DC DataCentreDebate.ca

Not every use is slop

What is all this compute for?

Useful AI does not excuse a weak project approval. It does mean the debate has to get past “AI bad” and “AI magic.”

Why I don’t dismiss the criticism

“My data needs are already being met. We do not need all these AI data centres.”

That sounds reasonable until you remember how infrastructure works. People usually do not need the next thing until the next thing exists and becomes boring.

The line I keep coming back to: Nobody needed today’s internet in 1994, when I built my first data centre. Until they did.

Useful AI already exists

A lot of what people see first is garbage: slop images, spam, lazy writing, bad search summaries, fake customer service, and surveillance dressed up as convenience. That does not mean the whole field is garbage.

The serious uses are less flashy and usually less annoying: medical imaging, drug discovery, accessibility, agriculture, wildfire modelling, weather forecasting, grid management, industrial design, robotics, cybersecurity, logistics, engineering, education, software development, network operations, and scientific research.

Medical tools

AI-enabled medical devices are already a real regulatory category, not just a pitch deck.

Drug and protein research

Protein-structure tools are one example of AI being used for actual science.

Accessibility

Computer vision and language tools can help people read, navigate, and understand the world around them.

Wildfires

Emergency modelling and boundary tracking are a lot more useful than another fake movie trailer.

Weather forecasting

Forecasting is one of the places where faster models can matter in the real world.

Energy systems

Grid operations, maintenance, forecasting, and industrial control are boring in the best possible way.
Real-world links on this pageThe tiles above link to real examples or authoritative sources. They are not endorsements of every product or company; they show that AI use is broader than chatbots, slop, and surveillance.

Where this could go

More data centre capacity does not automatically create public good. It can just as easily create more slop, more surveillance, and more foreign-owned infrastructure. But with the right rules, access, and public benefit, the next wave of compute could support things that are worth having.

Autonomous labs

AI-guided experiments, robotics, hypothesis generation, and faster scientific testing.

New materials

Batteries, solar cells, chips, catalysts, construction materials, and industrial processes that are hard to improve by trial and error.

Faster public modelling

Weather, wildfire, grid, flood, health, transportation, and emergency-response modelling that can run faster and more often.

Industrial systems

Predictive maintenance, robotics, plant optimization, quality control, and safer operation of complex equipment.

Education and training

Better tutoring, simulation, trades training, language support, and tools for people who do not learn well from one-size-fits-all systems.

Canadian research capacity

The part Alberta should care about: whether Canadian universities, companies, and public-interest projects get access, or just the impacts.

Read Canadian Compute

What Alberta should get out of it

AI has useful purposes. Alberta still needs to decide what it gets in return for supplying the land, power and infrastructure, and how Canadians keep a share of the productivity if AI eventually reduces the amount of human labour the economy needs.

Better future

  • Canadian compute capacity.
  • Research access.
  • Transparent rules.
  • Training and apprenticeships.
  • Public-interest uses.
  • Privacy and data protection.

Bad future

  • Foreign-owned infrastructure.
  • Local impact.
  • No Canadian access.
  • Vague jobs.
  • Surveillance and slop.
  • Public costs, private upside.

The economic question

What happens if AI needs far fewer workers?

The largest AI builders are spending enormous amounts on compute, data centres, chips, software and robotics because they expect those systems to create enormous economic value. One source of that value is efficiency: producing more goods and services with less human labour.

Taken far enough, that creates a distribution problem. Wages are still how most households receive purchasing power. If production rises while labour income falls sharply, the economy still needs a way to connect the wealth created by machines to the people expected to buy what those machines produce.

Production is only half of an economy. Somebody still has to be able to buy the output.

Sources and notes

Useful AI examples are broad and change quickly. The point of this page is not to endorse every AI product. It is to show that the technology is broader than chatbots, slop, and surveillance.