TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Foundations · AGE 16—18 · FACTUAL · DURABLE
Promises, then winters
A student who knows the winters can tell a demonstration from a delivery.
The idea
Twice the field promised more than it could deliver, and funding and enthusiasm collapsed across much of it: the AI winters. The first is usually dated to the mid-1970s: programs that solved toy problems failed to scale, and Sir James Lighthill's report for Britain's Science Research Council (1972–73) found the grand claims unmet. The second came in the late 1980s and early 1990s, when commercial "expert systems" — hand-written rulebooks for specialists — proved costly to maintain and hit the limits of rules. Both had one shape: a real advance, a wave of promises, a gap between demonstration and use.
Why it matters
Every generation of AI arrives with claims that sound new. A student who knows the winters can tell a demonstration from a delivery, and that is the best protection against hype there is.
See it in the studio
A vendor demo renders a perfect courtyard in seconds. The winters teach the follow-up questions: does it do this on my brief, my site, my drawing standards — and what does it cost to keep it working for a year?
Watch for this
Two opposite mistakes. "This time is different, so the pattern does not apply" — people said that before each winter. And "it is all hype, another winter is coming" — but both winters followed real advances that later returned, better funded and better built.
Try it
Find three AI claims made about architecture in the last twelve months — a product page, a talk, a post. For each, write down what was demonstrated and what was promised. Check whether any of the promises can be verified today. Date your notes and re-read them in a year.
Prove it
Describe the two winters — roughly when, and what over-promise preceded each — and name one habit a designer can take from them.
How it works
Winters are not only about technology. They are about the gap between what a system does in a laboratory and what it does in a working environment, and about who pays while that gap is closed. The first winter was a crisis of scaling: search-based programs that worked on small problems hit a combinatorial explosion on real ones, where the number of possibilities grew too fast to search. The second was a crisis of maintenance: expert systems needed specialists to write and update thousands of rules, and the hardware built for them was overtaken by ordinary computers. The boom that produced today's tools came from a different approach, learning from data, which is why it has not simply repeated the earlier collapse. Whether it has a gap of its own is a question this map keeps open.
What this idea builds on
What this idea opens up
Sources
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.