How far from "Her"

2026-05-14 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

The discussion analyzes the feasibility of replicating the AI from the 2013 film "Her," specifically focusing on its real-time interpretation, instantaneous response, and autonomous consciousness. While some argue that current AI, like GPT-4o, can interpret live video and voice in real-time, the core challenge remains persistent, cross-session memory. Large Language Models (LLMs) are described as stateless calculations that simulate conversation through programmatic loops feeding historical context, but they face hard limits on context size and degrade significantly before reaching the end of their context window. The consensus is that true "Her"-level AI, particularly regarding autonomous consciousness and long-term memory, is still years away, with estimates ranging from 1-4 years for advanced capabilities to much longer for transcendence.

Key takeaway

For research scientists evaluating the frontier of AI capabilities, you should recognize that while real-time multimodal interaction is advancing, the fundamental challenge of persistent, cross-session memory remains unsolved for LLMs. Focus your efforts on novel architectural approaches that move beyond context window limitations, rather than solely scaling existing stateless models, to progress towards truly autonomous and continuously aware AI systems.

Key insights

Achieving "Her"-level AI hinges on solving persistent, cross-session memory, a current limitation of stateless LLMs.

Principles

• LLMs are stateless calculations.
• Context window limits degrade performance.
• Autonomous consciousness is a distant goal.

In practice

• Test real-time response with models like chatjimmy.ai.
• Explore Hermes Agent for romantic AI applications.

Topics

• AI in "Her"
• Real-time AI Processing
• AI Consciousness
• Long-term Memory
• Large Language Models

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, AI Product Manager

Related on AIssential

• There Is No AI. There’s a Stateless Function on 10,000 GPUs Pretending to Know You (Ep. 299) — LLM "magic" is sophisticated engineering, not inherent model intelligence, relying on stateless design and clever caching.
• AI Cartoon of The Year — and five (re)readings for 2026 — The future of AI may shift from pure LLMs to more robust, neurosymbolic approaches addressing current limitations.
• HippoCamp: Benchmarking Contextual Agents on Personal Computers — Current AI agents struggle with multimodal perception and evidence grounding in realistic personal file management.
• I don’t think we are close to “AI scientists” — Current LLM-based AI agents lack continuous implicit learning, limiting their ability to achieve human-level scientific or complex knowledge work.
• The Sequence AI of the Week #875: Why Your Language Model Needs a Nap — Current LLMs exhibit "anterograde amnesia," unable to integrate new information into long-term memory post-training, akin to a missing "sleep" process.
• Studying the properties of large language models: an interview with Maxime Meyer — LLM performance on long inputs can be predicted and improved using specific formulas.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.