90% of a programmer’s time is spent on reread and correct the code already written. Especially because most software has already been created and must be not only corrected and updated, but also translated from one language to another. This is an extremely expensive process, also because it requires programmers to be experts not only in the latest innovations in the language market, but even those from decades agoas happens in the old servers called mainframes and used since the dawn of computing, when computers occupied an entire room.
IBM’s solution? Rewrite code using artificial intelligenceamong other things by fielding the modern version of an old acquaintance: Watson.
The goal is to create a basic model that can be infinitely applied to all existing languages, revolutionizing the rewriting times of business applications. IBM therefore plays the card of Watson Xwhich promises a revolution compared to automatic systems Microsoft e Google. All of this, obviously, as part of a broader strategy that intercepts that of customers, i.e. the companies that compete on the markets every day.
The test From vaccines to the war in Ukraine: 10 questions to Bard, Google’s generative AI that now speaks Italian by Emanuele Capone 13 July 2023
Digital transformation for IBM
After all, the recipe for winning on the markets is to make targeted investments in digital skills and technologies, as he claims Stefano Rebattoni, number one of IBM in Italyaccording to which “we invest to have a clearer vision of the market, more efficiency, better quality of products and services, but also resilience and operational continuity” and “one of the keys is artificial intelligence today and quantum computing tomorrow”.
IBM is therefore preparing Watson X, the new incarnation of first, great effort to commercialize the fruits of artificial intelligence research made by IBM over the last fifty years. The new version of Watson, officially created in 2011 to answer the quizzes of Jeopardy, a well-known American television program (where he already defeated the human champions by winning a million dollars), has a different objective: to speed up the work of programmers in a critical area for digital transformation. That is, the updating of old mainframe platforms, which manage complex systems and whose transformation costs made by human programmers are prohibitive.
IBM created Watson USA and LLM (what is it?) from approximately 20 billion parameters specialized for understanding and rewriting code.
Stefano Rebattoni, CEO of IBM Italy
Watson X Code Assistant
In fact the Watson X version for Cobol is just one of a number of versions they have the very ambitious goal, also shared by Microsoft and Googleto make a large part of the programmers’ roles obsolete: Watson X Code Assistant for Z (this is the full name) is part of the Watson expected by the end of the year.
The basis is common to all variations and the goal is to have cloud services capable of understand 115 programming languages learned from 1.5 trillion tokens with 20 billion parameters. The scale of this project is extremely ambitious for IBM, which aims to transform Watson The Watson X Code Assistant product portfolio is still limited but will expand over time to support other programming languages, speed up the time needed to convert code in modernization projects and help developers keep up with all the updates needed in their business. The revolution passes precisely from the possibility of creating a Foundation Model, which can then be declined and specialized several times for different languages but also in different sectors.
The importance of foundational models
The new Watson updating and rewriting done by human developers is very difficult and expensive. The automatic system analyses, understands and translates the source code of the programs in Cobol into equivalent source code, however, written in Java, which is more modern and above all more flexible because it can be virtualized and used much more efficiently on different architectures.
For developers to be able to use Watson rewriting and optimization work, which is one of the most expensive maintenance activities for companies, brings a series of advantages in terms of time and costs. According to IBM, it is possible to rationalize costs, have precise metrics on the modernization times of the Cobol code, operate on the same IBM Z systems already purchased and, once the conversion has taken place, expand the pool of possible developers to hire (because Java is a language known by many more programmers than Cobol) and reduce the training time of new recruits. In the end, Java is more up to date to the latest, good practices in terms of code efficiency and security, compared to Cobol.
Interview Mario Rasetti: “We will never build a machine as complex as the human brain” by Bruno Ruffilli 05 September 2023
Produce applications on an industrial scale
The idea of industrializing the creation of applications thanks to an AI system reduces not only times but also costs, opening up to an audience of potential users of medium and small sizes or with very small budgets, who today are essentially prisoners of old software architectures. Furthermore, according to market analyzes conducted internally by IBM, the strategy for the next 2-3 years for large and small companies it will be to enhance internal applications rather than create new ones.
According to IBM, the future will evolve even more rapidly with the arrival of quantum computing, of which the company is among the main players on the market. The Armonk company carries out research in the sector both in the USA and in Europe and, thanks to the work of the Swiss research center led by the Italian Alessandro Curioni, who is also an IBM Fellow: “It will still take between 2 and 5 years – he told us – to have operational solutions in various sectors”.