Digging in: “Two AI Truths and a Lie” dangerous extractions
Industry will take everything it can in developing Artificial Intelligence (AI) systems.
We will get used to it.
This will be done for our benefit.
“Two of these things are true and one of them is a lie. It is critical that lawmakers identify them correctly…. no matter how AI systems develop, if lawmakers do not address the dynamics of dangerous extraction, harmful normalization, and adversarial self-dealing, then AI systems will likely be used to do more harm than good.” - Woodrow Hartzog
"Two AI Truths and a Lie," dangerous extractions excavated
In my last post about the Woodrow Hartzog piece (freely available and a highly recommended read) I noted that each concept that we focused on deserved a deeper dive. Here is the beginning of that exploration.
Industry exploitation and data collection (dangerous extraction):
AI systems are voracious for personal data, essential for training models
Companies exploit the narrative that human information is a raw resource for economic production, creating a "biopolitical public domain" where personal data is free for exploitation.
Examples include targeted ads based on comprehensive data mining, showing the extensive reach of data collection even without direct eavesdropping.
"Companies cannot create AI without data, and the race to collect information about literally every aspect of our lives is more intense than ever."
AI Systems' Appetite for [Personal] Data:
We’ve explored the deep need for training data here before, and the toll (in terms of energy and water) that AI is increasingly taking on the environment. Just a quick reminder, (and from a new study by the folks over at Goldman Sachs):
On average, a ChatGPT query needs nearly 10 times as much electricity to process as a Google search.
AI systems rely heavily on vast amounts of personal data to train their models effectively. The more data these systems have access to, the better they can learn and make predictions. Their incredible appetite for data leads to widespread data collection from various sources, often without explicit user consent or awareness. (Large Language Models have been developed in relative silence and secrecy for decades).
For example, Facebook's data collection practices have been widely criticized. The platform collects extensive data on users' behaviors, preferences, and interactions to enhance its AI algorithms for targeted advertising and content personalization. In the not-too-distant past’s Cambridge Analytics scandal, Facebook allowed a third-party app to collect data on 50+ million users without their explicit consent. Their data was used to influence political campaigns, showing us exactly how personal information can be exploited for economic and political gain.
And the Temu app asks for more than 24 permissions for all kinds of information - like biometric data (some of the most personal data around) - that would not be needed for an online shopping app.
As the technologists behind AI look further to gather data to personalize experiences for us, we now have a whole new level of protection we need to be aware of for our personal data.
Human Information as a Raw Resource:
Companies frame human information as a raw resource for economic production, effectively creating a "biopolitical public domain." This concept - from Julie Cohen's work on the legal construction of the surveillance economy - is referenced to highlight how personal data is treated as a resource for economic gain, leading to extensive data collection and exploitation practices by companies.
Julie Cohen, in her writings, explores how human information is framed as a resource for economic production, which companies are free to exploit.
As a recent example, Google's data mining practices illustrate this point well. Amazon's Alexa devices collect vast amounts of data from users' homes. While the primary purpose is to improve the AI's performance and user experience, concerns have been raised about the potential for data misuse and the lack of transparency in data handling.
By treating Alexa and search queries, location data, and browsing habits as raw materials, Google refines the information into valuable (to them) insights that power its advertising algorithms and other AI-driven services. Never forget - we ARE the product.
Extensive Reach of Data Collection Without Direct Eavesdropping:
The reach of data collection is vast, often extending beyond direct user interactions to include inferred data based on patterns and behaviors. This comprehensive data mining allows companies to create detailed profiles and target individuals with precision, even without actively listening to their conversations.
Targeted advertising can feel eerily accurate. We might chat about a product or service offline, only to see ads for it shortly after. This phenomenon is typically due to sophisticated data mining and pattern recognition algorithms rather than actual eavesdropping.
As an example, here is an illustration of the 2024 Marketing Technology Landscape of 14,106 martech products (27.8% growth YoY) from ChiefMartech’s Scott Brinker:
As we’ve noted before, studies show that almost 70% of these products have AI built into them already. So the data gathering, collection and mining capabilities continue to expand exponentially.
These examples underscore the critical need for robust regulatory frameworks to ensure AI systems are developed and deployed in ways that prioritize user privacy and societal benefit over unchecked and truly dangerous extractions.
With gratitude for the thoughtful piece, thank you Woodrow Hartzog. Two AI Truths and a Lie (May 24, 2024). 26 Yale Journal of Law and Technology (forthcoming 2024). And we got additional support on this post from our friend ChatGPT.
Resources from AIGG on your AI Journey
Need training or specific support in building AI Literacy or AI Regulations for your organization? We’re a little different. We’re not approaching AI from a tech perspective, though we have techies on staff. We’re approaching it from a safe, ethical, and responsible use perspective because we’ve been through technology and business transformations before.
Whether you’re a government agency, school, district, or business looking to add AI to your tech toolkit, we can guide the way in a responsible manner. AiGg is here to support you in navigating ethics, governance, and strategy setting.
We have attorneys, anthropologists, data scientists, and business leaders to support you as you develop your Strategic AI Use Statements, which can guide your organization’s use of the tools available to you. We also offer bespoke educational workshops to help you explore and build your playbooks, guidelines, and guardrails as your adoption (and potential risk management) options grow.
Connect with us for more information, to get your free AI Tools Adoption Checklist, Legal and Operational Issues List, HR Handbook policy, or to schedule a workshop to learn more about how to make AI work safely for you. We are here for you.
Reach out for more information and to begin the journey towards making AI work safely and advantageously for your organization.
