Tech Tools for the Future: Zebras, AI, and Girls in ICT Day - Seite 3

Accessible AI and ML learning

Q: AI/Machine learning (ML) has been a subject of academic study for more than half a century. Why was last year such a milestone for this type of technology?

A: Last year it exploded, not because of the algorithm or the math, but it's about how you make that accessible.

Two things happened simultaneously. Firstly, there was a buildup of data available-with many caveats and asterisks that we're now revisiting. And secondly, modern machine learning is data hungry.

When you have the hardware to run these complex models and the data to feed it, you can start capturing the complexity of the world. But it would have been esoteric if not for this brilliant interface that allows everybody to interact with it.

And that's a huge lesson if you want to make any piece of technology useful. It's not about the technology itself, per se, it's about how you make it a partner, how you really make it accessible.

Observe. Experiment.

Q: Conservation of nature often faces complex questions about the natural world. Can AI help?

A: In Henri Poincaré's book Science and Method, he says what we now call the scientific method consists of observation and experiment. And all that a scientist needs to do is look carefully at everything.

AI doesn't fundamentally change the scientific method. It is still observation and experiment. But just like the microscope, the telescope, or genome sequencing, it expands the types of things that scientists can look at.

The fundamental thing that ML and more broadly AI approaches do is extract complex patterns and complex relationships. So, we can not only look at more things, but we can also look carefully at the complexity of the world.

The role of public data

Q: Does publicly available data help in this quest?

A: There is a lot of publicly available data from digitized biological collections, field studies, and citizen scientists. But the most untapped data by far is from social media posts. People love taking pictures of nature, sometimes unintentionally capturing trees and grass, bugs and spiders.

There's a lot of information already there but it is disconnected and disorganized, so we're not taking advantage of it. And we need AI's help to get useful insights from all of it.

Q: Can AI help discover the undiscovered?

A: If we want to discover new things about the world, we need to take a completely different computational philosophical approach and a new design framework of algorithms.