Blog

I’ve now begun my 1000 mile journey to get familiar with Artificial Intelligence. There are so many aspects of AI: Natural Language Processing (NLP), Computer Vision, Robotics, Machine Learning and so on. I can see my journey taking me a good number of years. I have to familiarise myself with High School and University Math and Stats.

Here is what I’ve gotten on Machine learning (ML) so far. It entails a three-step process: Data > Model > Action. You put in training data into your model, and when it is sufficiently “taught”, you put it into action with real-world data.

Machine Learning is distinct from Data Mining, though they both have similarities, such as using inferential methods to draw conclusions. Data mining is about cleaning up large data sets to find patterns. It is much more of a manual process. Machine learning tries to uncover the patterns in new data through experience derived from previous data (the training set). Once there is confidence that the model will (almost) accurately find the patterns in any data set, you release it out in the wild.

A book I am going through is Machine Learning For Absolute Beginners: A Plain English Introduction (Second Edition). I’ve gone through 2 chapters and can recommend it as you start your journey.

A lot can be said about a person based on what she shops. We can infer details like age, gender, income, whether single or married, whether having children or not. We probably do this passively all the time as we watch the attendant at the till work on the customer ahead of us in line.

A great way of tracking this is through a loyalty points scheme. Many of us believe that the supermarket is out to reward us for being loyal customers. Actually it can do a lot with that information including targeted marketing. If you pay with cash at the till the supermarket may not be able to catch you on your data.

We’ve had one or two supermarket chains in doldrums recently here in my country. And they are sitting on quite some data. A comeback strategy can utilize this data to get back on top of the game, faster than organic growth. Alternatively a division can be spun to do something with that information, it can be used by yet other supermarket chains or advice suppliers of products.

Did you carefully read the Terms and Conditions before accepting that loyalty card? Show me your shopping cart and I’ll tell you something about yourself.

I was amused to read that scientists looking for extra-terrestrial life are missing computer hardware. One team of radio astronomers said that the cost of Graphics processing units (GPU) went up and ate their contingency budget. GPUs are in short supply as crypto-currency enthusiasts put all their horsepower into mining. GPUs are built for vector processing.

It’s safe to say that the cost of mining Bitcoin today with a home based workstation is very unlikely to pay off – the cost of electricity would outweigh the rewards of mining, even in countries with low power costs. I suspect DIY miners are into Litecoin and Ethereum.

Nvidia said that retailers should make arrangements to make sure gamers’ demands were met. Can the gamers out there tell me whether the game consoles (Xbox, PS4) have increased in price? Is anyone using his game machine in its spare time to mine coins?

I’ve fancied a strategy for DIY folk to use the cloud for their work, for example Amazon AWS provides EC2 compute-optimized GPU instances. Of course you have to do your calculations to make sure your cloud costs are less than your coins.

You can read the news article here.

A lot of us who live in apartments are unaware of the gym right at the stairs. No need to join any other gym out there.

Just purchase a pair of dumbbells and walk up and down the stairs. You can do various exercises when you either reach the top or the bottom.  

Of course make sure you don’t feel faint as you carry your luggage as falling can lead to serious injury and property damage. I can recommend bending to make sure blood rushes to your head if you should be faint. You can pause midway and catch your breath.

The best thing about our indoor gym is it’s available 24-7, and no worries about personal security at night. Of course don’t hog food right when you get back to your apartment, start with fruits or veggies 😀

Reactive Systems are Responsive, Resilient, Elastic and Message Driven. They have the appeal to be decoupled, scalable and handle error gracefully.

Responsiveness is the ability to rapidly and predictably respond. Resilient systems remain responsive even in the face of a failure. Recovery is possible without compromising the entire system – just like the human body can heal itself. Elastic systems can scale resource usage up or down depending on the workload given to the system.

Asynchronous message passing defines Reactive Systems as Message Driven. Loose coupling, isolation and location transparency is achieved with these messages.

Read more about the Reactive Manifesto on https://www.reactivemanifesto.org.

So here in Kenya we recently got news of a bet win of KES 221,301,602 – (approximately USD $ 2.14 million). I heard this is the largest ever in Sub Saharan Africa. I can picture many things I can do with it 😀

So let us see how we can have a guaranteed way of winning this amount.

The bet rules are as follows. There are 17 soccer games to be played. Each game can have 3 outcomes: Home team wins, a Draw or Away team wins.

So the total number of potential outcomes is 3^17 or 129,140,163 (~ 1.29 E8)

Assuming it costs KES 100 to place a bet slip (one combination of the 17 games). The total amount for playing all the combinations then becomes KES 1.29 E10 (KES 12.9 billion)

So there you go folks. For an investment of KES 12.9 billion, you have very high chances of winning KES 221 million jackpot! Note I omitted SMS/data costs and labour costs to place the bets.

P.S: I don’t encourage betting as a means to get rich quickly. We have to make an honest living.

Let me define here what flavour of “mobile money” that I am talking about. This is stored digital value that is held in the cloud (not on the device) – a digital wallet. Funds can be transferred peer-to-peer, peer-to-merchant and even across borders. When it moves beyond our Kenyan borders, it typically lands in the wallet of another money transfer operator e.g. MNO, but it may also go to a cash-out kiosk.

The elephant in the room is MPesa, run by Safaricom, our leading mobile network operator. “Pesa” is money in Swahili. It is estimated that about 80% of Kenya’s GDP flows through MPesa. I typically have more funds in my MPesa account than in my physical wallet. It is convenient to use as I can transfer funds to individuals, pay for goods and services and even credit my bank account.

MPesa is not the only mobile money that exists. There are others such as Airtel Money run by Airtel and Equitel which is from Equity Bank.

So different postulations exist on why mobile money took off here with a bang. Some say it is because the post election violence of 2007-2008 created a problem in fund circulation (Mpesa was publicly launched approximately at the beginning of 2007). Others say it is because of the dominance of Safaricom as an MNO – they commanded probably over 90% of voice traffic in the country at some point.

One of the reasons I see is that there was a lack of a convenient means to move funds around. We did have the traditional Western Union and others in place though. But these were not affordable for everyone. Moreover they had outlets in fixed locations. Internet banking was unheard of, and Android and iPhones were yet to become widespread.

The lucky thing is that for a long time, MPesa was unregulated. At some point it raised the eye brows of banks who went to challenge its legality. It had begun to compete with their products. Luckily the Central Bank of Kenya Governor allowed for it to continue, with a new framework of regulation. Strictly speaking it does not puport to offer banking services, for example gain interest through deposits. This can only be done in conjunction with a bank.

Actually till now you can say it’s the unofficial money switch in Kenya. It remains to be seen whether the newly launched Integrated Payments Services Ltd (IPSL) by the Kenya Bankers Association (KBA) will have a dent on MPesa.

I can say the chief reason that the service won’t succeed as much in other African countries is because regulation has come in before innovation. I was seated at an office of an unnamed MNO in Nigeria in 2016 listening to their strategy on mobile money. They said that *they* wanted to be the gateway for all mobile money in the country. And the same argument is there with other money service businesses.

The government and other stakeholders have already seen the big pie in East Africa, and have put many regulations that hamper its success. It flew under the radar in Kenya for a while till regulation caught up. The infighting among the stakeholders in some other African countries has clipped the wings of a hatchling.

Let us see where matters evolve over the next 5 years. At the moment there is debate at the regulator level that MPesa is too dominant and needs to be broken off from Safaricom.

I am an aspiring data scientist.

Welcome to my blog.  Here I post what’s on my mind.