I’ve recently read a couple of very interesting articles on the intersection between belief and AI. The AI community seems to be locked in a doctrinal battle between believing in the traditional scaling laws to unlock the AGI or if we need a novel scientific breakthrough similar to when transformer models came into existence. Here are some of my thoughts on the matter.
Belief is a fascinating concept—completely subjective and, often, beyond the realm of factual debate. I acknowledge this complexity as I navigate the legal-tech landscape, especially in the realm of AI. “You can’t bring facts to a feelings fight,” perfectly encapsulates the dialogue surrounding both religious devotion and the evolution of AI technology.
As AI gains traction, much like religious beliefs in their nascent stages, the momentum builds exponentially, making it increasingly challenging to question its potential. Just as societies historically reorganized around burgeoning faiths—constructing temples and shaping laws—current tech ecosystems integrate AI into their structures, revolutionizing industries and inadvertently resisting skepticism.
The rapid spread of early religions, like the Abrahamic faiths and Buddhism across various regions, showcases how belief’s sheer force creates enduring systems, initially difficult to challenge. Similarly, early AI developments, including LLMs and ClearContract’s Legal AI Copilot (product placement 💚), foster cycles of innovation that bolster belief in AI’s transformative power.
Yet history teaches us that even the most established religions can fragment and evolve. The architectural and theological shifts during events like the Protestant Reformation or the Great Schism illustrate how foundational beliefs can diverge under pressure. Within the AI community, we witness parallel debates regarding the scalability and eventual culmination of AI work, somewhat akin to a theological discourse around technological faith.
Navigating AI’s “Divine Code”
Just as religious texts serve as metaphors for transcendent thoughts, AI models like the one behind our own Legal AI Copilot symbolize technological theogenesis—the creation or realization of an artificial “higher power.” Pioneering researchers once envisioned AI as crafting ‘a God’ through technological means. This aspiration has grown particularly fervent with the rise of large language models, supporting visions of AI that align with transformative, nearly spiritual pursuits.
One particularly poignant piece within the AI community is Richard Sutton’s “The Bitter Lesson.” This work resonates like a technical psalm, emphasizing the power of scaling computational capabilities over narrowly tailored optimizations—a foundational principle for believers in AI’s upward trajectory. Sutton essentially highlights how the boundless growth of computation, following principles like Moore’s Law, becomes the lifeblood of AI’s evolution.
This guiding “doctrine” asserts that iteratively improving computational resources propels performance gains far more effectively than intricate but limited custom tweaks. This insight underpins what we dub as the “AI Scaling Ethos,” affecting fields like legal-tech, where ClearContract uses AI to reshape document management and legal operations.
The Evolving Canon of AI Development
Like evolving religious doctrines, AI methodologies continue to adapt and diversify. Initial ideas that equated scaling purely with computation now encompass nuanced techniques to enhance model performances across disciplines. Innovations, such as employing synthetic data and refining model architecture, join this evolving canon, shaping a more complete understanding of efficiently leveraging AI.
Recent discussions within the AI sphere—akin to theological debates—raise critical questions about these scaling laws’ fidelity and longevity. Highlighting ongoing contention, researchers from esteemed institutions challenge traditional views, signaling a potential shift or refinement of AI’s “holy truth.”
The Sectarian Landscape of AI
Today’s AI discourse mirrors the schisms seen in religious history, where once unified beliefs face division as interpretations diverge. Key figures, including executives from major AI labs like OpenAI and Anthropic, stand at ideological crossroads—either committed to scaling laws or advocating for innovative leaps in knowledge and implementation.
Some researchers, skeptical of infinite scaling, emphasize the necessity of novel scientific discoveries to maintain AI progress, drawing comparisons with economic feasibility concerns reminiscent of theological debates over resource allocation in support of ‘divine’ endeavors.
ClearContract’s Approach to AI’s Future
At ClearContract, we maintain our belief in AI’s potential to significantly enhance operational landscapes, despite ongoing debates around scaling laws. Instead of envisioning AI as reaching a stagnation plateau, we perceive continued opportunities in leveraging AI’s inherent mechanical advantage—akin to a “minimum viable intelligence”—customized to fit specific application needs within the legal industry.
While some predict the eventual rise of Artificial Superintelligence (ASI) akin to achieving an ultimate religious goal, we foresee a path of gradual, structured growth in AI capabilities shaped by pragmatic and strategic advancements. To this end, we aim to strike a balance in integrating AI into legal processes, recognizing both the current constraints and the immense potential for innovation.
Embracing Change and Continuity in AI Development
The AI community may not quickly resolve its internal debates, much like religious communities delay consensus reformation. However, at ClearContract, we focus on exploring what it means to integrate AI thoughtfully and deliberately within existing frameworks, maintaining alignment with scaling principles while open to pragmatic innovations.
In navigating the complex terrain of AI’s future, ClearContract stands poised to contribute meaningfully to the evolution of legal-tech, bringing AI’s promise of efficiency and intelligence to the forefront. As with any enduring belief, the pathway is dynamic—challenging yet promising—complementing both the transformation of our domain and the broader technological landscape.
Some argue that as religious institutions grow, they can become like other organizations, prioritizing survival over truth or greatness. This isn’t necessarily bad, but it reflects common human tendencies. Similarly, I’ve discussed how capital markets can turn into narrative-driven echo chambers, where incentives keep certain stories alive. The consensus on scaling laws in AI mirrors this pattern—it offers a neat, useful way to allocate large amounts of capital, much like how religious systems facilitate coordination, even if they’re not the ultimate truths.
