OptionGreeks

OptionGreeks is an educational tool to help users understand option pricing. Options are derivative instruments, which can be traded on stock markets / exchanges around the world. They are derivatives in that their values and contracts are derived from the “price” of some other financial instrument (including individual shares, bonds, commodities, exchange rates, interest rates etc., or indices of these). Options come in many variants, but this app focuses on European options (only exercisable at expiry). It covers Calls and Puts, from the buyers (long) and sellers (short) perspectives.

The calculator will calculate the theoretical price of options, given the users choice of the relevant parameters which include the spot price of the underlying instrument, the strike price of the option, the risk-free interest rate (discount rate), the volatility of the underlying spot price, the yield of the underlying instrument, and the term to maturity (or expiry) of the option. It will also calculate the values of the “Greeks”, which includes delta, gamma, theta, rho, vega, and epsilon. It also displays all of these in charts against each of the underlying parameters.

Finally, it also includes many different option strategies (a combination of one or more options, as well as the underlying asset), so that you can see the payoff profile, profit and the greeks for these combinations.

MLFotoFun

MLFotoFun uses state of the art Machine Learning models that use Convolutional Neural Networks (CNNs) to classify photos taken with your camera or from your library using one of eight state of the art models. The classification includes the two most probable descriptive labels, as well as the probability associated with each label.

The eight models include: AgeNet (that classifies the age of the human subject); GenderNet (that classifies the gender of the subject); CNN Emotions (that classifies the emotion of the person); VisualSentiment (that classifies the human subject’s sentiment as positive or negative); Food101 (that classifies the food), Oxford102 (that classifies flowers); CarRecognition (that classifies the make of car); and GoogLeNetPlaces (that classifies the category of place in the image).

This app is for entertainment purposes only, and clearly demonstrates how bad such models can be, as well as the biases that they may contain, so no offence is intended with age or gender classification. It may however also surprise you in how far image recognition has come in the five years.

HRZones

HR Zones is a simple to use heart rate training zone calculator. All you need to do is pick your date of birth, and it will calculate the five distinct heart rate zones for different types of training outcomes based on the percentage of your maximum heart rate.

I’ve included two models for the calculation of the maximum heart rate (there other much more sophisticated ways of doing this). The first, is simply 220 less your age (last birthday) for males, and 226 less you age for females. The second is essentially the same for males and females and is the same as the one for males above i.e. 220 – Age. This is the model used by Discovery Vitality for points.

It is very important to consult your general practitioner before undertaking any sort of exercise to ensure that you are medically fit to do so. This is a simple calculator, not a replacement for sound medical advice, so please treat your health with the caution and respect that it deserves.

Photo Classifier

Photo Classifier uses state of the art Machine Learning models that use Convolutional Neural Networks (CNNs) to classify photos taken with your camera or from your library. The classification includes both a descriptive label of the scene, as well as the probability associated with the label.

Six models have been included, and the app provides the output from each of the models in a fraction of a second each. You can see more detail on the top five predictions for each model, along with the probability of each prediction.

This is a simple but very powerful tool that demonstrates how far artificial intelligence and machine learning has come, especially the power of deep neural networks and specifically, convolutional neural networks. These models are all freely available from the internet under the licenses provided by the links below.

GoogLeNetPlaces: Creative Common License. More information available at http://places.csail.mit.edu

Resnet50: MIT License. More information available at https://github.com/fchollet/keras/blob/master/LICENSE

Inceptionv3: MIT License. More information available at https://github.com/fchollet/keras/blob/master/LICENSE

VGG16: Creative Commons Attribution 4.0 International(CC BY 4.0). More information available at https://creativecommons.org/licenses/by/4.0/

SqueezeNet: BSD License. More information available at https://github.com/DeepScale/SqueezeNet/blob/master/LICENSE

MobileNet: Apache License. Version 2.0 http://www.apache.org/licenses/LICENSE-2.0

StatsNoisy

This app calculates the probabilities of certain performance outcomes in investment management, based on some simplifying assumptions.

It calculates the probability of a single manage outperforming the benchmark over various time periods (from 1 month to 20 years), given a user selected manager skill level (given by the Information Ratio).

It also calculates the joint probability of none, or at least one manager, underperforming the benchmark. The user can select the skill level (equal for all the managers), the time period, and the number of managers.

It also calculates the probability of a single manager outperforming the benchmark by a certain alpha target (user selected) given a user selected tracking error, under the assumption that the manager has no skill i.e. an Information Ratio of 0.

It now includes probability density and mass functions, to help visualise the distributions of the outcomes i.e. the probabilities.

HIIT Counter for Apple TV (tvOS)

This app provides a timer for High Intensity Interval Training. It allows you to select the training time and resting time for each set, and allows you to select the sets per exercise. The timer provides both a visual countdown display of the time remaining, for each set, and an audible beeping for the last five seconds of each set. You therefore know when a set ends without even looking at the display, allowing you to focus on your training.

There are displays for the number of exercises, and sets, as well as the total time of exercising. There is a start and pause button, and you can change the training time and resting time while exercising, allowing you to mix and match exercise and resting times.

HIIT Counter for iPhone and iPad (iOS)

This app provides a timer for High Intensity Interval Training. It allows you to select the training time and resting time for each set, and allows you to select the sets per exercise. The timer provides both a visual countdown display of the time remaining, for each set, and an audible beeping for the last five seconds of each set. You therefore know when a set ends without even looking at the display, allowing you to focus on your training.

There are displays for the number of exercises, and sets, as well as the total time of exercising. There is a start and pause button, and you can change the training time and resting time while exercising, allowing you to mix and match exercise and resting times.

  

 

ML Camera for iPhone and iPad (iOS)

This Artificial Intelligence Camera uses Machine Learning models to classify images real time from your iPhone camera. You don’t need to take a photo, you just point your iPhone camera at any object or scene, and the classification happens instantaneously. I’ve provided five different machine learning models available for iOS, including: GoogLeNet Places (for classifications of scenes), Inceptionv3, VGG16, Resnet50, and SqueezeNet (all trained to classify objects across 1000 categories). In addition to classifying things, it also provides the probability of the classification label given.

This is a simple but very powerful tool that demonstrates how far artificial intelligence and machine learning has come, especially the power of deep neural networks and specifically, convolutional neural networks.These models are all freely available from the internet under the licenses provided by the links below.

GoogLeNetPlaces: Creative Common License. More information available at http://places.csail.mit.edu

Resnet50: MIT License. More information available at https://github.com/fchollet/keras/blob/master/LICENSE

Inceptionv3: MIT License. More information available at https://github.com/fchollet/keras/blob/master/LICENSE

VGG16: Creative Commons Attribution 4.0 International(CC BY 4.0). More information available at https://creativecommons.org/licenses/by/4.0/

SqueezeNet: BSD License. More information available at https://github.com/DeepScale/SqueezeNet/blob/master/LICENSE

Beyond 2020 (Artificial Intelligence)

STANLIB MULTI-MANAGER MINDSET – Q3 2017

BY JOAO FRASCO, CHIEF INVESTMENT OFFICER, STANLIB MULTI-MANAGER

One of my favourite TV shows growing up was an Australian TV show called “Towards 2000”, which later changed to “Beyond 2000” for what I hope are obvious reasons.

The program explored how companies and other institutions (e.g. research labs at universities) were developing technologies at the “bleeding edge” (a play on “leading edge” that is so far on the edge that it is invariably traumatic). Many people may not remember the pain of dialling up to the Internet at a baud rate of 300 kbps and waiting endlessly for a very small graphic to download. Most people probably don’t even remember the internet before the World Wide Web (www) and hypertext transfer protocol (http). While many of the technologies explored in the TV show didn’t make it into our everyday lives, most were surpassed. Many futurists will recognise that technology (or the impact on our environment) is impossible to predict, and the Internet is a great example of this. What’s next on the horizon?

It’s a brave new world

As cliché as this may sound, this time could be very different, perhaps for reasons that are not that obvious. Whenever facing the unknown, it is normal to react with fear and apprehension. Some of the brave in these encounters, live to tell the story, while others perish in the annals of history. Historically though, we’ve at least come to know the “unknown”, even if it took some time to discover. This time however, we may nd ourselves in the “dark” forever.

While most futurists think the “singularity” is a long time off (and some think it will never happen), there are some that think it is around the next corner. The “singularity” (a term hi-jacked from astronomy) refers to the point in time where humans will have developed a piece of technology (deliberately generic) that surpasses

human level general intelligence, creating an intelligence explosion where machines become super intelligent, leaving the human race forever behind.

In this article, I will talk to some of what is happening in tech that is noteworthy (in my opinion), and introduce some of the topics that my fellow colleagues will explore in more detail.

Technology in general

Everyone understands the importance of technology within our personal lives and our economy more generally, but few may appreciate just how big and important it has become. Globally, Apple Inc. became the biggest company in the world (by market capitalisation) a couple of years ago, and recently became the rst company to breach a market cap of $800 billion. Alphabet Inc. (holding company of Google) is not far behind at $682 billion, and is the second biggest company in the world. In case you were wondering, the next three biggest companies are also technology companies, namely Microsoft at $544 billion, Amazon at $476 billion, and Facebook at $442 billion (all numbers at 30 May 2017).

Closer to home, Naspers became the biggest share on the Johannesburg Securities Exchange (JSE) a couple of years ago, and has continued to dominate our market, all on the back of its holding in Tencent (itself in the top 10 largest companies in the world by market cap), another technology company. Many professional money managers are sceptical of technology company valuations, as they often trade on terrible value metrics (like the Price to Earnings (PE) ratio). Some money managers stay away because of the dot.com bubble where spectacular losses were made by companies that failed to monetise their value proposition. Unfortunately, these stocks can no longer be ignored. Warren Buffet (the sage of Omaha and CE of Berkshire Hathaway) and Charlie Munger apologised to investors at their latest AGM for a couple of their technology company blunders, including not investing in Google. They have recently invested in Apple, having stayed away from tech stocks for a long time.

Artificial intelligence

Technology by itself is not new (albeit interesting), as we have lived with tech for many years. What is far more interesting, is how the technology has begun transforming our lives and economy in ways that are different from any other time in history. Here, I’m talking about technology in its broadest sense i.e. the application of scienti c knowledge for practical purposes. We are making tremendous progress across many disciplines, and technology is the key enabler in all of them. I want to focus on arti cial intelligence (AI) generally, and speci cally a branch of AI called machine learning.

Like many technologies that have come before it, machine learning adoption is undergoing massive acceleration because of the intersection of three big “pieces” (the proverbial “perfect storm”). The rst is computing power (both in its magnitude and its cost) on the back of Moore’s law, which shows no sign of slowing down. We have unprecedented access to more computing power at lower costs than ever before. To make this “real”, my desktop PC runs at about 100 GFLOPS (100 000 000 000 oating point operations per second), where the Cray-2 supercomputer (from the 80’s) ran at about 2 GFLOPS (50 times slower for the world’s fastest supercomputer at the time).

The second is “big data”, or access to unprecedented amounts of data (both structured and unstructured). To make this “real”, and I am sure that you will look this up as the numbers are unbelievable, by some estimates, we now produce more data in two days, than we did in the entire history of humanity up until 2003 (and this is an old estimate dating back to 2010). Another estimate

by IBM, has us producing 2.5 quintillion bytes of data (that’s 18 zeros) every day, or 90% of the world’s data was created in the last two years. The Large Hadron Collider in CERN alone, produces about 25 petabytes of data every year (that’s fteen zeros, or 25000 1 terabyte hard drives every year).

The third and nal piece of the puzzle relates to the advances in the state of the art, which includes everything from brain power (the advances that have been made by individuals and groups), to the software that utilises the computing power and data referred to above (speci cally “deep learning” enabled by deep neural networks). A lot of this knowledge and software has been made available to the masses (democratised), allowing individuals and small enterprises to leverage the advances being made. Let’s brie y explore some of the things that this technology has enabled.

Applications of machine learning

Back in 1997, IBM’s supercomputer “Deep Blue” beat the reigning world champion (Gary Kasparov) at Chess by a score of 3.5 – 2.5. While many saw this coming given the progress that had been made over the years, and the limited number of moves in a chess game (albeit a very large number), it was still seen as a major milestone. Most people thought that the same feat would not be achieved for a very long time for the game “Go”, as the number of possibilities in this game is larger than the number of atoms in the universe, making a “brute force” solution impossible (which is essentially how the feat was achieved in chess, with Deep Blue able to calculate many move possibilities before every move). It was therefore a big surprise when AlphaGo (an AI computer program developed by Alphabet’s Google Deepmind) beat 9-dan professional, Lee Sedol, in 2016. More recently, AlphaGo beat the world number one ranked player, Ke Jie. In 2011, IBM Watson (another AI program able to process questions in natural language), beat two former winners of the quiz show “Jeopardy”.

Nowadays, we all have access to AI on our smart phones (Siri on Apple’s iPhone, and Google assistant on various platforms), where we too can pose questions in natural language. This is enabled through two technologies that have made signi cant progress in the last decade, all based on machine learning. The rst is speech recognition, and the second is natural language processing. Again, this has been made possible because of computing power and huge amounts of data. Another application that some of you may have tried, is translation (between languages), which can now be done in “real time” by pointing your smart phone at any written text and seeing the translation in place (using another technology called augmented reality).

At Google’s recent I/O (developer conference) the CEO, Sundar Pichai, announced that Google was changing from a “Mobile-First” company to an “AI-First” company, before introducing a slew of technologies and initiatives in support of this, including support for cloud machine learning using their TPUs (Tensor Processing Units – chips developed speci cally for machine learning).

Some of these applications may appear silly or gimmicky, but there are far more serious applications being developed. Machine learning in image recognition has taken even bigger strides in the last decade and is being used in healthcare for diagnosis. The “killer app” for image recognition from machine learning has to be self-driving cars. Various companies are developing this technology, including Tesla (founded by none other than South African born, Elon Musk). What makes their solution particularly interesting (compared to Google say), is they are achieving this without LIDAR (the equivalent of laser radar – or the spinning cylinder you see on top of Google’s cars). Tesla uses eight video cameras and one radar (and ultrasonic devices), and of course a supercomputer powering the AI.

Investments?

Most of the topics discussed above make news headlines and are therefore known to most people. What most people may however not be aware of, is how AI and machine learning is being used in investing. This is however not something new, although it will have bene tted more recently from the same technologies already discussed. The ash crash of 2009

which resulted from algorithmic trading performed by computers looking to exploit pricing inef ciencies (resulting from information inef ciencies because of their ability to gather trading information on order books on various exchanges before their human counterparts), is very much a part of the AI revolution.

Unlike in other elds already discussed above though, when it comes to investing very little is known about who is doing what, for obvious reasons (why tell the world about how you are exploiting technology to make excess returns, or worse still, share the technology and knowledge). So why does this interest us, and why should it interest you? Let’s explore a couple of the main reasons.

The rst reason, is that we need to understand the world we live in, and the markets and companies that compete (do business) in that world. We need to understand the opportunities for investing, so that we invest in companies that not only provide great short-term returns, but will also survive longer- term and hence keep generating positive returns. If we understand the themes and transformative (or disruptive) power of technologies, we may be able to identify themes for longer term investing. We are entrusted by our clients to do all of this on their behalf, and cannot shirk the responsibility by simply passing the money to our underlying managers.

This leads us to the second reason. If we allocate capital (our clients’ hard earned money) to these markets through asset managers, we need to ensure that our managers also understand this world. It is pointless to allocate the money to people that don’t understand technology and therefore will never invest in technology companies (famous words from Warren Buffet). Some managers don’t invest in growth companies as they believe that they are invariably over- valued, again possibly missing great opportunities. We need to be able to assess whether managers have the skills and knowledge to exploit as many opportunities as possible as they arise, and not merely invest in “old world” companies.

The third and nal reason, is that we may ourselves be able to leverage these technologies in the work we do. We design and build solutions for many different clients, and machine learning may assist us with many parts of our investment process (from design to asset allocation, and from manager selection to portfolio construction). It is often the jobs / professions that we think are most secure from being automated by intelligent agents that will be the rst to go in the AI revolution. We are looking at all of these and many more applications (including robo tools powered by machine learning for advisers) to understand whether there are opportunities to meet tomorrow’s demands utilising these and other new technologies.

Conclusion

I think this is an exciting time to be alive and working in investments, but also believe that these technological disruptions will be very painful for many people and businesses. Despite these disruptions, we attempt to

have our eyes focussed on the horizon, with our feet rmly planted on the ground beneath us, thus ensuring that we make great investment decisions today for our clients for many years to come.

Technology broadly is changing our world, and AI/ machine learning is leading the charge. We can either embrace it with both arms or risk getting left behind.

In the articles in this edition, we will explore the topic of technology in investments in more detail by digging a little deeper into some of the things discussed above. In the accompanying podcast, I’ll go further down the “rabbit hole” and discuss the world of AI and machine learning in more detail, and how companies like Apple, Google and Facebook are investing heavily in these technologies in an attempt to capture the “hearts and minds” of their huge client / “follower” base.

I hope you enjoy and learn something new.