When Brains And Computers Become One
"Brain-in-a-Box" technology underway to foster cognitive computing.
"Brain-in-a-box". This term may sound like science fiction, but it really is a quest to change computing as we know it.
Our world is becoming ever more complex and data-rich -- ironically, it is information-poor. Today, the society is one big massive feed, with signals flowing in from cameras, microphones, and other sensors that stream real-time, noisy, parallel, spatiotemporal, and multimodal information. We have more data at our fingertips than ever before, but we are also more overwhelmed. Processing this flood of real-time data using traditional computers would simply be too slow and too costly, in terms of power consumption.
Computer vs Brain
Traditional computers were designed for sequential and centralised processing in accordance to a pre-defined programme. Although traditional computers are precise number crunchers, they tend to consume a lot of energy, and are very inefficient for brain-like pattern recognition.
In contrast, the brain, which operates comparatively slowly and at low precision, excels at tasks such as recognising, interpreting, and acting upon patterns. Plus, the human brain does all that while consuming the same amount of power as a 20W light bulb and fitting into the same volume as a two-litre bottle of soda.
That is why we are turning to the brain for inspiration.
The Quest To Building A Brain-In-A-Box
Inspired by the brain's ability to interpret, act upon, and learn from massive amounts of complex, ever-changing data from various sources in an extremely energy-efficient way, we aim to bring together recent advances in neuroscience, supercomputing and nanotechnology.
Our ultimate goal is to develop a system of chips that allows a hundred trillion synapses (electric signals that move from one nerve cell to another), occupies less than two litres of space, and uses only one kilowatt of power.
R-evolution in Financial Analytics
The American Century Investments (ACI) quantitative research team introduces a new open source investment platform to simplify the process of analysing the investment worthiness of companies.
The use of analytics in financial firms for research, operational optimisation, and risk management is a well-known and understood practice. ACI is utilising network/graph analysis to predict how information travels from one company/industry to the next, and to identify companies that are going to perform well in their portfolio strategies. In launching this new package, ACI walked away from the traditional, commercial analytics vendors, and embraced the open source R language together with the Revolution R Enterprise analytics software from open source specialist Revolution Analytics. The result is a new collaborative, quantitative investment platform.
Putting the r in ACI
"We began observing how much more quickly the open source community (in R and Python, specifically) was outpacing the analytical capacity of our commercial vendors. In an information and idea-driven business in which proprietary analytics are absolutely critical, we can't afford to wait for commercial vendors to play catch-up," says Tal Sansani, quantitative analyst and portfolio manager at ACI. Sansani and his colleague, Sampath Thummati, worked together to build their own package called rACI to serve their large research team, as well as a growing number of investment processes. Their focus on "open" infrastructure/analytics guided their choice as to which technology and data vendors they would work with to scale the package and their platform's capabilities.
Analysis on a Higher Level
"Our new infrastructure allows us to explore new and innovative ways to integrate "non-traditional" data sets with more common fundamental, macroeconomic analysis," said Sansani. "We expect economic network analytics, textual analysis, and data from other financial markets to better inform our stock selection models and analytics in the future." While the introduction of new, non-traditional data sets put the system to greater risks, the ACI team credits their new platform in providing more rigorous and sophisticated data quality controls.
A model to get the right fit
rACI has made it easier to introduce updated data sets to make the application of graph theory (via the "igraph" package for R) meaningful to their core stock selection models.