Learning from the Past: How Previous Technological Transformations Can Guide AI Development

Risto Miikkulainen, Bret Greenstein, Babak Hodjat

Abstract

Artificial Intelligence (AI) is rapidly changing many areas of society. While this transformation has tremendous potential, there are several challenges as well. Using the history of computing and the world-wide web as a guide, in this paper we identify pitfalls and solutions that suggest how AI can be developed to its full potential. If done right, AI will be instrumental in achieving the goals we set for the economy, the society, and the world in general.

Introduction: Navigating the AI "Wild West"

AI development currently feels like a "Wild West." We face a lack of global standards, significant security and privacy risks, and a small group of experts making high-stakes judgment calls. Many fear that AI is an uncontrollable force that might lead to disaster.

This paper addresses this anxiety by suggesting that AI is not as "different" as we think. By looking at how the personal computer (PC) and the Internet moved from niche research projects to foundational infrastructure, we can identify a historical roadmap. This comparative analysis helps identify specific pitfalls to avoid and policy actions to take to ensure AI reaches its full potential responsibly.

The Four-Phase Framework

To understand the trajectory of AI, we can look at the four-stage framework that characterized previous technological shifts like personal computing and the World Wide Web (WWW):

  1. Standardization: Establishing common protocols (e.g., IBM PC architecture, HTML).
  2. Usability: Making the technology accessible to non-experts (e.g., Graphical User Interfaces, CSS).
  3. Consumerization: Expansion to individual creators and mobile access (e.g., iPhone, Web 2.0).
  4. Foundationalization: The technology becomes "invisible" infrastructure (e.g., Cloud computing, the digital economy).
Table 1: A comparative framework of technological transformation across Computing, the World Wide Web, and AI.

By mapping AI onto this framework, we can see that while it follows a similar path, it exhibits unique characteristics that require specific policy interventions.

The AI Usability Paradox

One of the most striking insights is the "AI Usability Paradox." Unlike previous technologies, AI—specifically Generative AI and Large Language Models (LLMs)—achieved high Usability before Standardization. Because AI is so easy to interact with through natural language, humans tend to trust it prematurely. This exacerbates risks like hallucinations and misinformation, as the underlying "safety" and "standardization" layers are still being built.

Lessons from History: Interoperability and Safety

History provides two critical lessons for the current phase of AI:

  • The Need for an "HTML for AI": Just as HTML allowed different browsers to display the same web, we need open standards for "Agentic AI." This would allow different AI agents (e.g., a car's vision system and a city's traffic planner) to interoperate and be swapped out without breaking the system.
  • Safety as a "Day One" Requirement: The Internet is difficult to secure today because safety wasn't built into the original protocols. AI safety standards must be developed alongside the technology, not as an afterthought.

Furthermore, the "Browser Wars" of the 90s show that when one player dominates a phase, innovation stalls. Open-source AI is critical to prevent monopolies from stifling the current surge of progress.

A Strategic Roadmap for AI Policy

As AI moves toward the "Foundational" phase, it shifts from predicting what will happen to prescribing what should be done. This makes human-defined objectives—such as prioritizing environmental impact or social equity—more important than ever. The proposed roadmap suggests four key policy actions:

  • Phase 1 (Standardization): Ensure safety and interoperability.
  • Phase 2 (Usability): Make systems easy to use without harm.
  • Phase 3 (Consumerization): Ensure open innovation before heavy regulation.
  • Phase 4 (Foundationalization): Establish human-defined objectives to ensure positive impact.

Conclusion

If this guided path is followed, AI will eventually become as essential and "invisible" as electricity or plumbing. By learning from the past, we can move beyond the "Wild West" phase and build an intelligence-based society that empowers human decision-making rather than replacing it.

Summary updated

Abstract

Artificial Intelligence (AI) is rapidly changing many areas of society. While this transformation has tremendous potential, there are several challenges as well. Using the history of computing and the world-wide web as a guide, in this paper we identify pitfalls and solutions that suggest how AI can be developed to its full potential. If done right, AI will be instrumental in achieving the goals we set for the economy, the society, and the world in general.
BibTeX
@misc{miikkulainen2019learning,
  author = {Risto Miikkulainen and Bret Greenstein and Babak Hodjat},
  title = {{Learning from the Past: How Previous Technological Transformations Can Guide AI Development}},
  year = {2019},
  eprint = {1905.13178},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1905.13178},
}
Figure