Conversational AI industry today is like software industry back in 70's
Vlad Chernyshov10 minutes read
We live in an age of computer technology when programs and apps not only help the user but can completely replace them. One example of such technology is Conversational AI.
And don't be confused by today's seeming simplicity of what conversational AI can automate: it's just the beginning of a long journey. Everything always starts small. Remember that in the 1970s, the first-ever personal computers were lurking in the shade of large and mighty mainframes. And those first PCs were a very niche thing for geeks only (who you couldn't call developers because there was no such word yet, so they were called "hobbyists", which speaks volumes).
Back then, you could already buy those first computers (without even spending a fortune), but there was almost no software for them. As the very first software companies emerged, they developed and sold implementations of programming languages. For example, in 1975, a tiny 2-person company called Microsoft released Altair BASIC that ran on the now-forgotten Altair 8800 computer based on one of the first Intel 8080 microprocessors.
Why is that? Because any hardware is useless without software (apps), and to develop software you need at least: a) a programming language and b) a development environment. Since a developer environment in its simplest form is just a text editor, and since it's of no use without a programming language, we can understand why it all began exactly with the development of programming languages (by development I don't mean the process of designing a language, but its implementation for specific hardware).
And as the number of software products for what was then called "microcomputers" skyrocketed, so did the number of developers and computer manufacturers. But then they faced another problem: you couldn't program much in a primitive text editor, without a debugger, compiler, and assembly tool. As soon as larger projects came into play, the need for an integrated development environment (IDE) became obvious. Any IDE is aimed at increasing developer productivity.
It was around the same time that another problem arose: how do you compile and run written code without changes on any computer, regardless of the manufacturer? To do this, you had to abstract programs from the underlying hardware and create a standard API that all programs would use when they needed to access a particular device (say, a monitor or a hard drive). The solution was an operating system. The OS takes control over the complexity and diversity of everything below it (hardware) and standardizes everything above (applications). This leads to reduced complexity and development time.
The number of operating systems in the 1970s seems overwhelming now (there were about a hundred of them in the early '70s). In the late 70's one of the most popular OS was CP/M, designed for Intel 8080 processors. Computers were becoming increasingly popular, but you could hardly spot them outside company offices, scientific labs, and military facilities.
Things changed in 1981, when IBM developed the first open, extensible computer architecture, and released the first mass-market "personal computer" called IBM PC, which you could upgrade.
What followed was a barrage of IBM PC clones that rained over the world and appeared on the desks of ordinary people, which forced specialized operating systems to evolve into general-purpose operating systems. Unlike the first-generation OS, they were no longer just a set of drivers, but in fact, determined the capabilities and appearance of apps that were developed for these OS.
In the early 1980s, one of several rival general-purpose OS was MS-DOS made by a small company called Microsoft. By 1984, MS-DOS had started to dominate the market for IBM PC-compatible computers.
As general-purpose OS emerged and people were buying more and more computers, the demand for software development rose sharply, which required a sharp increase in programmer productivity. And in 1983, Borland released one of the first development environments, Turbo Pascal, designed for CP/M and MS-DOS. It got so popular that in the first two years after its release, more than 6,000 copies were sold (which was a lot for 1983).
At this point, you might ask: "why are you telling me all this?" You see, the invention of technology (microprocessors) brought about the emergence of a whole new industry (personal computers), which began with developing programming languages. And as it grew, it brought up the issue of standardization, which was addressed by developing mass-market operating systems (MS-DOS) and tools that accelerate development (IDE).
Let me give you this analogy. Around 2014, deep neural networks appeared (a technology), which led to the emergence of a whole new industry (machine learning and artificial intelligence in general, and conversational AI in particular), and now, in 2020 it’s necessary to standardize and create a conversational AI operating system (spoiler: Dasha Platform), and then we will be facing the need to create an integrated development environment designed to dramatically increase the productivity of creating conversational AI apps based on this OS when going into production.