Ali S. Razavian
I am an entrepreneur and researcher in the field of Artificial Intelligence with a passion for creating and learning. I was fascinated with AI since I was a teenager and to this day, I have never stopped learning about it.
Some websites, that rank scientists based on their publications, have put me in the top 100 most influential scientists in Computer Vision and Machine Learning in the last decade.
Besides that, sometimes, I lift heavy stuff repeatedly, and often daydream of sleeping for 8 hours straight. I like to think about the implication and impacts of technology on our culture and society. I try my best to show the potential of technology to as many as I can and warn them about the side effects of it.
My research had always been focused on making machine learning, not just better, but good.
Back then, I believed that the bottleneck of machine learning is data representation and I focused on a notorious family of algorithms called Deep Learning. Or, as it used to be called, Artificial Neural Networks.
The breakthrough came around 2014 where I – together with a small group of scientists – managed to convince our colleagues that deep learning can achieve wonders.
So if you think that deep learning should be the primary choice in machine learning, you are probably indirectly quoting me.
The success of Deep Learning in academia inspired the industry to copy academic solutions in their pipelines, which lead to a great disappointment.
Deep-med was founded as a research-based startup to solve the problem of Deep Learning in the industry.
After banging my head to the wall for more than a year, I came up with the conclusion that to achieve super-human accuracy on any data stream, three distinct problems must be solved: Long-tail distribution, ambiguity, and inconsistency.
You can read more about it here
FUTURE PROJECT
Deep Learning
at Scale
Now
A welcome but unintended solution to training deep learning on any data stream for any user was that the entire pipeline could be automated.
The final problem to bring AI to the real world is not technology, but energy. The process is exponentially devouring our entire energy production, and only the right tool can alleviate the concern.
I set out to build the economy of Process in a world where energy is scarce.