AI is the fastest growing technological demand in history. Much more than the cloud and the advent of personal computers.

‍

AI is everywhere and penetrates exponentially every domain of our professional or private life, be it finance, healthcare, industry or social…

‍

If AI is set to change our lives to the extend, we are being told - probably for the good - it has a rapid growing cost. This will soon become prohibitive for most of the users, would they be individuals, SMEs or large corporations. The demand grows way faster than the technology (Each ChatGPT training is said to cost over 100M$).

‍

The answer provided today to increase the computing power to train and use ever larger AI models, is really to build new data centers with always more and more servers. Even when considering initiatives consisting of improving today's technology efficiency by X percent (under 100% in the best case), AI will not make it.

‍

Can we continue like this?

‍

Sure not. If we continue like this AI will soon hit a wall!

‍

Furthermore, this growth in datacenter deployment has certainly a negative impact on our planet and will increase the discrepancy between the north and the south. Can we dedicate more and more electricity to train and use models (analysts warn that by 2027/2030 AI will need the equivalent of Japan's electricity consumption), can we use more and more water to cool more and more servers, can we extract more and more rare earth to manufacture more and more chips? Surely not. It is just not sustainable.

‍

Our vision

‍

At EvoChip we saw this coming more than 3 years ago and we decided to work on a solution that would not just improve the current technology but create a solid long-term differentiator.

‍

We gave ourselves an objective: Compute AI models in a chip 1000 times faster than existing solutions. In other words, reduce the hardware (servers, GPU’s, laptops…) necessary to train and use the same AI models, with the same quality, by 1000x.

‍

Three years later, #LiquidBioScience, the largest US biomarker discovery company, can attest that we have succeeded. We have just reached our initial goal of accelerating performance in a chip – #FPGA today - We have proven by just using the software version emulating our hardware implementation (it runs faster in a chip) that our technology is a magnitude of times faster and resource efficient than current practices. We have reached our objective and beaten our own expectations.

‍

How did we do that?

‍

We started with a blank sheet of paper. No preconceived ideas, no obligations.

‍

The brain behind the technology in our team is a mathematician, a data scientist, a developer and one of the few specialists of evolutionary algorithms (the ones that are almost not taught anywhere).

‍

It was also clear, from day one, that if we wanted to make a real impact, our technology needed to use less transistors to provide the same output.

‍

If we could use 1000 times less transistors, we would consume 1000 times less energy and resources, and provide much faster chip performance.

‍

There were two major domains to work on:

‍

• Use a different class of algorithms that would not complexify exponentially the relationships between data (as Neural Network does) but remain linear as the volume of data increases.
• Mathematics as we know them today were developed for our human brain. Machines do not work as our brain. They calculate differently. We had to create a complete new set of mathematical components directly usable at the gate level.

‍

By combining the two, and more innovative techniques, EvoChip core technology was born, and we achieved our objective. This is only the beginning. There is a lot to be done with this new technology.

‍

The invention of airplanes, nuclear fission or operating systems, etc. were fundamentally disruptive technologies. They came out of the brain of a single – genius - individual or a very very small team. They considerably evolved over time, but the core IP made the difference between before and after.  It changed our lives.

‍

At EvoChip we believe we have developed such a truly disruptive, good for the planet, game-changing IP.