NLP + FaaS would be the next combo bingo
Natural Language Processing (NLP) was on the top of the 2017 buzzword list with the GAFA’s competition around the personal assistants (Siri, Cortana, Alexa and Google Assistant) and the chatbots business.
FaaS (Function as a service) — Serverless declinaison in particular — was not so far from the first one in the list. What is exactly this computing approach that doesn’t need computers ?
In some words, Serverless computing is mainly a cloud service that transform a function (an algorithm with inputs and output coded in a maintream language) in API (most of the time an HTTP API). In the cloud implementations (AWS Lambda, Azure Functions, Google Cloud Functions or Alicloud Cloud Function Computing), it’s pretty efficient to reduce compute cost at the exact processing cycle you need. In order to be efficient your function need to be extremely smart (a nano-service ?) because you need to start your application each time you need to process an event (event processing model). To build an application, you just have to chain functions and PaaS services (like an event bus or an event storage system).
Some days ago the machine learning science has passed a huge milestone with the born of AlphaGo Zéro. Some month after the emergence of AlphaGo, Alphabet has given birth to a terrific competitor (After three days of learning Go, AlphaGo Zero win AlphaGo by 100 to 0). The short period between the born of theses two brothers mean that the learning capabilities improvement have nothing to do with the miniaturisation of transistors (the declining Moore Law) : since now, robots doesn’t need us anymore to learn, they just need to eat small amount of data and compute them. With this incredible capability to learn in a very short period (with at least an incredible backend of distributed computing) the datascientists have killed the datascientists. Who will be the next victims of the AI revolution ?
When a skilled product manager has a clever idea, he could use a lot of tools to launch his business. SaaS software could help him a lot to support functions and productivity tools. But in a software and data driven world, he probably needs also software engineers (at least one) to convert his brilliant product in reality. With the combination of NLP and the FaaS technologies, a Product Manager with algorithm capabilities could quickly build MVPs without any developer … you’ve got it ? Developers will probably be the next victims.
But don’t worry, the cloud era advent was supposed to eliminate all the bearded army (sysadmin) … and the bearded guys are still there. They pivot to DevOps and Site Reliability Engineering for the best of them and bearded guys are still alive… I’m sure Developers would not die to.
However, as the end of the Moore’s law it seems AI is strongly challenging the Schumpeter’s law to … including in the IT world itself.