About me

Oliver Goldenberg

Studying Computer Science BSc @ ETH Zürich

Currently
Project lead, Insolvenzstart A digital insolvency filing for the self-employed
Experience
Digitalisation and hardware integration AI integration for companies
Languages
English German

Handwriting recognition

A trained convolutional neural network reads 28 by 28 pixel letters: 16 and 32 convolutional feature channels, a dense layer with 64 units, and 26 letter outputs. The browser replays precomputed activations.

The human was the first model.

  1. 1490 Leonardo measures the human.
  2. 1958 The perceptron learns from examples.
  3. 1998 A network learns to read our handwriting.
  4. 2012 Deep networks learn to see.
  5. 2017 Attention changes how machines read language.
  6. 2022 Language becomes the interface.
  7. 2026 Agents plan, use tools, and do the work.

468 years to a machine that learns. 64 more to one we can talk to. Four to one that works on its own. What comes next?

The pace of progress in AI is staggering.

My vision is a future where AI moves beyond the screen and reshapes the physical world for the better, with humanity at its centre.

That is the future I get up every day to help build.