AI promises are everywhere. In breathless (and usually uncritical) coverage of AI on 60 Minutes, in the New York Times, and innumerable podcasts, we are told that AI will have a “profound” effect on all aspects of human life–from education, to commerce, to healthcare, to transportation, and more. At the same time, others predict AI will create a dystopian future where AI is used to surveil and manipulate the population with devastating results.

Given these extreme views on AI and its impacts, many of us are left with mixed feelings about AI. We have questions. We wonder: What is AI actually doing/delivering? Who benefits and who is left behind by AI? When do the harms outweigh the benefits and (how) can we prevent those harms? Are there better uses for our finite resources? What does AI mean for the things we care most about? What choices do we have as individuals?

How do we know if claims being made (“AI will be…more profound than fire or electricity” or “AI will most likely cause WW3”) carry any truth? If so–or even, if not–what does that mean for us?

These questions are complicated by how widely-used, but ill-defined, the term AI has become and by how quickly AI technologies are evolving and being applied in new ways. The term includes an expansive set of technologies. For example, generative AI is a subset of AI technology that has gotten a lot of recent attention because it mimics human writing, speech, reasoning, and artistic creation. It’s flashy and novel. It’s easy and sometimes fun to use. But generative AI is just a small piece of the AI picture. Like an iceberg, there is a mass of AI technologies that are less prominent (in this moment) and that have been in use at larger scales and for, arguably, more consequential purposes (insurance, health care, law enforcement, weather models, etc.).

To begin to make sense of AI and its impacts, we need to first understand what we’re talking about when we/they say “AI.” We can develop a deeper understanding of the various types and functions of AI, and their impacts. We can also develop frameworks for wading through the intense polarization brought on by these technologies so that we can make intentional choices about our use of them. These are the goals of this year’s Digital Detox–to help all of us develop lenses for evaluating and preparing for/responding to AI uses.

Predictive AI – Predictive AI supports or supplants human decision making. It aims to identify patterns and predict outcomes based on historical or simulated data. This AI is seen as a way to find answers in data that is too large and complex for humans to make sense of.  More data is seen as the path to better predictions which drives increasingly aggressive tracking and logging by companies. And, this AI brings up concerns about our individual rights to privacy, the quality of the data (objective, accurate, complete, representative), and whether we’re even using the right data/measurements for the outcomes being predicted. AI can be a form of tech-washing, where the technology creates a veneer of objectivity and makes already-complex decision processes even more opaque. AI can end up automating and compounding inequity. Clear examples of these issues have already been seen as AI has been used in health insurance and law enforcement. The same AI can, on one hand, alleviate gender bias in sentencing and, on the other hand, increase racial bias. Understanding, evaluating, and regulating the relationship between potentially biased data, potentially biased algorithms, and potentially biased application is going to be an ongoing challenge requiring significant effort and attention as AI spreads to more and more high-impact sectors.

Generative AI – Generative AI supports or supplants human communication and media creation, and it’s increasingly being used for analysis and reasoning. While text was the initial medium, images, audio, and video creation have expanded rapidly in quality and sophistication. This technology is built on huge troves of data that have been collected without permission from, or payment to, the people who created the content. There are a number of class action lawsuits currently underway alleging copyright infringement. Generative AI also brings up questions of accuracy, energy usage, and job security in a number of industries.

AI Agents/Agentic AI – Agentic AI supports or supplants human interactions in more complex processes. It can act independently to make decisions and often interacts with other tools to accomplish specific goals. Agents are often built using a mix of AI technologies and can include both predictive and generative AI. Claude’s demo video showing how their “computer use” feature can coordinate a sunrise view of the Golden Gate bridge is one example.

There are, of course, other ways of categorizing and understanding AI–and other types of AI not represented here. These categories begin to give us a framework by which we can evaluate applications of AI in our lives and in the world. We’ll come back to these categories throughout this experience to help us evaluate various applications of AI.

Dealing with polarization

AI is increasingly viewed through opposing extremes – either as a force for catastrophic harm or as the solution to humanity’s greatest challenges. AI is going to either cause the extinction of the human race or usher in a new era of peace and abundance for everyone. AI will exacerbate the climate crisis or solve it entirely. Few technologies have become symbols of both hope and fear simultaneously.

Narayanan and Kapoor highlight the dangers of overemphasizing AI’s capabilities for good or for harm. For example, in portraying AI as an all-powerful, existential threat, critics unintentionally bolster the claims of companies promising far more than AI can currently, or possibly ever, deliver (p 178). Excessive hype undermines our ability to scrutinize AI and simultaneously devalues the work of experts who are raising specific concerns. Wildly polarized views about AI serve to keep us disempowered, unable to ascertain what’s happening, how we can shape what’s happening, and how we can make informed decisions for ourselves and loved ones.

AI raises important questions about humanity, thought, art, and intellectual property while intensifying concerns about equity, environmental impact, mass manipulation, and data privacy. These are questions that need to be addressed first through what AI is capable of now.

To have space for these important questions and conversations, we should resist retreating to extremes and focus on what’s really happening. Narayanan and Kapoor suggest that (like all research), when claims are made about AI, we should emphasize the need for transparency, reproducibility, and accuracy in representing what AI is doing.

We will need to be willing to grapple with the human problems that are brought to the forefront with AI. For example, the hype surrounding the use of AI in education should cause us to grapple with the real problems in education. 86% of US K12 school systems reported issues hiring enough teachers in the 2023-24 school year and K12 teachers have one of the highest burnout rates of all industries (44%). With wages that continue to decline when compared to the average for other US workers, it’s clear there are foundational problems that are entwined with larger societal and economic issues. These aren’t AI issues but when sectors have long-standing and unaddressed foundational issues, AI is increasingly seen as a “solution.” And, as Dan Meyer suggests, we should be wary of AI solutions for education that continue the pattern of investing significant resources in tech companies, while further divesting resources essential for the success of teachers, schools, children, and their communities.

Dealing with the speed of change

Even when our goal is to be thoughtful and engaged citizens who make intentional choices about AI in our lives, that goal is complicated by the speed of change. New models, new tools, new possibilities and new fears emerge every day, it seems. Some changes are just minor updates, or UI tweaks, but some feel really significant. At the time of writing, ChatGPT has just finished 12 days of AI with a number of fairly major updates just in the month of December and with more promised soon. Claude, Copilot, Gemini continue to implement new features and model upgrades. We don’t know what the next month (let alone 4 years) will bring but under the Trump administration we can expect that the few regulations governing AI in the USA will be relaxed or removed. Competition with China (Qwen VL and many others), competition between the increasing number of AI companies, and massive investment (approaching $200 billion in 2025) will likely drive rapid expansion of models and capabilities.

While we can’t control the pace of change or the use of fear-of-missing-out (FOMO) marketing strategies, we can use some strategies from Patrick McGinnis (coiner of the term FOMO) to help us prioritize our attention. We’ll be using this pattern throughout this experience.

• Alignment
  • Clarify your personal and/or organizational values and goals relative to using AI.
• Prioritization
  • Limit your attention to AI that aligns with your particular values and goals.
  • Decide what information you need to determine this.
• Remove noise
  • To help avoid hype, follow a few credible sources that align with your values and goals. Cultivate a trusted network.
• Focused experimentation
  • Choose a few AI tools or concepts to explore more deeply rather than trying to keep up with everything.
• Reflective decision-making
  • Pause before making decisions about AI. Weigh pros and cons using data to analyze risks and benefits logically and strategically.
• Reframe FOMO as JOMO (Joy of missing out)
  • There is value in making deliberate, well-reasoned choices so you can focus on what truly matters.