What is superior intelligence? A practical guide to the idea

How we use the term

The Super Intelligence Times uses superior intelligence as an editorial lens for examining AI that exceeds human performance, the research that could extend those capabilities, and the decisions people must make along the way. It is the name of our coverage, not a certification we award to a model.

A system may be excellent at one task and unreliable on the next. A useful article should tell you which task was tested, what the comparison involved, and what the result leaves unanswered. That is more useful than treating intelligence as a single score.

Where artificial superintelligence fits

Artificial superintelligence, usually shortened to ASI, is the related research term readers will encounter. Google DeepMind's From AGI to ASI report examines the possibility of systems whose cognitive abilities exceed those of large human organisations. Its discussion concerns possible development after human-level general intelligence; it is not an announcement that such a system has arrived. Read the research abstract.

For this publication, the distinction matters. An impressive demonstration can deserve coverage without supporting a claim about general superintelligence. We will use the specific name of the capability when that is what the evidence establishes.

Read a capability claim with three questions

First, ask about the task. Was the system writing a short answer, completing a software change, checking an invoice, or coordinating several tools? These activities need different evidence.

Second, ask about the comparison. Was it measured against beginners, specialists, a previous model, or a team with access to the same tools? A percentage improvement has little meaning if the baseline is missing.

Third, ask what happened when the system failed. Did it flag uncertainty, stop for review, quietly return the wrong answer, or take an action that someone had to undo? Failure handling belongs in the capability story.

A local example

Imagine an Alberta service company testing an assistant that turns work orders into draft estimates. The useful comparison is the quality of those drafts, the time a qualified reviewer spends correcting them, and the mistakes that survive review. A polished estimate alone does not establish that the assistant understands every job.

That is our practical starting point: explain the wider research question, then identify what a reader can verify in their own work. Begin with our guide to AGI and ASI, or use the measurement guide before comparing systems.