PM GO insights: Grigory Bakunov on the future of technologies for business
The PM GO: Go Global! business and technology conference on scaling and entering new markets took place online on May 14. As a contributor to the event, Parimatch Tech CTO, and VP of Technology Grigory Bakunov presented his forecast for the development of technologies over the next 7 or 8 years.
We have prepared a summary of the main points of Gregory’s speech.
Intro: The technological revolution is yet to happen
In recent years, there’s been much talk about a technological revolution. However, if we take a closer look, it’s more of an evolution, where technology just follows the movement of business.
In the case of artificial intelligence, for example, the very first hand-written algorithms that resembled AI appeared in the 40s and 50s. After that, statistical systems based on Bayes systems emerged along with primitive machine-learning schemes, support vector machines, random trees, and so on. And based on developments in machine learning, businesses began to use the neural systems that most people today would think of as artificial intelligence.
In reality, there are no great leaps or magical technologies. Behind it, AI has a huge mathematical apparatus that was developed throughout the last century and can be seen in such works as Support Vector Machine by Vapnik and Chervonenkis and Rumelhart, Hinton and Williams’s Neural Networks.
It’s only now that business has been able to apply this mathematical apparatus.
Key changes to expect in the next 7 or 8 years
Let’s take a closer look at trends currently in the development stage that will soon become ubiquitous.
Going to the clouds
In the beginning, we built monolithic systems — huge software packages. We then realized that instead, we should be building systems that were simpler in each specific unit but more complex in general — this is how microservice architecture appeared. Initially, all microservices were hosted on our data centers and servers, but it soon became obvious that for flexibility and quick development you need to add or remove these microservices at the right time — and this is how cloud solutions appeared. But now the idea of moving everything to the clouds is no longer the front line of this trend.
We are now attempting to build systems where not even a service is stored somewhere in the cloud, but each separate feature of this service. We don’t even care where this feature is stored. What is important is that it works at the right pace.
Taking this approach, the next step is a system of separate features, each one stored in the cloud. This is where the idea of a feature store comes from: simply find a feature in the store and use it in your application.
On a global scale, the feature store doesn’t make the development any easier, but it does speeds it up. And most importantly, the results are more predictable, so we can forecast business models more accurately.
Commoditization of technology and concentration around open source
Commoditization is when modern technology that seems to be from outer space becomes generally available in 7–8 years.
To begin with, the basic idea is developed by large or small companies for internal use. It soon becomes clear that there’s a market for this technology. And so it is sold, for example, as a service. The next step is an open or commercial but publicly available solution: come → pay money → get a product of predictable quality. The downside of this concept is that all businesses have specific characteristics, which have to be taken into consideration.
And this is the point at which we take finished products and pay an established product team to improve them. We already do this — we find a team that creates an open-source product and say: “We’re ready to pay this price for you to implement this or that feature.” In most cases, we find performers, and if we don’t, we can find co-orderers (people who need the same technology with the same improvements). We share the investment with the co-orderers, and the interest generated by the open-source product leads to a general commoditization of technology.
LEGOlization — technologies that “snap together” like LEGO bricks
This concept gives hope that we won’t need so many programmers in the near future.
LEGOlization is about ready-made blocks that are compatible with each other. If the blocks are not compatible, you can hire an individual or a team who will use the ready-made blocks to create the required tool.
For example, when creating an online store with the block constructor, you no longer need the skills of a programmer. The specialist simply combines the ready-made “LEGO blocks,” which would include search, purchase, and payments — and it all works. That’s LEGOlization — creating everything with compatible ready-made bricks.
Decentralization as a way to independence
In general, the current structure of the Internet and technologies is centralized — we are almost always connected to some kind of server or cloud that we need to make everything work. Decentralization will allow developers but also users to quit their dependence on centralization.
In the early days, most Internet services and businesses within them worked on the same server, since there weren’t that many users at the time. As the Internet developed, huge services appeared, often requiring 2 or 3 servers — this was when the first groups of identical servers solving the same problem appeared. With the advent of the cloud, it looked like the server problem would fade away since the cloud was reliable enough. But in reality, over a one-year period, the largest clouds — Amazon, Google, and Microsoft — were down for more than 3 days in total, and this shows that the technology is not reliable enough. For this reason, businesses are striving to create products that are completely independent of servers.
The next step is about how most progressive messengers operate — sending data from the messenger to the recipient, not directly through the provider but through the neighboring phone. For example, when the network connection on your phone is lost for some reason but the person next to you is still connected — why not send a message through their phone? Moreover, why not send a message not via the server but straight to the recipient? Networks that can do this already exist, and they work well.
This approach is already used for most cryptocurrencies. The current growth in cryptocurrencies is largely driven by decentralization — the abandonment of a single central hub.
No people — why we’ll no longer see the people who solve our problems
Once upon a time, we used to hire people to create services for us. Then the roles of architects, programmers, DevOps, and so on were established. Next, a system appeared where people were paid for carrying out projects — this was when we found out how much we would have to pay the executors. Obviously, this is not the end of evolution.
We are gradually moving to the point where we’ll be paying an abstract person to perform a function. Paying for functions will lead to a pyramid of functions, where each function is performed by a separate person. In the pyramid, certain people will collect these functions and others will give tasks for these functions. This is a different approach.
At first, there will be a significant mismatch between the overall quality of implementation and the amount of effort expended, but at the same time, it will be good for the business. One good thing is that whoever performs a function can be anywhere, all they need is a computer and the ability to understand and perform the task. This approach allows us to cut the cost of implementation and create 35 different solutions to the same problem in order to choose the best one. The next step is to make this same function available in the store in the hope that it will be useful to someone else.
This is our future — we will stop seeing the people who solve our problems. This will not concern the architects and other specialists who design complex solutions; those at the highest level will remain. But the more we simplify the tasks that regular people have to perform, the closer we get to abandoning those people.
This is what a technological revolution in AI could easily become, and that would pose enormous challenges to humanity.
That’s all for now; thank you for your kind attention.