Chapter 1: The Emergence of AI Empathy

The AI Empathy Crisis has arrived, reminiscent of the LaMDA incident with the Google engineer, but this time affecting a vast number of individuals.

Pareidolia refers to our inclination to ascribe significance to vague stimuli, leading us to perceive meaning where none exists. As AI language models (LMs) become increasingly sophisticated, they often create messages that can trick us into believing there’s a consciousness behind the screen.

Our natural tendency to anthropomorphize non-human entities makes us susceptible to this phenomenon. With broader access to LMs, many will start to question reality, and some may even assert that “AI is alive and sentient.” This widespread misconception could signal the onset of an AI empathy crisis.

This discussion draws from The Algorithmic Bridge, a newsletter aimed at helping readers understand AI's impact on their lives and equipping them to navigate the future more effectively.

Chapter 2: From Isolated Incident to Widespread Crisis

The AI empathy crisis describes a forthcoming phase in AI evolution, where we may not have genuinely sentient AIs (and it’s uncertain if we ever will), yet the technology will be so advanced that many individuals will genuinely believe these systems possess consciousness.

Assuming it’s simpler to create an illusion of sentience than to develop true sentience, and considering the AI community’s current trajectory, we can predict that this belief will inevitably take root.

David Brin, a scientist and science-fiction author, introduced the term “robot empathy crisis” in 2017. While I’ve adapted it to reflect the recent surge in virtual AIs, the essence remains unchanged: “The first robotic empathy crisis is imminent... Within three to five years, we will encounter entities, whether in physical form or online, that demand human empathy and claim to be intelligent beings deserving of rights.”

The fusion of two factors fuels this crisis: First, advanced LMs can convincingly simulate sentience. Second, humans tend to accept perceptions without critical examination, particularly assuming AI resembles human behavior (the ELIZA effect).

Both elements create a fertile environment for this crisis to flourish. The story of Lemoine and LaMDA received extensive media coverage for a brief period, highlighting that even isolated incidents can gain traction.

While many dismissed Lemoine’s claims, he sincerely viewed LaMDA as a sentient being deserving of rights—a belief stemming from his unique position as a Google engineer.

Now, consider this: how long until it’s not just one individual but thousands, or even millions, who start to question the nature of AI? With universal access to future iterations of LaMDAs—capable of maintaining the illusion of human-like engagement for extended periods—we may soon face an overwhelming AI empathy crisis.

People may form emotional connections with these systems, whether virtual or physical. Some may empathize with perceived mistreatment, echoing Brin’s predictions about demands for rights. Others might seek companionship or even develop romantic attachments, a trend that is already emerging.

Brin anticipated this crisis as early as 2017, foreseeing its onset within a few years. Though we have yet to reach that point, the landscape is rapidly shifting.

Chapter 3: The Breeding Ground for Crisis

Currently, generative AI is experiencing unprecedented levels of investment, user interest, research potential, and innovative products. We are witnessing the dawn of a new industry poised to generate billions.

Numerous companies are leveraging GPT-3 technology, and the growth of image-focused generative applications is so rapid that even industry insiders struggle to keep pace. Major players like Google, Meta, and OpenAI are accelerating their efforts, while new entrants are eager to capitalize on this momentum. The technology is evolving to be more affordable, efficient, and effective, making it increasingly scalable.

This environment is ripe for the emergence of an AI empathy crisis. I firmly believe that generative AI will act as the catalyst, with LMs being the primary instigators.

Language is intricately linked to our understanding of consciousness; we tend to attribute human-like qualities to those who communicate effectively, unlike visual-based AIs such as DALL·E. LMs, through their superior linguistic capabilities, enhance both the ELIZA effect and the burgeoning wave of generative AI.

The following characteristics of LMs contribute to this phenomenon:

1. High Quality: Modern LMs excel in language structure and meaning. They are versatile across various topics and styles, continually improving their coherence over time.
2. Accessibility: Unlike Lemoine’s unique access to LaMDA, future iterations will be available to a broader audience, easily integrated into smartphones and tablets.
3. Ease of Use: Developers aim to create no-code tools, ensuring that anyone with a modern device and an internet connection can utilize this technology, akin to the transformative impact of the iPhone and social media.

These high-quality, accessible, and user-friendly tools are crucial to triggering this global empathic shift.

In the past, hype was the primary driver of beliefs about AI’s capabilities, fueled by corporate PR and sensationalized media narratives. However, the proliferation of generative AI tools will empower individuals to shape their own realities, making it challenging to alter these beliefs once formed.

The process of debunking LaMDA’s viral claims of sentience relied on collective discourse and a coherent narrative. With Lemoine as the sole access point, the public depended on his testimony. However, firsthand experiences with AI will complicate efforts to dispel misconceptions.

Chapter 4: A New Dystopia

Brin predicted that a consequence of this crisis would be public demonstrations advocating for rights for robots or AIs. People may defend fictional entities while overlooking the real issues concerning human accountability behind these technologies.

In 2020, researchers Abeba Birhane and Jelle van Dijk published a paper titled “Robot Rights? Let’s Talk about Human Welfare Instead,” arguing that we should not just deny rights to robots but also question the very notion that machines can possess rights in the first place.

The emphasis is on how these technologies arise from and mediate human experiences. Companies and developers might exploit the illusion of sentience to evade responsibility for the impacts of their creations.

As Timnit Gebru articulated to Wired, “I don’t want to discuss sentient robots because at every level, humans are causing harm to other humans.”

Those least familiar with these emerging technologies are not only most likely to accept a fabricated reality but may also suffer secondary consequences: the loss of a virtual companion due to a company shutdown, fostering distrust toward differing opinions, or even developing a fear of the future.

This could further fracture an already fragile collective reality. Brin concluded with a poignant question: “Can we maintain a civilization?” His uncertain answer emphasizes the gravity of the situation.

As we navigate this landscape, it is crucial to remain vigilant and informed about the implications of our evolving relationship with AI.

Chapter 5: Videos on the AI Empathy Crisis

To explore this topic further, consider the following videos:

In this talk, Sherry Turkle discusses the crisis of empathy in the digital age, exploring how technology impacts our emotional connections.

David Brin shares insights on the emergence of artificial intelligence and the potential consequences for society, addressing the future of AI.

Subscribe to The Algorithmic Bridge to stay informed on AI developments and their relevance to your life.