The crucial link between excitement and earnings
Here’s a revised version of your content with the requested changes:
—
Produced by Pause AI, an international activist group that co-organized the protest, it concluded with a plea: “Pause AI until we know what the hell Step 2 is.”
In the South Park episode “Gnomes,” which aired in 1998, Kenny, Kyle, Cartman, and Stan discover a community of gnomes that sneak out at night to steal underpants from dressers. Why? The gnomes present their pitch deck: “Phase 1: Collect underpants. Phase 2: ? Phase 3: Profit.”
The gnomes’ business plan has since become iconic among internet memes, often used to satirize everything from startup strategies to policy proposals. High-profile figures, such as Elon Musk, have referenced it to illustrate ambitious plans, like funding a mission to Mars. Currently, it encapsulates the state of AI: companies have developed the technology (Step 1) and promised transformative results (Step 3). However, how we arrive at that transformation remains uncertain.
As far as Pause AI is concerned, Step 2 should involve some form of regulation. The specifics of what this regulation will entail and who will enforce it are subjects of ongoing debate.
On the flip side, advocates of AI believe that Step 3 represents a significant breakthrough, often glossing over the critical middle stage. They see us advancing toward a promising future supported by what they describe as an “economically transformative technology.” While they have a general idea of the destination, the path forward remains unclear. Will everyone reach their goal? Will anyone?
For every bold assertion about the future, there exists a more cautious evaluation of the practical implications, which tempers the hype. Consider two recent studies. One, from Anthropic, predicted which job sectors will be most impacted by large language models (LLMs). A key takeaway: employees in management, architecture, and media should prepare for change; those in groundskeeping, construction, and hospitality may be less affected. However, these predictions are largely speculative, based on LLM capabilities rather than their real-world performance.
Another study, released in February by researchers at Mercor, an AI hiring startup, evaluated several AI agents powered by leading models from OpenAI, Anthropic, and Google DeepMind across 480 workplace tasks often carried out by human bankers, consultants, and lawyers. The results showed that every agent tested struggled to complete most of its responsibilities.
Why is there such divergent opinion? Several factors contribute to this. First, it’s essential to consider who is making the claims and their motivations. For instance, Anthropic has a vested interest in the outcomes. Moreover, many assertions regarding imminent advancements largely stem from the rapid progress of AI coding tools. However, not all tasks can be addressed solely through coding; other studies indicate that LLMs struggle with strategic judgment, for example.
—
This revision maintains the core content and tone while removing specific external references.
