Computation

Scaling intelligence

Creating ventures that unlock the potential of computational leverage
Computation is human leverage. It allows us to connect and orientate more effectively, transfer and manipulate abstract information and concepts, create beyond our imagination and magnify our manual capability in automation.

These capabilities have the potential to accelerate us towards a much better future, they also risk making things much worse. We’re focused on how to move the world to one where the full breadth of society benefits from the step change in creativity, connection and automation across health, food, energy, happiness and abundant capital. We’re starting with abstracting hardware out of the way to accelerate the pace of computational capability whilst minimising the energy and resource impact.

Thank you!

Contact

To speak with our computation team, please send us a message using the form below.

Areas of venture creation:

Universal Compute Fabric

Problem to solve

There is an ever expanding range of specialist processors from neural network focused architectures to dedicated ASIC circuits for specific data processing operations. The problem is that most developers do not have the time to optimise for any given architecture, in reality most of the world’s computational workloads are run on CPUs on over provisioned cloud resources. This is enormously costly for business and academia, $100bn on cloud compute alone last year, often millions of dollars training one model, enormous CO2 output (44 millions tonnes on AWS alone) and huge resource waste with tonnes of precious metals being thrown away each year.

Reason to solve

Moore’s law is breaking down and we have become increasingly reliant on just one or two firms that can place transistors at sufficient density but with very high geo-politcal risk. The next wave of performance will come not from denser chips but horizontal integration, minimising memory movement and optimising code on the fly to the specific array of processing architecture. This has the potential to create 10-100x speed and power improvements with any new hardware or any hassle for the developer. That’s what we’re building in Xonai.

Low Loss Compute

Problem to solve

Computing hardware selection and design is a multi-dimensional problem with complex trade-offs (performance, power, area, cost, flexibility, composability) that must be determined years before the required workload is known. The result is that the market is barbelled into flexible but relatively poor performance (CPUs) or a long tail of high performance extremely specific designs that are costly to make, generalise poorly and hard to work with.

Reason to solve

What if we could map the distribution of computational and memory requirements and quickly optimise hardware on demand and in an extremely heterogeneous fashion, extending the plethora of existing specialist hardware so that every process ran in the fastest, lowest power manner possible. That has the potential to create absolute optimal performance from existing nodes and extend performance far past the tradeoffs associated with Moore’s law.

Our companies

  • computation
  • Seed
Dissolving boundaries to accelerated computing

Opportunity

Solution

  • computation
  • Seed
Transparent noise cancellation

Opportunity

Solution

  • computation
  • Seed
Software eats engineering design

Opportunity

Solution

  • computation
  • Seed
Non-toxic adhesives powered by ML

Opportunity

Solution

Content