Point: Lower-skilled workers benefit more from increased model compute, with a 4x larger improvement in task completion speed compared to high-skilled workers.
Supporting Explanation: The study found that while high-skilled workers reduced their time taken by 4.9%, low-skilled workers saw a 21.1% reduction. This has implications for wage inequalities and the potential for LLMs to bridge skill gaps.
Key Quote: "Those with ‘high-skill’ reduced their time taken per task by 4.9% whilst those with ‘low-skill’ saw a more than 4x larger reduction of 21.1%." This quote underscores the significant disparity in benefits between high-skilled and low-skilled workers.
4. Implications for Future AI Improvements
Key Point: The results suggest that future frontier model scaling, which is currently estimated at a 4x increase per year, may have significant economic implications.
Supporting Explanation: The study implies that as model training compute continues to increase, the economic benefits will become more pronounced. This has implications for productivity enhancements and wage inequalities.
Key Quote: "These results imply further frontier model scaling—which is currently estimated at 4x increase per year—may have significant economic implications." This quote highlights the potential future impact of model scaling on the economy.
Conclusion
The paper concludes by discussing the implications of the findings for various domains, including labor economics, machine translation, and skill-biased wage premia. It acknowledges limitations and the need for further research to generalize the findings to other domains and larger model training compute sizes.
Why It Matters
The study provides valuable insights into the economic impacts of AI model scaling, particularly in the context of professional translation. It highlights the potential for LLMs to enhance productivity and reduce wage inequalities by benefiting lower-skilled workers more significantly. The findings have broader implications for the future of AI in various industries and the need for continued research to understand the full scope of these impacts.
What a 160-year-old theory about coal predicts about our self-driving future - The Verge
In his article "What a 160-year-old theory about coal predicts about our self-driving future," David Zipper explores the potential implications of the Jevons paradox on the adoption of self-driving cars. The Jevons paradox, a 19th-century theory by economist William Stanley Jevons, posits that increased efficiency in resource use can lead to increased total consumption. Zipper applies this theory to the modern context of autonomous vehicles, arguing that even if self-driving cars become more efficient and safer, their widespread adoption could lead to increased total emissions and crash deaths because people will use them more frequently.
The article begins by discussing the current state of self-driving cars, highlighting how companies like Cruise, Waymo, and Tesla are expanding their services despite public wariness. Zipper then introduces the Jevons paradox, explaining its historical context. In the 1800s, improvements in steam engine efficiency led to increased coal consumption because the enhanced efficiency made coal-powered activities more economically viable, leading to a surge in demand. This theory has since been applied to various modern efficiency improvements, such as electric lights, where increased efficiency has led to no decline in total energy consumed due to increased usage.
Zipper extends this theory to the realm of transportation, particularly highway expansions and autonomous vehicles. He argues that if highway expansions or new traffic technologies relieve congestion, more people will decide to drive, leading to increased emissions and crash deaths. This phenomenon, known as induced demand, negates the intended benefits of these improvements. Similarly, the availability of autonomous vehicles could lead people to take longer motor vehicle trips or opt for a car when they would have otherwise used transit, biked, or stayed home. The result would be a lot more (now autonomous) cars on the road, potentially counteracting the theoretical safety and efficiency benefits of self-driving cars.
The article also discusses the potential second-order effects of autonomous vehicles on land use. Just as the advent of car ownership fueled suburbanization in the 20th century, AVs could lead people to relocate to larger, less energy-efficient homes on the urban fringe, where car trips are longer. This could exacerbate the negative environmental impacts of increased driving.
Zipper concludes by emphasizing the need to consider the Jevons paradox when assessing the impacts of autonomous vehicles. While the safety and efficiency of self-driving cars may improve over time, the increased total driving induced by their availability could negate these benefits. The article underscores the importance of a holistic approach to evaluating the benefits and costs of new technologies, particularly in the context of transportation policy, environmental economics, and urban planning.
Key Takeaways
1. The Jevons Paradox and Its Relevance to Self-Driving Cars
Key Point: The Jevons paradox, a classic 19th-century theory, predicts that increased efficiency in resource use can lead to increased total consumption, which has implications for self-driving cars.
Supporting Explanation: The Jevons paradox explains that even if autonomous vehicles become more efficient and safer, their widespread adoption could lead to increased total emissions and crash deaths because people will use them more frequently.
Key Quote: "As a classic 19th-century theory known as a Jevons paradox explains, even if autonomous vehicles eventually work perfectly — an enormous 'if' — they are likely to increase total emissions and crash deaths, simply because people will use them so much." This quote underscores the potential unintended consequences of widespread AV adoption.
2. Historical Context of the Jevons Paradox
Key Point: The Jevons paradox originated from the observation that improvements in steam engine efficiency led to increased coal consumption in the 19th century.
Supporting Explanation: William Stanley Jevons argued that despite improvements in steam engine design, total coal use continued to rise because the efficiency gains made coal-powered activities more economically viable, leading to increased demand.
Key Quote: "Jevons drew from then-recent history to show that steam engines’ efficiency had led people to deploy more of them." This quote highlights the historical basis of the Jevons paradox and its relevance to modern efficiency improvements.
3. Application of the Jevons Paradox to Modern Efficiency Improvements
Key Point: The Jevons paradox applies to modern efficiency improvements, such as electric lights, where increased efficiency has led to increased consumption.
Supporting Explanation: People have responded to improved light bulb efficiency by installing more lights, resulting in no decline in total energy consumed by lighting. This demonstrates how efficiency improvements can backfire and cause increased resource consumption.
Key Quote: "Electric lights are often cited as an example: people have responded to improved light bulb efficiency by installing so many more of them that there has been no decline in the total energy consumed by lighting." This quote illustrates the modern application of the Jevons paradox.
4. Implications for Highway Expansion and Autonomous Vehicles
Key Point: The Jevons paradox suggests that highway expansions and the introduction of autonomous vehicles could lead to increased driving, negating the intended benefits.
Supporting Explanation: If an added lane or new traffic technology relieves congestion, more people will decide to drive, leading to increased emissions and crash deaths. Similarly, the availability of autonomous vehicles could induce more driving, counteracting the theoretical safety and efficiency benefits.
Key Quote: "The Jevons paradox reveals a blind spot in such claims. If an added lane or new traffic technology does relieve congestion, more people will decide to drive due to a drop in the 'cost' of using a car — in this case, the time sitting in traffic." This quote highlights the potential unintended consequences of highway expansions and AV adoption.
5. The Potential Impact of Autonomous Vehicles on Driving Behavior
Key Point: The availability of autonomous vehicles could lead people to take longer motor vehicle trips or opt for a car when they would have otherwise used transit, biked, or stayed home.
Supporting Explanation: As AV companies' ads show, the raison d’être of autonomous vehicles is making driving easier and more pleasant, which could lead people to spend more total time in vehicles and use them for even more tasks.
Key Quote: "Similar to highway expansion, the availability of autonomous vehicles will likely lead people to take longer motor vehicle trips or opt for a car when they would have otherwise used transit, biked, or stayed home." This quote underscores the potential behavioral changes induced by AVs.
6. Second-Order Effects of Autonomous Vehicles on Land Use
Key Point: Autonomous vehicles could lead to increased suburbanization and the adoption of less energy-efficient homes on the urban fringe, where car trips are longer.
Supporting Explanation: Just as the advent of car ownership fueled suburbanization in the 20th century, AVs could lead people to relocate to larger, less energy-efficient homes on the urban fringe, exacerbating the negative environmental impacts.
Key Quote: "Just as the ascent of car ownership fueled suburbanization in the 20th century, AVs could lead people to relocate to larger, less energy-efficient homes on the urban fringe, where car trips — now more tolerable — are longer." This quote highlights the potential second-order effects of AV adoption on land use.
Why It Matters
The article provides valuable insights into the potential unintended consequences of widespread autonomous vehicle adoption. It highlights the importance of considering the Jevons paradox when assessing the impacts of efficiency improvements and technological advancements. The findings have broader implications for transportation policy, environmental economics, and urban planning, emphasizing the need for a holistic approach to evaluating the benefits and costs of new technologies.
What is a correlation, and how do you think clearly about it? (clearerthinking.org)
The article "What is a correlation, and how do you think clearly about it?" by Travis M. explores the concept of correlation, its relationship to causation, and common pitfalls people encounter when interpreting correlations. The article aims to help readers understand these concepts better to avoid being misled or making incorrect inferences.Key Takeaways
Definition of Correlation
Explanation: A correlation is a numerical measure between -1 and 1 that indicates the strength and direction of a relationship between two phenomena.
Key Quote: "A correlation is a number between -1 and 1 that expresses the strength and direction of the relationship between two phenomena."
Why It Matters: Understanding what a correlation is helps in quantifying relationships and making sense of data.
Correlation vs. Causation
Explanation: Just because two things are correlated does not mean one causes the other.
Key Quote: "Just because two things are correlated doesn’t mean that one causes the other."
Why It Matters: This distinction is crucial to avoid mistakenly inferring causation where none exists, which can lead to flawed decisions and beliefs.
Confounding Variables
Explanation: Confounding variables are hidden influences that affect both variables being correlated, making it seem like one affects the other.
Key Quote: "Confounding variables are hidden influences that affect both of the two variables of interest in a correlation, making it seem like one affects the other when, in reality, it’s really the confounding variable that is affecting both."
Why It Matters: Recognizing confounding variables helps in accurately interpreting correlations and avoiding misleading conclusions.
Strength of Correlations
Explanation: The strength of a correlation is indicated by how close it is to -1 or +1. Weak correlations can be overstated, leading to misinterpretations.
Key Quote: "Frustratingly, a lot of science journalism simply reports that there is a correlation between phenomena, without telling us the strength of that correlation."
Why It Matters: Understanding the strength of correlations helps in making informed decisions and avoiding overinterpretation of weak relationships.
Generalizability of Correlations
Explanation: Correlations found in one context may not apply universally. They can vary across different groups, environments, or situations.
Key Quote: "Sometimes people assume that a correlation found in one context applies everywhere, but correlations can vary across different groups, environments, or situations."
Why It Matters: Recognizing the limitations of correlations helps in avoiding erroneous generalizations and making more accurate predictions.
Explanation Supporting Key Takeaways
Definition of Correlation
Correlation measures how much two variables change together. A positive correlation (+1) means both variables increase together, a negative correlation (-1) means one variable decreases as the other increases, and a correlation of 0 means there is no linear relationship.
Example: The relationship between coffee intake and productivity at work.
Correlation vs. Causation
Correlation does not imply causation. Two variables may be correlated due to coincidence or a third factor influencing both.
Example: The correlation between Elijah Wood’s movie appearances and the number of orderlies in Oklahoma does not imply causation.
Confounding Variables
Confounding variables can lead to spurious correlations. For example, the correlation between ice cream sales and sunburn rates is due to sunny weather, not a direct causal link between ice cream and sunburn.
Example: The French paradox regarding red wine and heart disease, which was influenced by various confounding variables.
Strength of Correlations
The magnitude of correlation values should be considered. Weak correlations (close to 0) may not be significant, while stronger correlations (close to -1 or +1) indicate a more substantial relationship.
Example: Correlations in different fields may be interpreted differently, such as a correlation of 0.7 being considered strong in psychology but moderate in medicine.
Generalizability of Correlations
Correlations may not apply universally. They can differ based on context, such as different populations or situations.
Example: Heart attack symptoms in women differ from those in men, and behavioral correlates of autism are different in men and women.
By understanding these key points and the supporting explanations, readers can think more critically about correlations and avoid common pitfalls in interpreting data. This critical thinking skill is essential for making informed decisions and avoiding being misled by superficial relationships.
Eleven Predictions: Here's What AI Does Next - by Ted Gioia (honest-broker.com)
In the article "Eleven Predictions: Here's What AI Does Next," Ted Gioia presents a series of predictions about the future of artificial intelligence (AI). He highlights the rapid and sometimes chaotic advancements in AI technology and discusses the potential impacts on society. Gioia warns of the unintended consequences of AI, including its potential to disrupt industries, create existential crises, and reshape the nature of work and human interaction.
Key Takeaways
Ownership of AI as the Real Threat
Explanation: Gioia argues that the real threat from AI comes not from the technology itself but from the people who own and control it.
Key Quote: "Stop worrying about AI taking over. It’s the people who own the AI who pose the biggest threat."
Why It Matters: This perspective shifts the focus from technological fears to the ethical and power dynamics among those who control AI.
Multiple AI Agents and Competition
Explanation: AI will not be a single force but will involve competing agendas from governments, corporations, hackers, and others, each with their own AI agents.
Key Quote: "Hundreds of governments and corporations are already competing in building their AI empires."
Why It Matters: Understanding the complex landscape of AI competition helps in anticipating conflicts and strategic alliances.
AI War Zone
Explanation: The future will see AI agents constantly battling each other for control, leading to potential conflicts and disruptions.
Key Quote: "We will soon be living in an AI war zone—because competing AI agents will constantly battle each other for control (often over us)."
Why It Matters: This prediction highlights the need for robust regulatory and ethical frameworks to manage AI interactions.
Impact on Creative Professionals
Explanation: Companies will view creative professionals as liabilities and may launch stealth attacks on their own core talent base.
Key Quote: "Companies will launch stealth attacks on their own core talent base."
Why It Matters: This trend could lead to significant job displacement and a restructuring of creative industries.
Dumbness Crisis in AI
Explanation: Despite massive investments, AI will face a decline in quality due to exhausted high-quality training materials, leading to a "dumbness crisis."
Key Quote: "The bots will actually get dumber, despite trillions of dollars in investment."
Why It Matters: This prediction underscores the challenges in maintaining and improving AI performance over time.
Disruption Over Profit
Explanation: AI will create more disruption than profit, with companies struggling to implement it successfully and generate returns.
Key Quote: "AI will create more disruption than profit."
Why It Matters: This highlights the economic risks and uncertainties associated with AI investments.
Job Replacement and Vulnerability of Tech Workers
Explanation: AI will replace jobs, with digital workers, particularly software developers, being the first casualties.
Key Quote: "The first casualties will be digital workers, especially software developers."
Why It Matters: Understanding the job market's evolution is crucial for career planning and policy-making.
Value of Human Skills
Explanation: Holistic thinkers and empathetic individuals with human skills will be the actual winners in the AI age.
Key Quote: "The actual winners will be holistic thinkers and empathetic individuals with human skills."
Why It Matters: This prediction emphasizes the enduring value of human qualities that AI cannot replicate.
Proliferation of Malicious AI
Explanation: Malicious and deceptive AI will proliferate at warp speed, outnumbering caring human agents.
Key Quote: "The actual winners will be vastly outnumbered by malicious and deceptive AI."
Why It Matters: This highlights the need for vigilance and security measures to combat AI-driven threats.
Preference for Human Interaction
Explanation: Even when AI can handle a job, people will pay more for human interaction, which will become a status symbol.
Key Quote: "Even when AI can adequately handle a job, people will pay more for a human being."
Why It Matters: This prediction underscores the continued value of human interaction in a digital world.
Existential Crisis
Explanation: AI will create an existential crisis, blurring the lines between reality and deception, truth and lies, fact and error.
Key Quote: "AI will create an existential crisis of epic proportions, as all the dividing lines that are foundational to society—between reality and deception, truth and lies, fact and error—collapse."
Why It Matters: This prediction warns of the profound societal and philosophical challenges posed by AI.
Explanation Supporting Key Takeaways
Ownership of AI as the Real Threat
Supporting Explanation: Gioia argues that the true danger of AI lies in the intentions and actions of those who control it. Psychology and game theory will play a significant role in how AI is used and misused.
Example: The example of companies like Adobe attacking their own creative talent highlights the human-driven threats posed by AI.
Multiple AI Agents and Competition
Supporting Explanation: The AI landscape will be characterized by multiple competing agendas, leading to a complex and potentially conflict-ridden environment.
Example: The competition between the US and China in AI is just one example of the broader dynamic of competing AI empires.
AI War Zone
Supporting Explanation: AI agents will engage in constant battles for control, leading to disruptions and potential conflicts that could affect human lives.
Example: The Minecraft simulation where AI agents created a religious cult and resolved conflicts with nuclear weapons illustrates the potential for AI conflicts.
Impact on Creative Professionals
Supporting Explanation: Creative professionals will be seen as liabilities in the new AI economy, leading to job displacement and strategic attacks on talent bases.
Example: Adobe's attack on photographers and artists showcases how companies may view creative professionals as expendable.
Dumbness Crisis in AI
Supporting Explanation: The exhaustion of high-quality training materials will lead to a decline in AI performance, despite massive investments.
Example: OpenAI admitting that its latest chatbot will have performance issues supports the prediction of a dumbness crisis.
Disruption Over Profit
Supporting Explanation: AI investments have often failed to generate profits and have led to significant disruptions, with companies struggling to justify their spending.
Example: The recent sell-off in AI investment funds and the challenges companies face in implementing AI highlight the economic risks.
Job Replacement and Vulnerability of Tech Workers
Supporting Explanation: AI will first replace digital workers, particularly those in software development, as it is most reliable in writing software code.
Example: The reliability of AI in software development makes tech workers the most vulnerable to job displacement.
Value of Human Skills
Supporting Explanation: Human qualities such as love, care, trust, and compassion cannot be replicated by AI, making holistic thinkers and empathetic individuals valuable.
Example: The preference for human interaction in areas where trade-offs must be made highlights the enduring value of human skills.
Proliferation of Malicious AI
Supporting Explanation: Malicious and deceptive AI will proliferate rapidly, leading to a proliferation of fake content and increasing the risk of scams and deception.
Example: The proliferation of fake images, articles, and videos illustrates the growing problem of AI pollution.
Preference for Human Interaction
Supporting Explanation: People will pay more for human interaction as a status symbol, valuing authentic human qualities over AI-driven interactions.
Example: The preference for laboratory-made diamonds over authentic ones illustrates the enduring value of reality over artificial substitutes.
Existential Crisis
Supporting Explanation: AI will create profound societal and philosophical challenges, blurring the lines between reality and deception, truth and lies, fact and error.
Example: The need for personhood credentials and safe words to prove one's existence highlights the existential crisis posed by AI.
By understanding these predictions and their implications, readers can better prepare for the future of AI and the challenges it presents.
Pluralistic: Unpersoned (22 Jul 2024) from Cory Doctorow
In the article "Pluralistic: Unpersoned (22 Jul 2024)" by Cory Doctorow, the author discusses the implications of putting critical infrastructure into the hands of unaccountable tech giants. The article highlights the risks associated with cloud-based services and the unilateral power that tech companies have over their users. Doctorow uses real-life examples to illustrate how individuals can be "unpersoned" by these platforms, leading to significant disruptions in their lives. He also explores potential solutions, such as the EU's Digital Markets Act, to address these issues.
Key Takeaways
Unpersoning by Tech Giants
Explanation: Tech giants like Google have the power to lock users out of their accounts without explanation, leading to significant disruptions in their lives.
Key Quote: "Last March, Renee's Google account was locked, and she was no longer able to access ten manuscripts for her unfinished books, totaling over 220,000 words."
Why It Matters: This highlights the unchecked power tech companies have over users' data and livelihoods.
Lack of Transparency and Accountability
Explanation: Tech companies often provide opaque customer service and do not offer clear explanations for account suspensions.
Key Quote: "Google's famously opaque customer service – a mix of indifferently monitored forums, AI chatbots, and buck-passing subcontractors – would not explain to her what rule she had violated, merely that her work had been deemed 'inappropriate.'"
Why It Matters: The lack of transparency and accountability makes it difficult for users to understand or challenge these decisions.
Impact on Personal Data and Livelihood
Explanation: When accounts are suspended, users can lose access to critical data, leading to significant personal and professional consequences.
Key Quote: "Without contacting Mark, Google sent a copy of all of his data – searches, emails, photos, cloud files, location history and more – to the SFPD, and then terminated his account."
Why It Matters: This underscores the importance of users having control over their data and the need for robust protections against arbitrary suspensions.
Potential Solutions
Explanation: Doctorow discusses potential solutions, such as the EU's Digital Markets Act, which aims to reduce switching costs and increase competition.
Key Quote: "The EU has a new solution to this problem. With its 2024 Digital Markets Act, the EU is requiring platforms to furnish APIs – programmatic ways for rivals to connect to their services."
Why It Matters: These solutions could help mitigate the power of tech giants and provide users with more control over their data.
The Cloud as a Trap
Explanation: Cloud-based services can trap users, making it difficult for them to switch to competitors or access their data.
Key Quote: "The cloud is many things, but most of all, it's a trap. When software is delivered as a service, when your data and the programs you use to read and write it live on computers that you don't control, your switching costs skyrocket."
Why It Matters: This highlights the need for regulations that reduce switching costs and increase competition in the tech industry.
Explanation Supporting Key Takeaways
Unpersoning by Tech Giants
Supporting Explanation: Tech giants like Google have the power to lock users out of their accounts without explanation, leading to significant disruptions in their lives. Users can lose access to critical data and services, which can have severe personal and professional consequences.
Example: The case of K Renee, a romance writer whose Google account was locked, leading to the loss of over 220,000 words of manuscripts.
Lack of Transparency and Accountability
Supporting Explanation: Tech companies often provide opaque customer service and do not offer clear explanations for account suspensions. This makes it difficult for users to understand or challenge these decisions, and they are often left without recourse.
Example: When K Renee's account was locked, Google's customer service did not explain what rule she had violated, merely stating that her work had been deemed "inappropriate."
Impact on Personal Data and Livelihood
Supporting Explanation: When accounts are suspended, users can lose access to critical data, leading to significant personal and professional consequences. This can include the loss of important documents, photos, and other personal information.
Example: Mark, a tech worker whose Google account was terminated after a photo of his son's infected penis was flagged as CSAM. He lost access to all his data, including his phone number, email archives, and stored passwords.
Potential Solutions
Supporting Explanation: Doctorow discusses potential solutions, such as the EU's Digital Markets Act, which aims to reduce switching costs and increase competition. This could help mitigate the power of tech giants and provide users with more control over their data.
Example: The EU's Digital Markets Act requires platforms to furnish APIs, allowing rivals to connect to their services. This could help reduce the pain of switching to a rival and increase competition.
The Cloud as a Trap
Supporting Explanation: Cloud-based services can trap users, making it difficult for them to switch to competitors or access their data. This gives tech companies significant power over their users and makes it difficult for them to leave.
Example: Adobe's decision to cancel its Pantone color-matching license, which led to all colors in users' files turning black until they paid an upcharge. This highlights how cloud-based services can be used to enshittify users.
By understanding these key points and their implications, readers can better appreciate the risks associated with putting critical infrastructure into the hands of unaccountable tech giants. This awareness can help inform discussions about the need for robust regulations and protections for users in the digital age.
