Providing AI Startups With Computing Power

The recent growth in artificial intelligence has largely been driven by two things: a tremendous increase in the amount of data upon which to train algorithms, and an increase in computing power to do that computational work.

It’s a field that has attracted the heavy weights of the industry.  For instance, last year saw supercomputing giant Cray Inc. undertake a deep learning collaboration with Microsoft and the Swiss National Supercomputing Centre.  The project aimed to improve the ability of companies to run deep learning algorithms at scale.

The partnership worked to leverage their collective computing expertise to scale up the Microsoft Cognitive Toolkit onto a Cray XC50 supercomputer.

The aim is to speed up the training process, and thus obtain results in hours that would typically take weeks, or even months.  This increased speed opens up a raft of new possibilities for customers, not only to solve existing problems more efficiently, but to start tackling challenges that were previously too difficult to perform computationally.

Affordable computing

That kind of computing power is out of the reach of many however, and especially startups looking to make their way in the industry.  A recent report by the UK’s Digital Catapult Centre found that around 50% of AI startups in the UK are currently constrained by a lack of computing power.  Indeed, the report revealed that a single training run for a machine learning application can cost in excess of £10,000.

To rectify matters, the Centre has created a ‘Machine Intelligence Garage’ to try and provide startups with affordable computing.  Startups utilizing the facility will gain access to computing power from the likes of Amazon Web Services (AWS), the Google Cloud Platform, Nvidia, Graphcore, STFC Hartree Centre, EPCC, Newcastle University, the Alan Turing Institute, Bart’s Health Trust and Capital Enterprise.

Access to the right kind of computational grunt can be a real leveller in allowing startups to compete with the big players.  Of course, they still need to obtain access to the huge datasets required for training, but this is one less hurdle for them to overcome.  Contact the Digital Catapult to find out how you can apply to the program.

Related

Facebooktwitterredditpinterestlinkedinmail