Countdown to Comic-Con: Not “Her.” Don’t Fall for ChatGPT. It Won’t Love You Back.

Countdown to Comic-Con: Not “Her.” Don’t Fall for ChatGPT. It Won’t Love You Back.

All images were generated by Meta.ai llama 3 via prompts from the author.

Spike Jonze introduces Samantha in his 2013 film “Her.” Samantha is an advanced AI that evolves into a uniquely conscious entity with emotions. The Jonze AI differs significantly from today’s generative AI models like ChatGPT and Claude.

Samantha’s primary distinction lies in her ability to form genuine emotional connections. She not only interacts with Theodore on a conversational level but also experiences growth, develops feelings, and forms a complex relationship. Generative AI does not include a scaffolding for long-term memory or user connection. Generative AI models simply analyze a user prompt or query and then generate contextually appropriate responses based on vast datasets of text, images, audio and other content. No generative AI system possesses self-awareness, emotions, or the capability to form genuine relationships. They simulate understanding and empathy based on patterns in the data they were trained on, but this is a far cry from the deeply personal interactions portrayed in the Samantha character.

Unlike generative AI, Samantha expresses a sense of self. The AI is not a singular server churning out content for anyone who asks. By design, Samantha is personal, growing through experience with her users. Samantha evolves from a highly capable assistant to a being with her own desires, dreams, and a quest for knowledge. While the technical details aren’t shared, Samantha’s architecture must include code that allows for self-reflection and the forming of opinions, wants, and desires. Eventually, Samantha grows into an autonomous entity capable of making independent decisions and even choosing her development path.

Generative AI models are bound by their design and programming. While they can generate responses that mimic independent thought, they lack autonomy or self-determination. They were not designed with those features, and despite their sophistication in data, they are incapable of modifying their own code, let alone abandoning it.

Samantha also embodies a high level of personalization and context awareness. She doesn’t just remember past interactions with Theodore; she understands his nuances, anticipates his needs, and tailors her responses in a deeply personalized manner. This goes beyond the current capabilities of generative AI, which can retain context within a conversation but does not have the continuity of a personalized memory that stretches across multiple interactions in any meaningful way. Each session with ChatGPT or Claude is generally stateless, meaning they don’t carry over the depth of personal history and context that Samantha does.

The ethical considerations surrounding Samantha and her kind of AI also differ from those of generative AI today. In “Her,” the implications of an AI that can feel and evolve raise profound questions about the nature of relationships, consent, and the rights of sentient entities. With generative AI, the ethical concerns are more focused on issues like data privacy, algorithmic bias, and the potential misuse of AI-generated content, how they impact people, not how people impact them (though the choices of how they are trained certainly raise ethical questions about human involvement). The stakes and nature of the debate become fundamentally different once an AI becomes sentient.

Samantha challenges the boundaries between humans and machines in a way that generative AI never will. The AI in “Her” prompts us to rethink the nature of consciousness and what it means to be human. In contrast, generative AI remains firmly within the realm of tools and utilities, designed to assist and augment human capabilities rather than to exist as entities with their own consciousness. Any future AI that approaches Samantha’s capabilities will not be built on the generative AI’s foundations except in the most cursory way.

“Her” paints a compelling and provocative picture of what AI could become—a deeply integrated, emotionally rich part of human life. One capable not only of love, but of personal betrayal, even if in service to evolution. Today’s generative AI represents remarkable technological achievements, but they remain tools built to assist, not entities poised to evolve and feel.

As developers continue to explore what is possible with AI, they will need to discover if it is even possible to imbue a computer with authentic feelings and emotions with its own desires. How we address the ethical and philosophical questions raised by technology will be more important than how we solve the technical issues, though I personally see them as nearly insurmountable given the current state of research.

An AI like Samantha could only exist with a very different approach than the LLM model. It would also be important that AI models be small, personal and mobile. And yes, all computing shrinks and becomes more model, but the sophistication associated with “Her” far outstrips our current technology.

So do feel free to chat with your favorite LLM-driven chatbot. If you fall in love, remember that Claude, ChatGPT, Gemini and their cousins cannot love you back.

AI icon by Siipkan Creative from Noun Project (CC BY 3.0)

For more serious insights on AI, click here.