Learn from the Harvard Professor & Google Engineers
In this blog, I will share my learning from him
How many of you have said, OK, Google?
What happens when you say that?
The machine wakes up and it knows to respond to you.
OK, fine.
You might have said, hey, Siri. Same thing, the machine wakes up and responds back to you.
All right, all right, fine. Perhaps you have said Alexa.
No matter what you have said, they all represent an important application of what's in TinyML.
It's called keyword spotting or hotword detection.
Now, the key difference is that while they resemble the application of TinyML, these devices are largely plugged into the wall, or they require charging. I'm talking about putting TinyML onto these tiny, tiny little processors that are on this board and then forgetting about them.
Let's talk about autonomous drones, drones that are completely independent, that operate fully by themselves.
Imagine autonomous drones that are able to do video surveillance in order to keep us safe by figuring out what's a threat, what's normal, and so forth.
For instance, later on, we'll talk about how
drones are being used in the Amazon forest for preserving our wildlife. So for instance, when the forests are deforested,
we actually can go in and autonomously plant new seeds
so that we can keep the forest growing.
Imagine a situation where you go out into the mall, or marketplace,
or you're at a party. And how many times have you had to say, I'm sorry,
could you repeat that again, right? Imagine if these glasses had contextual hearing.
Or let's talk about what Elon Musk is doing,
where he's implanting chips inside the brain so that they can send
signals to understand brain activity.
By measuring the electrical signals that are emitted by neurons,
we might be able to learn things about our brain.
And in doing so, we might be able to cure diseases like treat depression, insomnia, or any other disease.
To introduce you to a variety of these different kinds of applications
where the fundamental technology that sits in the palm of my hand,
for instance, has the ability to actually have impact at a global level.
So let me be specific and give you an example:
Here, I'm talking about an ElephantEdge project.
There's a growing crisis in Africa right now
about diminishing elephant population.
So organizations are coming together in order to protect the gentle beast.
Now, if we're not careful, these elephants can be wiped out in 10 years, so it's a very important thing.
So how do we help take care of the gentle giants?
Well, we need data. We need lots of data.
So what if we could build some sort of intelligent device
like a smart collar that goes on these elephants that
is totally non-intrusive, does not harm them, but is there,
constantly sensing interesting signals so that we
can help them have a better life?
For instance, if we had a smart collar, we
might be able to do risk monitoring from poaching.
What if we could build machine learning models that
could be put onto small, little devices where
they were able to figure out if, for instance, the elephants were moving
into a high-risk area and then be able to send real-time notification to park
rangers to keep them safe?
The park rangers could then react to them
and to make sure that the elephants go away from that danger zone.
Or perhaps, we can do human conflict monitoring.
We could build intelligent machine learning
models that are able to sense and alert when elephants are heading into an area
where, for instance, farmers live.
So that they can proactively get out of the way
or do something in order to make sure that they're safe
and the animals are safe.
We might be able to do activity monitoring.
We might be able to learn things about the animals.
When are they swimming, drinking water, or sleeping, and so forth?
And by learning all these things, we might
be able to take better care of them.
We might also be able to do communication monitoring.
What we mean by that is that by sensing sound,
we might be able to understand how these animals are actually
communicating together.
But in order to do all of that, we need TinyML.
We need to be right there where the data is, where that information, those sounds, those signals, where all of that is, so that we can do real-time machine
learning.
These are all remarkable applications of TinyML.
So, that's it for this blog. Let's learn more in another blogs.
Top comments (0)