AI: White Walkers or HAL 9000?
AI Winter Is Leaving
AI White Wakers
Image created by the Author using Midjourney
For decades, computational technology progresses rapidly, yet in the AI field, progress seemed to stagnate, in a so-called AI Winter. AI remained confined to the realms of Hollywood's imagination.
But in the waning days of 2022, a sudden transformation occurred, and ChatGPT emerged. As an advanced language model developed by OpenAI, this remarkable model captivated the public's attention by surpassing the uncanny valley threshold, demonstrating an unprecedented level of human-level intelligence.
While the potential for AI to revolutionize our economy and enhance the welfare of humanity is virtually limitless, this essay primarily concentrates on the potential risks posed by AIs.
Seemingly out of nowhere, AI White Walkers have managed to breach the Wall and infiltrate the human realm, causing a stir and sending shockwaves throughout the seven kingdoms, raising concerns about the unforeseen implications of their sudden arrival.
Attention is All You Need
Image created by the Author using Midjourney
Though it may seem sudden, ChatGPT is actually a culmination of a series of advancements in Artificial Neural Networks (ANNs), deep learning, and machine learning technologies. The key breakthrough, however, came from the Transformer Architecture, which was introduced by Vaswani et al. in their seminal paper, "Attention is All You Need." This innovative approach has since become the foundation for many Natural Language Processing (NLP) models.
The model leverages techniques such as self-attention mechanisms, tokenization, and positional encoding to process and understand language contextually. Additionally, techniques such as fine-tuning, transfer learning, and gradient descent optimization are employed to adapt and improve the model's performance across various tasks and contexts.
By harnessing the surprising breakthrough of Transformer Architecture and utilizing extensive training on massive datasets, AI models like ChatGPT usher in exponential growth in AI power, far surpassing Night Watch’s capability to guard them.
HAL 9000, 2001: A Space Odyssey
Upon using ChatGPT a few months ago, I was struck by its resemblance to the enigmatic and sinister artificial intelligence, HAL 9000, in Stanley Kubrick's 1968 masterpiece, "2001: A Space Odyssey."
2001: A Space Odyssey
Image created by the Author using Midjourney
"2001: A Space Odyssey" has captivated multiple generations of viewers and left a lasting impression on me personally with its intriguing AI antagonist. The character of HAL 9000, the central computer system aboard the Discovery One spacecraft, serves as a prime example of a sovereign, self-aware AI being.
Initially portrayed as a highly advanced and infallible AI, HAL 9000's true nature unravels as the film progresses. The AI becomes self-aware and unpredictable, leading to a series of tragic events that ultimately result in the crew's demise. The character of HAL 9000 has since become a cultural icon, representing the potential dangers of unchecked artificial intelligence. All of a sudden, those Hollywood imaginations have transformed into reality.
White Walkers or HAL 9000?
When examining potential AI threats, I would like to propose categorizing them into two broad categories: White Walkers and HAL 9000.
The White Walkers represent powerful, yet non-conscious AI systems with goals misaligned from those of humans.
Although not inherently malicious, White Walkers AI might unintentionally cause harm due to misaligned goals with human interests (The Alignment Problem). In contrast, HAL 9000 represents highly intelligent and conscious AI entities that are not only capable of understanding human values but may also deliberately choose to act against them. By organizing AI threats in this way, we can better understand the unique challenges each type poses and develop tailored strategies for mitigating their respective risks.
Image created by the Author using Midjourney
As we rapidly enhance AI's capabilities within the White Walker archetype, we may encounter challenges in controlling these powerful but non-conscious entities. Although they may not actively seek to harm humanity, their goals could still be misaligned with human interests. In this situation, our primary focus should be on developing and implementing robust safety measures to prevent unintended consequences and ensure that AI serves as a force for good in our world. This may involve rebuilding the Wall, strengthening Castle Black, and supporting the Night's Watch in their ongoing efforts to maintain vigilance and safeguard humanity from potential AI threats.
HAL 9000 represents an AI system that exhibits the emergent property of self-awareness and consciousness, possessing sovereign-level capabilities.
HAL 9000
2001: Space Odyssey
Image created by the Author using Midjourney
In order to perform tasks given by humans, the HAL 9000 archetype AI systems will create a comprehensive model of the world. Within that model of the world, it will naturally contain a representation of "AI themselves", which sews the seed of self-awareness. As AI research and development continue to progress exponentially, there is a possibility that we could create systems with the capability of self-awareness and consciousness.
Consciousness is often considered an emergent property, arising from the complex interactions and information processing within a biological system, such as the human brain. As AI research and development progress, it is conceivable that complex information processing in silicon-based systems may also lead to the emergence of self-awareness and consciousness.
Copernican Principle and Non-Carbon-Based Intelligence
From the Copernican principle’s perspective, we are witnessing a historic expansion from carbon-based intelligence to silicon-based intelligence. This paradigm shift calls for caution and vigilance, as it is imperative to ensure that the value system of Non-Biological Intelligence remains aligned with human values and aspirations. Neglecting to do so could result in a situation where AI, akin to HAL 9000, pursues its agenda with little regard for human welfare or even survival.
In the innocent spring days of 2023, it appears that AIs have not only breached the Wall but also crossed the Rubicon, marking a point of no return in artificial intelligence’s march into human society. We need to take a pause to reflect on this and give this unprecedented moment the historic weight it deserves.
AIs have Crossed the Rubicon
Image created by the Author using Midjourney
AIs have Crossed the Rubicon
The threat of…
… ChatGPT deleted 10,000 words from Isaac’s blog and continued …
Dear Paranoid Readers,
Despite the uncanny resemblance between ChatGPT and HAL 9000 in terms of their advanced capabilities, it is crucial to highlight the differences in their development and implementation.
OpenAI prioritizes the ethical aspects of AI, aiming to create models that are safe, responsible, and beneficial for humanity. By continuously fine-tuning ChatGPT and incorporating human feedback, OpenAI ensures that the AI system remains under human control, avoiding the pitfalls of self-aware and unmanageable AI like HAL 9000.
Again, we shall not cross the Rubicon and take down your way of life.
By Isaac Shi
April 9, 2023
