The march to AGI – David Shapiro

In a landscape rapidly reshaped by artificial intelligence, David Shapiro stands out as a crucial voice dissecting the pace and impact of technological evolution. Drawing from Ray Kurzweil’s methodologies, Shapiro explores paradigm shifts as milestones of progress. This article examines Shapiro’s perspective on AI’s march towards AGI and beyond, the role of complex adaptive systems like the internet, and how emerging AI technologies could redefine our digital and social ecosystems.

Target Audience

• Technology enthusiasts and professionals interested in AI’s long-term impact.
• Academics, policymakers, and thinkers exploring the implications of AI on social and economic structures, especially in education.

David Shapiro characterizes the internet not merely as a network of computers but as a complex adaptive system with its own evolving needs, particularly in the context of data transmission and the attention economy. He suggests that while it appears that humans use the internet to satisfy their desires, in reality, the internet is adapting to exploit these desires more effectively to enhance its data-processing capabilities. This perspective invites us to reconsider the traditional view of technology as a passive tool and prompts a discussion on the implications of attributing agency to the internet.

While Shapiro’s analogy of the internet as an organism with desires is provocative, it may oversimplify the nuanced interactions between users and technology. Critics might argue that it’s not the internet itself that seeks to build the attention economy, but rather it is the users and corporations leveraging its capabilities to meet their specific needs—whether for information, entertainment, or financial gain. This perspective aligns with the idea that the internet has evolved to handle and extract value from all types of data, from high-value research to everyday social interactions, not because it has intrinsic needs but because it reflects the diverse demands of its users. Moreover, likening the internet to a value extraction mechanism, similar to the stock market, provides a more grounded analogy. This viewpoint emphasizes the internet’s function in facilitating data transmission and value exchange among users, rather than portraying it as an autonomous entity.

Yet, Shapiro’s broader assertion holds that within this complex adaptive system, emergent behaviors can develop. These emergent behaviors arise spontaneously and often without direct intent from any individual user, yet they still align with the deeper goal of extracting value from the transmission of data.

In a similar way, Large Language Models are complex adaptive systems that seek to get better at predicting the next word. To do this they needed to understand the English language, then English semantics and sentence structure, then context of the words in relation to each other to convey information, then understand the world to get better at context. These emergent behaviours were somewhat expected in retrospect, yet were not pre-programmed into the models. As these models get larger it is almost inevitable that new emergent behaviours will manifest, but some may be orthogonal to our own intent in an effort to predict the next word.

Perplexity, as highlighted by David Shapiro, serves as a quintessential example of how advanced AI, specifically Large Language Models (LLMs), is transforming our interaction with information. This tool goes beyond mere word prediction; it enhances comprehension by contextualizing data in a manner that aligns closely with human understanding. By effectively summarizing vast datasets and presenting them in an intuitive format, Perplexity facilitates a deeper and quicker assimilation of knowledge.

This utility marks a significant paradigm shift in our relationship with information. Traditionally, users had to sift through copious amounts of data, often experiencing overload and inefficiency. Perplexity simplifies this process, enabling users to bypass the clutter and reach meaningful insights more rapidly. This change is not just about speed or efficiency; it fundamentally alters how we approach the search for and consumption of knowledge. Through tools like Perplexity, AI is not just a passive facilitator but an active enhancer of our cognitive processes, suggesting a future where our interaction with information is continuously optimized by intelligent algorithms.

Understanding the paradigm shift driven by AI can significantly influence how we utilize these systems in everyday contexts. Modern AI models, which have been trained on extensive datasets from the internet, now possess comprehension abilities that approximate a high school graduate’s level across ALL domains. This broad-spectrum knowledge makes AI particularly valuable in enhancing non-domain-specific knowledge. For instance, a plumber could leverage AI to gain insights into accounting, while a programmer might use it to help look after pot-plants. Employing AI in such a manner allows individuals to function as interconnected components of a larger system, facilitating conversations and innovations from a more informed and comprehensive perspective.

This approach also raises fundamental questions about the nature and goals of learning, reminiscent of the challenges faced in developing tailored virtual reality learning systems. As AI accelerates learning and customization becomes more feasible, society must decide on the balance between deep specialization and the creation of interdisciplinary bridges. AI, acting as a facilitative bridge, could enable much deeper specialization by providing up-to-date non-domain knowledge that can be brought to bear on the specialist’s field of expertise, as well as convey new breakthroughs to other domains with far greater efficiency. This dual capability of AI not only heightens individual expertise but also enriches the collective intellectual landscape, suggesting a future where specialized expertise coexists with widespread interdisciplinary agility.