A neural network learns the same way we do. Follow it forward.
We take in information.
Information enters the network. Each connection carries it forward, layer by layer.
We test what we've learned.
The network meets data it never trained on. We test ourselves on real problems and past projects, where being wrong shows us what to fix.
We trace our mistakes back.
Networks call this backpropagation: the error travels back through every layer that produced it, and each connection adjusts a little. It is the entire reason a network can learn, and it is how we work too.
We go again, deeper.
This is what research is. Not a moment of insight, the loop, run seriously, repeated until something true comes out.
We build something real.
The loop earns its output: projects, papers, and demos. Real work, with proof behind it.
We learn. We research. We build.
UoBD SAIL equips students to contribute meaningfully to AI, through hands-on research and a culture of intellectual curiosity.
We are a community where students learn by doing. Membership is open to every UoBD student; contributors apply to work on actual projects. We are honest about what we do not know yet, and serious about finding out.
Where things stand
- UoBD SAIL is a registered club of the UoBD Students’ Association.
- Our first project begins in September 2026.
Attention Is All You Need (Vaswani et al., 2017)
The paper that introduced the Transformer. It removed recurrence from sequence models and replaced it with attention: a mechanism that lets every token look directly at every other token. Nearly every large model today is built on this architecture.
Our first project begins in September 2026. The groundwork is happening now: reading, planning, building the team.