OP-ED: The new language of AI defined
The discussions taking place over artificial intelligence (AI) are complicated. Moreover, the descriptive terms that define AI may make sense to software engineers, but are baffling to the average person. This commentary will focus on the new language of AI to help the reader become more AI literate.
In my view, there are three topics that are critical to the AI debate. First, the good. What technological advances will the AI revolution bring into our lives? Second, the bad. What are the possibilities that the developers lose control of the technology and that it poses a threat to humankind? Third, the inconvenient. This would include the uproar against data centers that are being built in residential and/or environmentally sensitive areas.
There is basic agreement among the players on the AI field concerning the new language of AI. However, regarding AI’s future, there is widespread disagreement on whether the good outweighs the inconveniences and the possibility of very bad results that AI could spawn. By deciphering the new AI language, we can form our own opinions and take part in the debate.
Data centers. In 2026, there is no more divisive topic than the building of data centers. Local residents facing the prospects of one in their back yard view them as a substantial inconvenience affecting property values and the quality of life.
Data centers have existed in small numbers since the 1940s. In the 1990s, internet providers began building larger facilities. In the mid-2000s, Amazon Web Services (AWS) launched modern cloud infrastructure, followed by numerous hyperscale campuses built by Google, Microsoft, and Meta. These centers provide the massive computing power required for AI development.
The good news is that the outcry against data centers has forced the tech industry to develop alternatives. SpaceX plans to build the next generation of data centers (thousands of solar-powered orbital data center satellites) to run artificial intelligence infrastructure in space. Industry experts think that a 2028 rollout is reasonable. In addition, to combat the strain on local resources, technology companies and engineers are deploying highly advanced cooling technologies, smart AI software, alternative water tracking, and decentralized power solutions.
My view is that in five years no one will be talking about the inconvenience of data centers. Very soon, less intrusive technology will provide the same service.
Artificial intelligence (AI). Generally speaking, AI is the field of computer science dedicated to creating systems capable of performing tasks that require human intelligence. Engineers often divide AI into three categories. First, is “narrow intelligence,” which are systems that have been with us for years to handle specific tasks like voice assistants (Siri) and chatbots. Next, is the more advanced “generative intelligence” that uses what are called “large language models” to create text, images, and audio. This type is also used for training new models. Lastly is the most advanced, theoretical form of AI, termed “general or super intelligence.” Super intelligence will exist when machine learning matches or surpasses human-level cognition.
AI agent. Unlike traditional chatbots or assistants that only generate text and wait for your next prompt, an AI agent uses generative intelligence to solve problems for its human “principal.” It acts like a “digital worker” rather than just a conversational tool. Millions of agents are functioning today.
AI Primacy. This unassuming term should actually carry the same weight and concern that “nuclear arms race” sparked in the last century. AI primacy means achieving dominant leadership and a decisive competitive or geopolitical advantage in artificial intelligence development. The topic is a priority in every AI company boardroom and at the highest level of every nation’s security apparatus. How fast, and with what regulations should AI be rolled out and who should control it?
The Trump administration has cited the need for AI primacy to justify not slowing down the private development of new models, even if the developers have concerns about its safety. Trump and some AI developers contend it is most important to maintain a lead over rival nations like China in developing highly advanced AI models. Doing it safely is a secondary consideration.
Recursive self-improvement (RSI). This “techy” concept is closely related to AI primacy because it is used to justify both the need to speed up and slow down AI development. RSI defines the process where AI becomes advanced enough to rewrite its own code, rapidly accelerating its own intelligence beyond human control. Presumably, at the endpoint, an RSI system could completely design its own architecture, training procedures, and goals without human oversight.
Doomers. This term defines the faction in the AI landscape who call for AI governance and slowdowns due to their concern over an existential catastrophe. This would include the permanent loss of human control over our own future and possibly human extinction. A more subtle “doom scenario” focuses on the erosion of human capability. Instead of a sudden killer AI outbreak, this path implies a future where humanity surrenders its critical thinking, decision-making, and creativity to automated bots.
Bloomers. Bloomers advocate for rapid AI development with fewer regulations. They believe the next generation of AI can be tested and controlled. They argue that it will fundamentally reshape society in a positive manner and that the benefits outweigh the risks.
For those who want to expand their AI knowledge, I can recommend two books: “Artificial Intelligence Basics: A Non-Technical Introduction,” and “Co-Intelligence: Living and Working With AI.”
Gary Stout is a Washington attorney.