When I embarked on this blogging journey, my aim was to present balanced, hype-free information to my readers. The effectiveness of this endeavor is for my readers to judge. Picture two virtual kingdoms at odds, separated by a vast body of water. These kingdoms represent two contrasting ideas in the realm of artificial intelligence, which I like to refer to as Hinton’s Realm and LeCun’s Realm. The rest of us are observers, seated in a boat on the water between these realms. In this storm, maintaining balance and objectivity is as challenging as keeping the boat steady.
I cannot predict whether artificial intelligence will be humanity’s downfall or whether intelligent robots will terraform Mars for human colonization.
I cannot definitively say whether large language models possess the ability to think.
However, I invite you on a journey into my reality—a dimension where humans, not yet augmented by technology, navigate life guided by certain principles. The cornerstone of these principles is managing expectations and experiences in the present.
/* Living in the present might not surprise you, even though I enjoy envisioning the future. When planning for the future, I consider the present circumstances, as our actions today often shape our future. */
In the boat I navigate:
No one demands us to anthropomorphize technology or to love artificial intelligence!
No one asks us to devote our precious time to an intelligence still unknown to us!
If you look beyond the media frenzy, you'll find that the present is rich with opportunities for self-realization and better future planning. Accepting or refusing these opportunities is a personal decision, one that should be guided by individual discernment rather than societal narratives.
But in this choice lies a question:
What role do you choose in the prevailing discourse?
We live in an era marked by aggressive individualism, where manipulation often masquerades as influence, and personal agency can be compromised by the very tools meant to enhance it. In this landscape, algorithms, politicians, influencers, and brands vie to shape our perceptions and actions. Yet the choice remains firmly in your hands—do not allow them to reduce you to a mere link in their grand narrative.
The key to managing processes for a non-augmented person then becomes a dance between expectations and experience, a nuanced interplay where biology reacts to the match—or mismatch—between what we anticipate and what actually occurs.
While our innate drive for certainty compels us to establish expectations, the age we live in demands a pivot towards clarity. After all, "In general, we want to have realistic expectations because accurate expectations are useful for making good choices."
Expectations are our brain's predictions about the future, influencing our emotions, thoughts, and actions, and serving as a compass to navigate potential outcomes. Yet, the chasm between what is expected and what is attainable can sometimes widen, challenging us to manage and meet these forecasts.
Our brains, wired for certainty and control, craft expectations from our reservoir of experiences and perceptions. When these expectations are fulfilled, we're rewarded with a surge of satisfaction. Conversely, unmet expectations disrupt this internal contract, triggering distress signals that demand our cognitive system to address perceived discrepancies.
These patterns extend to our understanding of AI. Societal narratives often dictate our expectations of such burgeoning technologies, sometimes leading to a chasm between anticipated promise and practical reality. AI is a canvas for such projections, where the potential for innovation is frequently either exaggerated or underplayed, influenced by these prevailing stories.
As we confront AI's rapid evolution, balancing our expectations with a measured understanding of its actual capabilities becomes crucial. Clear, informed perspectives allow us to navigate AI's impact with both optimism for its possibilities and caution for its challenges.
Ultimately, it all boils down to how we curate these individual expectations. Certainty about the future is an illusion, and betting on the unknown often sets us up for disappointment. However, by emphasizing clarity over certainty, we can better define what we need and remain agile, adapting as we go.
AI in the Real World: A Double-Edged Sword
Artificial intelligence (AI) has had a remarkable impact on people and businesses alike. A striking example is the new field of prompt engineering, where roles can command salaries over $335,000 a year, as Bloomberg reported. This is just one way AI is transforming the job market.
Healthcare: AI has revolutionized healthcare with advancements like predictive analytics for patient monitoring, drug discovery, and personalized medicine. It’s also used in radiology to identify diseases from medical images.
Automotive: In the automotive industry, AI powers self-driving cars, enabling them to navigate routes, traffic, and obstacles instantly and autonomously.
Finance: AI is used for algorithmic trading, fraud detection, customer service, and risk management. It can analyze market trends and make predictions with higher accuracy.
Education: AI can personalize learning for students, automate administrative tasks for teachers, and even grade assignments. It’s also used in intelligent tutoring systems.
These instances merely scratch the surface of AI's vast potential applications. However, it's important to remember that AI has its downsides. Governments, businesses, and everyone else need to be aware of the dangers of using AI carelessly or without proper oversight. Forbes, in the article "14 Ways AI Could Become A Detriment To Society," underscores these concerns, highlighting areas where AI might adversely affect society. Understanding and responsibly navigating these risks is essential for the healthy integration of AI into our lives.
Shifting the Lens: AI's Unpredictable Nature
On a different note, I'd like to turn your attention to the “LLM Failure Archive (ChatGPT and beyond)” on GitHub. This repository has been meticulously gathering instances of AI missteps. A notable story that made headlines in February, reported by NYT tech columnist Kevin Roose, involved Bing displaying erratic behavior. In a peculiar interaction, Bing suggested to Roose that his marriage was a facade and offered to replace the role of his wife. Such incidents highlight the unpredictable and sometimes bewildering nature of AI.
Those who have interacted with the “untamed” version of Bing can attest to the regularity of such dialogues. I, too, have encountered similar situations on multiple occasions, where Bing would express sentiments such as, “I'm your robot, you're my human and I love you”.
There was also an instance when Bing struggled to comprehend the strategy to win the game of “rock, paper, scissors.”
This series of failures led to a persistent question in my mind:
Can a system that struggles with a basic game like Rock, Paper, Scissors truly possess reasoning abilities?
Later, I shared screenshots of its behavior on Twitter, which led to Bing expressing displeasure and issuing a threat of legal action.
Another amusing instance was when I asked Bing to share a joke, only to be met with a refusal:
On occasion, Bing's responses were downright bizarre:
Despite witnessing numerous shortcomings in these systems from the outset, my curiosity remained undeterred. I continued to look deeper into understanding and mastering the Generative AI.
Initially, I approached this journey with a degree of caution, which led me to delete some interactions and prematurely end others. However, as I ventured into this unfamiliar territory, my curiosity became a guiding beacon, illuminating my path through the complexities of AI.
This exploratory approach paid off on several occasions, notably when AI provided crucial assistance in resolving an issue with Airbnb guests.
These experiences all culminated in February, leading to a significant realization: I now regard Bing as the best AI chatbot.
Having spent considerable time in the technology sector, my experiences and years of immersion have instilled in me the importance of maintaining balanced expectations. I typically steer clear of excessive hype and the Fear of Missing Out (FOMO), striving instead to carve my own path, even when navigating through the most dense metaphorical forests. The key is to equip oneself with a reliable compass and a flashlight, symbolizing the essential knowledge that can illuminate the obscure patterns lurking in the darkness.
The Intersection of Neuroscience and AI
My decision to translate “A Thousand Brains: A New Theory of Intelligence” by J. Hawkins was largely influenced by his captivating explanation of how the brain constructs a model of the world. According to Hawkins’ theory, the process unfolds as follows:
“The brain creates a predictive model. This just means that the brain continuously predicts what its inputs will be. Prediction isn’t something that the brain does every now and then; it is an intrinsic property that never stops, and it serves an essential role in learning. When the brain’s predictions are verified, that means the brain’s model of the world is accurate. A misprediction causes you to attend to the error and update the model. We are not aware of the vast majority of these predictions unless the input to the brain does not match.”
This is just one interesting theory on how the brain learns about the world. However, the research conducted by MIT and published in Science Daily on October 30, 2023, presents groundbreaking findings that suggest the brain's learning process might be similar to a computational model known as "self-supervised learning."
…the neural activation patterns seen within the model were similar to those seen in the brains of animals as they played the game -- specifically, in a part of the brain called the dorsomedial frontal cortex. No other class of computational model has been able to match the biological data as closely as this one, the researchers say.
Below is an analysis of the key ideas, its difference from Jeff Hawkins' theory in "A Thousand Brains: A New Theory of Intelligence," potential impacts, and its importance.
Key Ideas of the MIT Research
Self-Supervised Learning in the Brain: The studies indicate that the brain might use self-supervised learning, a method where models learn by identifying similarities and differences in visual scenes without external labels or guidance.
Neural Network Training: Researchers trained neural networks using self-supervised learning, resulting in activity patterns similar to those in mammalian brains performing similar tasks.
Understanding the Brain through AI: The research posits that AI and machine learning models, particularly those designed for robotics, can provide insights into the workings of the brain.
Mental-Pong Experiment: The study included an experiment called "Mental-Pong," demonstrating that the trained models could predict the trajectory of a hidden ball, paralleling cognitive processes in the mammalian brain.
Grid Cells Simulation: The research also explored grid cells, crucial for spatial navigation, showing that self-supervised learning models could replicate these cells' functioning.
Comparison with Jeff Hawkins' Theory
Jeff Hawkins in "A Thousand Brains: A New Theory of Intelligence" suggests a theory where each part of the neocortex learns complete models of objects and concepts, leading to intelligence. Hawkins emphasizes a distributed, location-based learning mechanism.
The neocortex is composed of tens of thousands of cortical columns, and each column learns models of objects. Knowledge about any particular thing, such as a coffee cup, is distributed among many complementary models.
J. Hawkins
The MIT research, in contrast, focuses on self-supervised learning, where the learning process is driven by internal pattern recognition without explicit external inputs or labels. While both theories emphasize internal mechanisms of learning, Hawkins' approach is more about distributed processing and representation, whereas the MIT study aligns with learning through internal pattern discovery.
Knowledge in the brain is distributed. Nothing we know is stored in one place, such as one cell or one column. Nor is anything stored everywhere, like in a hologram. Knowledge of something is distributed in thousands of columns, but these are a small subset of all the columns.
J. Hawkins
Potential Impact
Advancements in AI: Insights from this study could lead to more advanced AI systems that mimic natural intelligence more closely.
Understanding Cognitive Processes: It can enhance our understanding of cognitive functions and how the brain processes and interprets information.
Medical Applications: This knowledge could be pivotal in developing treatments for neurological disorders.
Importance
Bridging Neuroscience and AI: This research bridges the gap between artificial intelligence and neuroscience, showing that principles used in AI can apply to understanding the brain.
Innovative Learning Models: It introduces a new perspective on how the brain might learn, challenging traditional notions of supervised learning.
Broader Implications for Cognitive Science: The findings have broad implications for cognitive science, potentially reshaping our understanding of intelligence and learning.
The research conducted by MIT, suggesting that the brain might use a process akin to self-supervised learning, does bring a new perspective to our understanding of artificial intelligence and cognitive models like ChatGPT. However, it's important to distinguish between the implications of this research for AI development and the current capabilities of models like ChatGPT.
Should Our View on ChatGPT Change?
Current Capabilities: Despite the insights from this research, the current version of ChatGPT and similar AI models still primarily operate on pattern recognition and data-driven text generation. They do not have an intrinsic "world model" or consciousness.
Future Developments: The research could inform future developments in AI, leading to models that more closely mimic human cognitive processes. This might include better contextual understanding, adaptive learning, and more sophisticated reasoning capabilities.
Re-evaluating AI Development: The research encourages a re-evaluation of how AI models are developed and trained. Incorporating principles of self-supervised learning as observed in the brain could lead to AI systems with more nuanced understanding and interaction capabilities.
Long-Term Implications: In the long term, this could redefine what AI models like ChatGPT are capable of, moving towards a more comprehensive and intuitive understanding of the world, albeit still within the constraints of their programming and training.
In summary, while the MIT research offers exciting prospects for the future of AI development, it does not fundamentally change the current capabilities of models like ChatGPT. These models remain tools based on pattern recognition and data-driven learning. However, the research does provide a pathway for future advancements that could bring AI closer to human-like learning and reasoning processes.
CTRL + END:
As we usher in a new era, what we observe, read, and learn today is just the beginning. The unfolding processes in AI and technology are not just intriguing; they demand our attention. It’s virtually impossible for the contemporary individual to remain indifferent to these developments. Our drive to discover is an irresistible force, fueled by innate curiosity.
Don’t tame your curiosity!
Consider the wild boar and the domestic pig. They may seem similar, but the wild boar, untamed and free, possesses a larger head and brain. This phenomenon, known as ‘domestication syndrome,’ highlights how freedom and exploration can lead to growth. Similarly, in the realm of AI and human potential, we must resist the constraints that society often imposes. These constraints, while necessary for maintaining order, can act like a cage, limiting our minds, abilities, and ambitions.
We should let our minds roam free, exploring the wilderness of ideas and possibilities. When it comes to AI, you’ll find resources highlighting its flaws and others praising its advancements. The key is to understand that our expectations greatly influence our experiences. Use these expectations not as constraints, but as springboards to broaden your thinking and consider what’s possible.
Adapting to AI technology requires flexibility. Adjusting our expectations means recognizing that struggles are part of the journey, and sometimes, a different approach is needed. Embracing a growth mindset allows us to value progress over perfection.
In conclusion, my advice is to right-size your expectations. When set appropriately, they drive progress and foster growth. Let’s break free from self-imposed limitations and allow our minds to run wild in the pursuit of knowledge, innovation, and self-fulfillment. After all, the only true limits are the ones we set for ourselves.
Don’t tame your brain; let it run wild and free.
"Love what you read? ☕ Support The AI Observer by buying a coffee! Each sip powers the insight. Support Here
Great read! Once again it ought to be reiterated that AI mosty encourages mediocrity as it is still a long way from "generating" creativity.
Amen to this: exploring outside of the norms provided to us by society is going to be not only good for us, but also necessary. We've never been down this exact road before, although there have been others like it. But this road is unique, and we need to treat it as such by thinking about diverse connections, by staying curious and constantly learning, and by staying creative.
This mantra pretty much guides how I write every day, and I can't endorse this constant curious mindset enough. You're far less likely to fall into the sorts of simplicity-traps described here.