OpenAI's Chat-GPT 4o has arrived. Apple is working towards a on-device AI. Google is pushing Gemini to its limits. Meta has come out with Llama 3.
There are absolutely killer startups like Perplexity that are redefining the AI startup space.
So, one can only wonder, why has Artificial Intelligence suddenly become so mainstream? Why is everyone hopping on the AI hype train? And why now?
What's Deep Learning?
This isn't a Deep Learning explanation blog, so I'll keep it short.
Deep Learning is the technique of performing Machine Learning on a dataset using multiple layers of Neural Networks.
Is Deep Learning New?
Well, it's not 2010's new, and it's not as old as Newtonian Physics. It tracks its origin from the 1950s.
And having studied Newtonian Physics all the time at high school, I consider Deep Learning relatively new!
And that's why some of the pioneers of Deep Learning like John Hopfield, Geoffrey Hinton, Yann LeCun, Ilya Sutskever, and Andrew Ng are still around. (there are many more!)
That's like having giants like Newton, Tesla, Einstein, Archemedes and many more for Physics still around in terms of Deep Learning, and that's crazy to me!
Why Is Deep Learning Taking OFF?
As said by Andrew Ng in his Neural Networks course, Scale Drives The Deep Learning Progress.
Here are the four reasons why Deep Learning is becoming increasingly relevant today.
Increasing availability of Data due to rapid digitisation of society.
Significant improvement in computation speeds and efficiency by breakthroughs in hardware by companies like NVIDIA.
Constant Algorithmic breakthroughs by researchers all over the world.
Superb Marketing; Pioneers like prof. Andrew Ng coming up with courses on Deep Learning and people wanting to upskill and get up to speed with the rapid developments in the AI space.
Moreover, Neural Networks are very versatile in nature.
They can work well for both supervised (Price Prediction, Weather Prediction, Image Classification etc) and unsupervised (Recommender Systems, Robotics, Self-Driving cars etc) tasks.
And, since Deep Learning is entirely based on Neural Networks, its application is becoming ever increasing with people having more digital footprint than ever.
While deep learning may seem all mathematical and software, it heavily relies on hardware for computational efficiency.
Deep Learning is an iterative process.
It involves constantly experimenting with your data to see what works and what doesn't.
And to do that, the hardware: Graphics Processing Unit (GPU) and Central Processing Unit (CPU), must be very efficient and handle multiple tasks at once.
Thanks for reading the blog! Giving credit where it's due, the inspiration for this blog was taken from DeepLearning.AI course on Neural Networks and Deep Learning.
I hope you have a better idea about why Deep Learning is taking off now!