People are often arguing that artificial intelligence will be capable of nearly everything humans can do. Specifically, it’s not about dumb machines programmed to do very specific tasks —it’s about AIs that learn and get better by watching us and parsing our data for patterns. It will certainly replace a lot of jobs, but often you also hear that creative jobs will be spared, as AIs lack emotional intelligence and proactivity which are essential human characteristics.

But can computers be creative? – This can be a very philosophical question, but looking at what Google is doing right now, you will be very surprised. Last year, they launched an open-source research project called Magenta, which aims to explore the limits of what artificial intelligence can do with machine learning in arts. Google uses a different approach of learning compared to how classical AIs work and builds intelligence.
They developed an algorithm, which uses neural networks — a complex mathematical system that can learn tasks by analyzing large amounts of data. In recent years, it has been proven to be a very effective way of recognizing objects and faces in photos and identifying commands spoken into smartphones. Today they are using it to teach machines to synthesize new sounds, on notes generated by different instruments. So far, the experiment has yielded classical piano compositions that are actually hard to distinguish from human-composed music.

Magenta will not only produce music but will also provide musicians with a completely new range of tools to make music. One great example is the infinite drum machine, which organizes thousands of everyday sounds through machine learning and you can generate beats with it. These tools are all part of a number of initiatives Google launched, including Magenta and Creative Lab, to introduce free-to-use AI tools to the greater mass. They hope that everyday users can provide them with helpful insights on how they should build and improve it, as well as to help independent developers and musicians to create new experiences. There is definitely more to come and I’m very excited to see where these projects are heading. If you want to try out the infinite drum check out this website: https://experiments.withgoogle.com/ai/drum-machine

Hutson, M. (2017, August 08). How Google is making music with artificial intelligence. Retrieved October 14, 2017, from http://www.sciencemag.org/news/2017/08/how-google-making-music-artificial-intelligence

Metz, C. (2017, June 03). Finally, Neural Networks That Actually Work. Retrieved October 14, 2017, from https://www.wired.com/2015/04/jeff-dean/

One of the hottest areas in the finance industry these days is quantitative investing, which is using AI (artificial intelligence) to scan through huge amounts of data identifying signals that are not visible to human beings. This change will be profound. More and more investors and funds adopt the aspect of analyzing alternative datasets to get an edge against market participants, the market will react much faster and can anticipate what traditional data sources may convey (e.g. quarterly corporate earnings, macroeconomic data, etc.). At some point, traditional data sources will lose their predictive value.

But what is actually meant with alternative data?

Alternative data includes data which can be generated by individuals (social media posts, product reviews, search trends, wifi data etc.), data that is generated by business processes (company energy consumption, credit card data, commercial transaction, etc.) and data which is generated by sensors such as satellite image data, foot and car traffic and ship locations, etc.). However, this type of data is often larger in volume, velocity and variability and therefore requires a deep level of analysis before it can be used for trading.

Using alternative data gives an advantage to those willing to adapt and learn about it. New types of datasets that capture ‘Big Data’ will increasingly become standardized in the near future. It will be an ongoing battle to uncover new higher frequency datasets with even greater granularity. Together with Machine Learning techniques, it will become a standard tool for quantitative investors. More and more traditional strategies such as risk premia, trend followers, equity long-short will need to follow to stay competitive. Already today, there are big data ecosystems which involve firms specialized in collecting, aggregating and selling these new datasets. Especially data which is generated by sensors is in high demand.

This type of data can be categorized into three groups: satellite data, geolocation data, and data generated by other sensors. One of the most popular alternative data offerings is satellite imagery. 20 years ago, launching a traditional satellite cost about millions of dollars and years of preparation. Nowadays, companies (such as Planet Labs) are able to launch a fleet of nano-satellites (the size of a shoe-box) into low-earth orbits. These nano-satellites have significantly brought down the cost of satellite imagery. Image recognition is also standardized and comes with Deep Learning architectures.

The second category of sensor-generated data is geolocation data. By tracking the location of smartphones through either GPS, Wifi or mobile phone signals, these firms can determine foot traffic in and around store locations. Blix Traffic is one of those companies and collects anonymous smartphone data from customers to understand the walk-by traffic beyond the front door of retail stores and walk by conversion rates.

One example illustrates how alternative data was used to be ahead of the market. When JCPenney reported results for the second quarter, the news came as a surprise — for most investors. Hedge funds however were analyzing satellite images and could tell that traffic into the stores was rising in April and May and they traded on it. JCPenney’s shares jumped more than 10% in mid-August.

Many companies in this segment are currently focusing on tracking foot traffic in retail stores; for the future, we can definitely expect real-time data from business processes. At JPMorgan’s macro quantitative and derivatives conference on May 19, the bank surveyed 237 investors, and asked them about Big Data and Machine Learning. It found that 70% thought that the importance of these tools will gradually grow for all investors. A further 23% said they expected a revolution, with rapid changes to the investment landscape.

Sources:

Kammel (2016), Alternative Data – The Developing Trend in Financial Data. Retrieved October 10, 2017, from https://blog.quandl.com/alternative-data

Kolanovic (2017), Big Data and AI strategies – Machine Learning and Alternative Data Approach to learning – JP Morgan.

Turner (2015), This is the future of investing, and you probably can’t afford it. Retrieved October 10, 2017, http://www.businessinsider.com/hedge-funds-are-analysing-data-to-get-an-edge-2015-8?international=true&r=US&IR=T