• Pascal's Chatbot Q&As
  • Posts
  • The article "How to Say No to Our AI Overlords" discusses the increasing prevalence of AI technologies from major companies like Google, Microsoft, Meta, and Apple in everyday consumer products.

The article "How to Say No to Our AI Overlords" discusses the increasing prevalence of AI technologies from major companies like Google, Microsoft, Meta, and Apple in everyday consumer products.

Even when users opt out of direct data collection, AI companies can still potentially access user or usage-related data through various indirect and creative means.

Question 1 of 3 for ChatGPT-4o: Please read the article "How to say no to our AI overlords" and tell me what it says

ChatGPT-4o’s answer:
The article "How to Say No to Our AI Overlords" discusses the increasing prevalence of AI technologies from major companies like Google, Microsoft, Meta, and Apple in everyday consumer products. The author, Brian X. Chen, points out that while these tech giants claim AI is becoming indispensable for tasks such as writing emails, generating images, and summarizing information, many people outside the tech industry are not enthusiastic about these developments. Instead, they find AI tools intrusive and sometimes irritating, especially when they don’t function as intended.

A significant issue highlighted is how these companies use consumer data to train and improve AI systems, raising concerns about privacy. The article explains how users can opt out, where possible, from contributing their data to these AI systems. For example:

  • Google: Users can filter out AI-generated summaries in search results by clicking on the "Web" tab and can stop Google from tracking their activity via myactivity.google.com.

  • Meta: Although there is no way to completely turn off Meta's AI, users in regions with strong data protection laws (such as the EU and UK) have options to deny access to personal data.

  • Microsoft: Users can disable the AI Copilot in the Edge browser and LinkedIn settings, preventing the platform from using their content for AI training.

  • Apple: Users will soon be able to opt into or out of Apple Intelligence, the company’s new AI system, and Apple claims it uses a privacy-first approach where data is inaccessible to the company itself.

The piece emphasizes the tension between user privacy and the growing influence of AI, offering practical steps for users to mitigate the impact of these technologies on their daily digital activities​.


Question 2 of 3 for ChatGPT-4o: Are there any other instruments users or AI makers should consider, other than the ones mentioned?

ChatGPT-4o’s answer:
Yes, beyond the specific tools and settings discussed in the article, there are several other instruments and strategies that users and AI makers can consider to further protect privacy, manage AI interactions, and promote responsible AI development. Here are some additional approaches:

For Users:

  1. Ad Blocking and Privacy-Focused Browsers:

    • Privacy-focused browsers like Brave, Tor, or Firefox (with privacy-enhancing plugins) help block AI-related trackers and cookies.

    • Ad blockers like uBlock Origin can also block AI-based ads and tracking scripts that are used to collect data for training AI systems.

  2. Decentralized Data Management Tools:

    • Tools like Solid (a project by Tim Berners-Lee) allow users to take control of their personal data and decide how it's shared with AI systems or other entities.

  3. AI Transparency Tools:

    • Algorithm transparency platforms or browser extensions can help users understand how AI is affecting the content they see, such as AI-generated news or search results. Examples include tools like Ad Observer, which tracks and explains targeted ads on platforms like Facebook.

  4. Data Minimization Practices:

    • Regularly delete unnecessary data (search histories, messages, etc.) that could be used to train AI algorithms. Consider using services that offer auto-delete or encrypted storage options.

For AI Makers:

  1. Federated Learning:

    • Implement federated learning, where AI models are trained locally on devices (rather than centralized servers), so that personal data never leaves the user’s device, enhancing privacy while still improving AI performance.

  2. Differential Privacy:

    • AI makers can incorporate differential privacy techniques to ensure individual data points cannot be reverse-engineered or traced back to users. Apple and Google already use differential privacy in some of their products, but expanding this practice across more AI systems could offer better user protection.

  3. Explainability Tools (XAI):

    • Implement tools for Explainable AI (XAI), where AI systems offer transparency by explaining how they made decisions, which increases user trust and awareness. Users can then decide whether they want to interact with these AI tools.

  4. Opt-in AI Systems:

    • Instead of making AI tools default, AI makers could adopt an opt-in policy for AI usage, giving users the power to explicitly choose whether or not to engage with AI features. This ensures that AI isn't being pushed onto users by default.

  5. Data Portability and User Rights:

    • Companies should enable robust data portability options where users can easily transfer or delete their data across platforms. Users should have clear options to audit and control how their data is being used by AI systems. This aligns with principles from the General Data Protection Regulation (GDPR).

  6. Ethical Review and Audits:

    • AI makers should conduct regular ethical reviews and third-party audits to assess the societal impact of their AI tools. This helps ensure AI technologies aren’t being used in ways that infringe on user rights or manipulate users unfairly.

By combining these tools and approaches, both users and AI makers can create a healthier ecosystem that respects user privacy while enabling responsible AI innovation.


Question 3 of 3 for ChatGPT-4o: List all creative ways in which the AI companies can still access user / usage related data, even when the users have opted out

ChatGPT-4o’s answer:
Even when users opt out of direct data collection, AI companies can still potentially access user or usage-related data through various indirect and creative means. However, it's important to note that many of these methods may raise significant ethical concerns, and in some cases, could violate privacy laws like the General Data Protection Regulation (GDPR) in Europe or the California Consumer Privacy Act (CCPA). Here’s a list of creative methods AI companies might use:

1. Aggregated and Anonymized Data Collection:

  • Service Metadata: Even if users opt out of personal data collection, companies may still collect metadata (e.g., timestamps, device type, general location, and network data) about how their services are being used. These are often considered less sensitive but can reveal usage patterns.

  • Device or Application Logs: Companies might collect logs from services or apps, which can include usage data in an aggregated, anonymized form. Over time, even anonymized data could be re-identified with other datasets.

  • Cross-Service Data Integration: AI companies that own multiple services or partner with third-party services can pool data from different sources. Even if a user opts out on one platform, data from another service could still be used to infer behavior or preferences.

  • Predictive Modeling: AI companies can develop models that predict user behavior or preferences based on data from users who haven’t opted out. These inferences can be generalized to the entire user base, including those who have opted out.

3. Behavioral Tracking via Cookies and Similar Technologies:

  • Browser Fingerprinting: AI companies can use browser fingerprinting techniques (device configuration, fonts, screen resolution, etc.) to track users without relying on cookies or other traditional identifiers.

  • Zombie Cookies: These are cookies that are reinstalled even after users delete them, often through browser exploits or using Flash cookies, HTML5 storage, or other mechanisms. This technique has been used historically, though it’s increasingly scrutinized.

  • Pixel Tracking: Invisible tracking pixels embedded in websites or emails can capture user interactions without directly collecting personal data, such as whether a user has opened an email or visited a webpage.

4. Data from IoT and Connected Devices:

  • Smart Devices: AI companies may collect usage data from Internet of Things (IoT) devices (e.g., smart thermostats, appliances) that users may not realize are collecting data. Even if direct consent has not been given, these devices may still provide valuable insights into user behavior.

  • Wearables and Connected Vehicles: Data from wearables, smartwatches, or connected cars could be used to track movement patterns, physical activity, or even health-related metrics, offering insights into user habits.

5. Proxy Data from Non-Personal Sources:

  • Network Analysis: Companies can analyze general network activity to understand patterns at a population level. Even if individual users have opted out, understanding broader network usage trends can still provide valuable insights.

  • Third-Party Data Brokers: Data purchased from third-party brokers (who may have obtained the data from other apps or services the user uses) can help fill in gaps, even if the user has opted out from the primary service.

6. Shadow Profiles and Ghost Data:

  • Shadow Profiles: Companies can build shadow profiles on users based on their interactions with other users who haven’t opted out. For example, if User A opts out but interacts with User B, who hasn’t opted out, data about their interactions may still be collected and used.

  • Ghost Data: If a user has previously provided data to the company before opting out, this historical data may still be used to infer future behavior, even though new data is no longer being collected.

7. AI Model Feedback and Learning Loops:

  • AI Model Training on Aggregate Data: AI models can be trained on data from users who haven’t opted out, but the resulting model can still apply to users who have opted out. The system doesn’t need personal data to predict preferences or make recommendations for all users based on similar behavior patterns.

  • Feedback Loops: Even without directly collecting personal data, companies can use user interactions to indirectly optimize AI models. For example, the AI might track which recommendations or actions users respond to without recording personal details.

8. Data from Public Sources or Open Platforms:

  • Public Data Mining: Companies can scrape data from public social media posts, websites, or other open data sources where users may inadvertently reveal behavior, preferences, or other usage patterns, even if they’ve opted out of direct collection.

  • Crowdsourced Data: Some platforms use community or crowdsourced inputs (e.g., reviews, ratings, feedback) that, while not directly linked to a specific user, provide usage data that can inform AI systems.

  • Third-Party API Data Sharing: If users interact with third-party services integrated into the platform (e.g., using a third-party app through a social login), data from those third-party services could be shared back with the AI company, depending on the integration agreements.

  • Indirect Consent Through Bundled Agreements: Some services bundle consent for data sharing with other services or apps. Users might unknowingly give consent to data collection by agreeing to terms for seemingly unrelated services.

10. Using Environmental or Contextual Data:

  • Location-Based Inferences: AI companies can collect general geolocation data from cell towers or IP addresses to infer behavior based on where users are located, even if they haven’t given explicit permission for direct location tracking.

  • Contextual Data from App Interactions: Even if specific data collection is disabled, companies can gather insights from how users interact with the app (e.g., time spent on certain pages, button clicks, etc.), which can provide enough data to model usage trends.

11. In-App Data Collection from Other Users:

  • Social Graph Data: In social networks or collaborative apps, AI companies can collect data from other users that might indirectly reveal information about those who have opted out. For example, messages, comments, or tags from others may still include personal details or interactions.

12. Data from Offline Purchases and Loyalty Programs:

  • Loyalty Programs and Offline Tracking: Even if a user opts out of data collection online, companies can collect data from offline purchases, loyalty programs, or third-party retail partners. Linking this data with payment methods, loyalty cards, or offline behavior can still give companies insights into user behavior.

While these techniques may allow companies to access data even when users opt out, many are legally questionable and may lead to regulatory action, such as fines under GDPR or CCPA. AI companies must balance data-driven innovation with respect for user privacy and legal compliance to avoid reputational damage and penalties.