Avenga Labs Update 012

October 2021 IT update

There was a lot going on in October, this time it was hard enough for me to select the best of the news with comments to fit the 2500 words boundary.

Microsoft launched financial cloud on November 1st

Cloud is more and more ubiquitous. From my perspective as a head of Avenga Labs, in new areas, truly innovative experiments, and technology experiments, there’s not even talk about cloud computing anymore. It’s so obvious. Of course, regular businesses sometimes lag behind and want to move faster.

→ Explore Avenga Cloud services and solutions for business

What is a new interesting trend is that cloud vendors start to address specific needs for particular industries and business sectors.

Microsoft announced they will deliver sector-specific services in their cloud offering and they delivered. The first one delivered is the financial services cloud, the next two are clouds for manufacturing and non-profit organizations.

What is different about the financial cloud?

The most important thing is the compliance with IT regulations for financial services and added security. Added are financial fraud detection and handling, account protection and payment protection.

It’s worth noting that Microsoft is not the first company to offer sector focused services, Amazon launched their FinSpace financial analytics platform in May 2021.

VSCode.dev

Speaking of new, open, Microsoft it’s hard not to notice one of their best free tools (if not the most popular).

One of the most popular code editors embraced on all major OSes (Windows, MacOS, Linux) is Visual Studio Code. Even staunch Linux developers prefer to use it over Atom, Sublime, and venerable VIM and emacs.

There are regular updates, thousands of plugins, visual themes, and a lot of flexibility in the configuration of the tool.

Now Visual Studio Code is available in the browser at vscode.dev.

It looks the same, it works almost the same, you can connect to the repositories from GitHub and work on the individual files in the editor. The main difference is, of course, a local filesystem.

Not everything is working as it would have on local desktop installation, for instance kernel for Python notebooks did not work in web version at

Nevertheless it’s great to be able to use Visual Studio Code online without installation, sometimes we need to just jump in, take a look, and for that purposes it is great.

Ubuntu 21.10

To add a little balance to so-far very Microsoft-y news feed, let’s enjoy another release of Ubuntu, the most popular Linux distribution on desktop, which means many of the developers’ workstations and laptops in particular.

This release is filled with many small changes which in a result make a difference compared to the previous one. We can enjoy a refreshed installer, tweaked version of Gnome 40 (which behaves differently but that’s the Ubuntu way of adapting Gnome), new 5.13 kernel for improved hardware support, and multitude of small improvements.

On the server-side there’s an interesting neat feature to assess which services have to be restarted in case of component update, which allows administrators to decide when and what to restart.

Short term support versions are used to showcase the new incoming capabilities which will eventually be available in Long-Term Support release, but they are perfectly usable, and it’s not a beta release. Another Long Term Support release is around the corner, it will be released in April 2022, so 21.10 can be considered as a good indication of what’s to come in another LTS release.

However, users focused on stability will likely prefer to wait another few months for LTS release in 2022.

Kaggle survey of machine learning

Kaggle, the organization that doesn’t need introduction, released the results of a survey among data scientists.

For instance the most popular IDEs for data scientists are Jupyter Notebook (easy one to guess) but also Visual Studio Code, PyCharm and RStudio (naturally). From the trends perspective we can observe that Matlab is dropping fast and the new rising star is Visual Studio Code.

From the algorithms and techniques used, the number one still is linear or logistic regression (no surprises for data scientists), second is decision trees and random forests, the next Gradient boosting machines, CNN (Convolutional Neural Networks) are 4th in terms of popularity.

The result means that more predictable, controllable methods still are much more popular than more bleeding edge but less explainable techniques. It comes from the natural and growing expectation for explainability and fairness of data science project results.

From the trends perspective it’s even more interesting. The transformer networks (GPT, BERT) are the only ones seeing growth in popularity, while all other techniques are stagnant (in terms of percentage of popularity) including gradient boosting, CNNs and recurrent neural networks.

This is probably caused by the growing popularity of NLP applications.

What about machine learning frameworks?

The most popular is the most universal one which is scikit-learn, Tensorflow, xgboost, and Keras are almost equal in popularity. PyTorch is merely 60% as popular as Tensorflow or Keras, but both PyTorch and TF ecosystems gain in popularity year over year. LightGBM, another gradient boosting framework, follows PyTorch as another proof of the importance of gradient boosting techniques.

What about cloud providers in data projects?

The most popular one in the survey is Amazon Web Services, Google is second, followed by Microsoft Azure. No surprises here, I suppose.

In case of data science projects the rankings are different than in case of overall cloud providers popularity.

→ Explore Top Cloud Service Providers: A Quick Comparison

In enterprise machine learning tools in the cloud the most popular choice (55%) is… not to use cloud at all for machine learning. But when cloud is used the most popular choices are AWS Sagemaker, DataBricks and Azure ML Studio. Cloud providers have a long way to go to win over the (null) cloud vendor which means on prem machine learning servers, and we as Avenga know exactly why.

→ A look into NLP for Business – With or Without the Cloud?

There’s no data project without data, and most of the data is stored in enterprise databases. The most popular ones are MySQL, PostgreSQL and Microsoft SQL Server followed by… lack of database (so other source than enterprise database).

Facial recognition systems for payments

Speaking of AI and Machine Learning, especially in the context of people’s privacy, two news stories shocked the world.

Moscow started to use massive face recognition in their metro subway system to verify the payments of the tickets. Faces are recorded, scanned and analyzed, this is considered as an example of creeping state wide surveillance. The same technology was already deployed in multiple supermarkets.

This is advertised by authorities as convenient and as an example of great customer experience, just look at the camera and the rest is done automatically.

As usual, when people’s privacy is violated, it is explained it’s happening for their own good and protection.

→ Reframing Data and Privacy. Essentially, Data Is Good. It’s the Use Cases That Can Be Problematic

Many commenters were not surprised as Russia is not widely known as a beacon of democracy and human rights.

But… the similar news came from the United Kingdom. Nine schools in England started to use face recognition to verify childrens’ payments for school meals. “With Facial Recognition, pupils simply select their meal, look at the camera and go, making for a faster lunch service whilst removing any contact at the point of sale” explains the flyer distributed by the schools. 97% of children and parents consented to this solution. Many of them were allegedly under pressure from teachers and school authorities. The “benefit” of massive privacy violation is cutting the time it takes to verify payments by a few seconds.

And the similar systems were also deployed in the United Stated at schools, but for surveillance with childrens safety as a goal.

Massive surveillance techniques are to be banned in European Union member states according to the new AI regulations. Also, in the US, individual states have banned facial recognition. And this is the trend to follow and support.

→ Journey Towards Trustworthy AI in the European Union

Digital banking surpasses social media

Pandemic accelerating digital transformation is sometimes we just feel and keep on repeating, but when the real data comes to back it up it is much more convincing.

The latest data comes from the US (“The Fintech Effect” report by Plaid), where digital banking usage penetration grew from 58% to 88% between 2020 and 2021. There’s no going back to the old ways, there’s more percentage of mobile banking users than social media users!

Convenience, speed and safety are the main reasons for adoption among all the groups, no matter what age, ethnics etc. 88% means it’s not “just” digital natives naturally embracing the obvious digital choice, it means the entire society.

This is great news for entire global digital transformation programmes and efforts. However the consequence of the saturation of users means there’s not too many users which haven’t joined the digital revolution in banking. And it means even stiffer competition, more need for speed and even lower operational margins.

→ More about Banking Trends in 2021

Facebook Metaverse announcement

Speaking of social media, let’s focus for a moment on the most popular social network, Facebook. Its popularity is diminishing, and even drowning in the younger population, and according to the majority of the analysts it will end up as social media for “boomers”. Even attempts to gain attention back with their Instagram as a “young” alternative for the platform are not as successful as expected.

Facebook is also riddled with scandals, including privacy leaks (which I addressed before at the  Avenga update)  and recently with the insider news from whistleblowers about Facebook promoting violent and racist content to increase traffic.

I can also share my personal experience with Facebook, thay have never reacted positively (removing content etc.) for clearly racist, anti-semitic, COVID-deniers etc. content which I personally reported. It’s a publicly known “secret” of not only Facebook, but also LinkedIn and other social networks. Violent, controversial content is grabbing attention and their business models are based almost entirely on traffic.

So what is Facebook doing now? It seems to totally ignore all the problems without addressing them properly and escaping forward with a big leap forward towards a new mixed reality ecosystem called Metaverse.

The first step is to rename the company from Facebook to Meta to distance itself from their main product (Facebook). It’s kind of similar to what Google did in the past with Alphabet transition.

What is Metaverse vision about? How does it relate to enterprise IT? Is it going to be successful?

First of all it’s worth noting that the concept was envisioned tens of years ago by science fiction writers and until recently there were no technical capabilities for massive adoption.

Meta is a mixed reality concept in which physical users and objects collaborate with virtual avatars and virtual reality / augmented reality environments. It’s a blend of real world and virtual world. And not just for fun and social activities but also for work.

I’m a big fan of virtual reality in consumer space, especially during lockdowns, virtual experiences could ease the sense of isolation and travel to virtual worlds with the relatively high immersion and emotions.

For instance there are co-workers sitting at the physical and virtual desk talking to each other to reduce mental distance between physical people and remote avatars, collaboratively painting ideas on virtual whiteboards.

(screenshot from official YT video)

Shared workplaces, working environments, are already a thing in the Oculus Quest ecosystem of Facebook (now Meta Quest of Meta ecosystem). Virtual desks, virtual screens, virtual keyboards are going to be the norm in 5 to 10 years according to optimistic predictions.

Horizon Workrooms are already here for those who would like to try them.

Why it can work

• Facebook/Meta is very successful in consumer space with their Oculus Quest and Oculus Quest 2 headsets, they are the market leaders not only in the category of autonomous headsets (not requiring any cables) but even in PC VR.
• More and more games, fitness apps, experiences and also work related apps are delivered to their app ecosystems. It’s really impressive for me as Quest 2 user myself to see the speed of the progress and wide global adoption from all the major gaming and software companies.

• Meta has a lot of money to invest and they do invest, for instance 10 bln EUR in European Union to hire ten thousand specialists to build software which will deliver Metave

Why it may fail

• Meta does not own any enterprise ecosystem, for instance enterprise behemoth Microsoft is also present in mixed reality space, especially for corporate users
• Meta is not a cloud player (unlike AWS, Microsoft and Google) but it does have technical capabilities to serve billions of users
• Meta heavily depends on Google and Apple for mobile experiences
• Headsets and AR glasses are much cumbersome and inconvenient than smartphones or laptops
• Current affordable headsets such as Oculus Quest 2 are really poor experience when it comes to text reading and processing

Time will tell, as usual, the vision is bold and backed up with initall VR success plus a mountain of money, startup acquisitions and

But the deciding factor is going to be the answer to the question. Do you want your professional life to be a part of the Meta/Facebook ecosystem, do you want your voice, image, business data to be accessible by Facebook as well?

Summary

A very interesting month of October has come to an end, with many big announcements, interesting statistics that I was glad to share with you.

Let’s see what November brings this year, until next time!