← All research storiesEngineering
An AI ‘copilot’ lets engineers steer computer-designed structures mid-design, making them stronger and easier to build
MIT’s system uses machine learning to point out which parts of an evolving design a person would likely want to change, improving strength and buildability while adding about 15 seconds.
What the scientists did
Topology optimization is a computer method that designs a part by deciding where to put material and where to leave empty space, to meet goals like strength and lightness. It has produced parts for ultralight aircraft and racing cars, but engineers usually see the result only at the end, and some designs are hard to build, such as joints where eight or more beams meet at one point.
The MIT team put a human in the loop. While the algorithm builds a design, a machine-learning model highlights the regions a person is most likely to want to change, and the engineer can make those regions thicker or thinner. To train the model, the researchers created a synthetic data set of likely human feedback on thousands of designs, covering looks, performance and how easy a design is to manufacture.
Words to know
Topology optimization
A computer method that finds the best shape for a part by adding or removing material to meet design goals.
- Lightweight brackets for aircraft
- Parts for Formula 1 race cars
- Frames for buildings or bridges
Buckling
When a slender part under squeezing force suddenly bends or folds sideways instead of just compressing.
- A soda can crumpling when you stand on it
- A thin ruler bowing when you push its ends together
Human-in-the-loop
A system where people review or adjust an AI’s work while it runs, instead of only at the end.
- An engineer adjusting AI-suggested parts of a design
- A doctor checking an AI’s flagged scans
What they found
In demonstrations, the human-guided designs were easier to manufacture and showed a 39% improvement in linear buckling load, meaning they could carry more load before buckling, compared with fully automatic designs, as reported by MIT. The human step added only about 15 seconds to the total design time.
Why it matters to you
The bridges you cross and the buildings you’re in are designed by engineers who must trust their tools. Letting people guide AI designs could bring lighter, material-saving structures into everyday construction sooner. It’s also a clear example of AI helping experts rather than replacing them.
Where this could lead
Lighter, buildable structural parts
Engineers could use guided design for building and bridge components that use less material but are still practical to make.
3–10 yearsDesign tools for students and researchers
The team hopes the approach opens this kind of design to students and researchers, not just big industry teams.
1–3 yearsThese are possibilities the research points toward, not promises. Most early findings take years of testing before they reach everyday life, and some never do.
Why scientists care
Engineers often distrust “black box” designs they can’t control. Keeping people involved builds trust and adds knowledge the algorithm lacks, which may help computer-generated designs move from aerospace and racing into everyday buildings and bridges.
Keep in mind
Results come from computer demonstrations, not real structures that were built and tested. Right now users can only make highlighted regions thicker or thinner. The model learned from synthetic, simulated feedback rather than records of real engineers. This summary is based on MIT’s news release.
How far along is the science?
- Peer-reviewedReviewed by independent experts for a journal before publication. Peer review is a quality check, not a grade of evidence strength: confidence still depends on study design, limitations, replication and the wider evidence.
- PreprintA manuscript shared publicly, often before or alongside journal review. Check its separate journal and review status (each story says what we know). Results may change.
- Conference talkPresented to other scientists at a meeting. Usually short and early; a full paper may come later.
- Agency reportPublished by a science agency such as NASA, NOAA or NIH. Reviewed internally, but not by an outside journal.
- Press releaseAn announcement from a university or company. Useful, but always check the research it describes.
How soon could it reach you?
- Could reach you soonCould show up in products, services or advice soon. Our rough editorial guess is about 1 to 3 years, not a forecast from the research.
- A few years outPromising, but needs more testing or engineering first. Our rough editorial guess is several years, with a lot of uncertainty.
- Long-term scienceBasic research that builds knowledge. Real-world uses, if any, are far off.
- Small stepA small update to earlier work. Useful to researchers, but not a change you would notice yet.
Every story links to its original source: the paper itself, or the conference program when no paper is out yet. Explanations are written with AI, usually from the paper’s abstract (each story says what it was written from), then checked against the source. We only use journals, recognized preprint servers, science agencies, scientific conferences and established science news outlets.