This is actually an interactive animation that students can play around. Like what happens if they click on this neuron? This neuron will send out an impulse that triggers other neurons. And then over time, these, as I start clicking on these connections will get sort of reinforced or not, and these connections will get sort of amplified over time. So this is how biological neural networks learn.
So if somebody has been bitten by a dog and this is done in therapy, you’re sort of really afraid of dogs. You can’t, you want to avoid them. But that’s not the way to sort of get over the fear of these, of dogs. You repeatedly expose yourself to dogs and say, well, they’re not so bad.
As time goes on, I can repeatedly give signals to these neurons and they are sending signals to each other and these connections get reinforced over time. And this is what’s called Hebbian learning.
The idea behind artificial neural networks is to really encapsulate these, but within computers and we can do lots of interesting stuff with it
In today’s day and time we hear about big data. That is, for example, in this particular case, each of these dots may be a single cell. Each cell has thousands of genes within them. There is no way we can visualize this huge amount of data. But using the magic of a machine learning algorithm, say a neural network or something called
PCA, I can squish this data down into this three dimensional plot and I can play around with this data.
So for example, this particular cell maybe is associated with a disorder. How far apart you are from each other tells you what different diseases each of these cells have.
Concept 🧩 🚀 So machine learning/AI is an intelligent way to really do interesting things with data and give you certain insights
💡 🛠️Activity on bias detection
🧩 🚀 Even though these machine learning algorithms appear fancy, they can make stupid mistakes. It is incumbent upon us to curate the data that is going into this: garbage in, garbage out. That is a big concept.
CCTV cameras are getting used for say, racial discrimination and racial profiling.
💡 Thornier issues that are coming up in machine learning. Not everything in AI is about terminators and them taking over the world.
There are multiple immediate risks that AI poses to us. Bias in AI is one of them and we can fix it.