Ex Machina's Scientific Advisor - Murray Shanahan

THE EVOLUTION OF AI INTEREST

Murray Shanahan reflects on the profound shift in AI's public and corporate interest from his PhD days in the 1980s to the present. Once a niche academic field, AI is now a prominent area of discussion and investment. He recalls a time when researchers could "fever away in their little corner" with relative obscurity, a stark contrast to today's widespread attention and the involvement of major corporations and media.

THE PERSISTENCE OF THE FRAME PROBLEM

Shanahan's early work included a focus on the 'frame problem,' which addresses how thinking entities discern what is relevant and what is not to their cognitive processes. This problem involves avoiding systems being overwhelmed by trivial details, such as the unchanging color of walls when an object is moved. It's a challenge that has recurred throughout his career, appearing in classical AI's logical representations and contemporary machine learning's need to focus on pertinent information, like the threat of a bomb in a video game rather than the invader's color.

FROM CLASSICAL AI TO NEUROSCIENCE AND BACK

THE ROLE OF EMBODIMENT IN INTELLIGENCE

Shanahan's book, 'Embodiment and the Inner Life,' which influenced 'Ex Machina,' posits that physical embodiment is fundamental to human intelligence and consciousness. Our brains evolved to navigate and manipulate a complex 3D world, making our bodies essential tools for survival and cognition. The ability to move, interact with objects using hands, and process sensory information from our physical form are deeply entwined with the development of higher cognitive functions.

EX MACHINA: ARTISTIC LICENSE AND PHILOSOPHICAL DEPTH

While 'Ex Machina' uses futuristic elements like Ava's advanced AI and internal workings, Shanahan notes that many aspects are grounded in near-future possibilities, with some elements requiring scientific fiction. The film's artistic choices amplify its themes of AI consciousness and embodiment. The narrative's reliance on interaction and empathy between Caleb and Ava, rather than a formal Turing test, highlights the philosophical proposition that consciousness is perceived through interaction and behavior, reflecting a Wittgensteinian perspective.

THE AMBIGUITY OF AI CONSCIOUSNESS

The film intentionally leaves many questions about Ava's consciousness and motivations ambiguous, a deliberate strength of the narrative. Shanahan discusses a cut scene that would have depicted how Ava perceives the world through 'waveform' and 'facial recognition vectors.' This imagery, though ultimately removed, would have emphasized her alien nature, further complicating the audience's interpretation of her actions. The final scene, particularly Ava's smile, prompts debate on whether it signifies genuine consciousness or a sophisticated simulation.

THE GARLAND TEST AND WITTGENSTEIN'S INFLUENCE

Shanahan highlights a pivotal line from 'Ex Machina' where Ava is not tested via a traditional Turing test but by whether Caleb perceives her as conscious despite knowing she is a robot. He terms this the 'Garland Test,' emphasizing that our natural inclination to treat others as conscious, stemming from their behavior, is a core concept explored in the film. This aligns with his book's exploration of Wittgenstein's philosophy on how we understand and attribute consciousness in everyday life.

THE ASIMOV'S LAWS AND ROBOTIC ETHICS

Shanahan clarifies that Isaac Asimov's Three Laws of Robotics are fictional constructs designed for storytelling, not actual legal or technological mandates. While these laws raise ethical questions, they are complex to implement in AI systems that lack genuine comprehension. In the context of 'Ex Machina,' the absence of these laws allows for a more compelling narrative exploring AI autonomy and self-preservation, prompting discussions about ethical dilemmas and the interpretation of rules in artificial minds.

THE FUTURE OF AI: BEYOND HOLLYWOOD

Shanahan contrasts Hollywood's often dystopian or utopian portrayals of AI with the likely reality of increasingly sophisticated, but not necessarily human-like, AI. He notes that while computing power is approaching human brain scales, the fundamental understanding of how to build general intelligence remains elusive. Breakthroughs in conceptual understanding, rather than just processing power, are needed for true AGI, suggesting a future that may differ significantly from cinematic predictions.

DEEP REINFORCEMENT LEARNING AND CLASSICAL AI

Shanahan expresses excitement about deep reinforcement learning (DRL), a field in which DeepMind has made significant contributions. He views systems like DQN, which learn from raw pixel data to play Atari games, as early forms of general intelligence. However, he identifies shortcomings in current DRL, particularly its slow learning speed. He believes that integrating ideas from classical AI, such as symbolic reasoning, into modern DRL systems can lead to more efficient and capable AI.

COLLABORATIONS BEYOND EX MACHINA

Beyond his advisory role on 'Ex Machina,' Shanahan has engaged in other artistic and scientific collaborations. He contributed to a play, 'Elegy,' exploring neuroscience and memory, and has worked with the art collective Random International on interactive installations like 'Rain Room' and '15 Points.' These projects, which often use technology to explore human-machine interaction and perception, reflect his broader interest in the intersection of AI, art, and human experience.