Reasonable Explanations for Widespread Disappointment in AI

Understanding AI: Beyond the Hype and Towards Practical Automation

In the current landscape, it’s no surprise that many feel underwhelmed by the advancements in artificial intelligence. A significant reason for this sentiment is the pervasive presence of sensationalist narratives driven by some tech entrepreneurs and capitalists. These figures often promote superficial solutions—think plastic wrap infused with PFAS—to problems that don’t truly exist, fueling a noisy, profit-driven environment that often feels more like a digital slot machine than a genuine technological revolution. Sadly, much of this content is spammy and lacks substance.

However, beneath the cluttered surface, a quieter revolution is underway. Skilled programmers and researchers are leveraging AI to create tailored automation solutions that streamline workflows and enhance productivity. Unlike one-size-fits-all tools, these custom automations are highly specific, often requiring just minor adjustments when applied to different contexts. We shouldn’t expect a single universal AI solution to solve every problem overnight—such an achievement is still on the horizon.

The real breakthrough isn’t about eliminating jobs through automation per se. Instead, it’s about automating the process of automation itself. This involves orchestrating a series of tasks—comprising multiple steps, possibly looping, and interacting with various APIs—into cohesive workflows. In essence, the goal is to automate the creation of automated systems, enabling AI to build and refine automation pipelines independently.

At first glance, automating a task might seem straightforward, involving just a handful of steps. But in reality, each step demands carefully crafted prompts, an appropriate sequence, and integrated memory systems. Connecting these steps with APIs adds another layer of complexity, necessitating multiple specialized agents: one to generate prompts, another to design the architecture—including memory management—and a third to handle API interactions.

Encouragingly, these capabilities are not science fiction. AI systems have been writing their own prompts for some time—an area explored in research such as the 2023 paper “[https://arxiv.org/abs/2310.08101]”. More recently, protocols like MCP (Memory-Control Protocol) have emerged, allowing APIs to embed instructions directly within communication protocols. Additionally, tools like AgentForge now support YAML-defined architectures, empowering AI to autonomously build complex workflows from scratch: sequencing prompts, managing memory, and connecting APIs—all without coding.

The key takeaway? We are on the cusp of a new era where the automation of automation will become commonplace.