Will AI era become the end of code?

25

The progress of AI makes many jobs are gradually disappearing. In the era of digital transformation, whether AI will replace programmers or create new opportunities? This is the question that will be shown in this article.

Artificial Intelligence (AI) has been there for decades but recently with the shape of digital transformation umbrella, everyone is talking about AI. This post does not discuss the fact that is every enterprise use AI in real cases but I will try to describe the digital world in the eye of a digital citizen one that lives in the world that many jobs have disappeared.

Before we are investigating on our current situation, let we go back to a report made by Carl Benedikt Frey and Michael A. Osborne in 2013: The future of employment: How SUSCEPTIBLE are jobs to computerization?  They developed a model to estimate the probability of computerization for 702 detailed occupations, using Gaussian process classifier, showing that about 47 percent of total US employment is at risk.

On January 2019, an artificial intelligence expert & venture capitalist Kai Fu Lee predicts that automation will cause major changes in the workforce. Speaking on an interview of CBS News, he said that he believe 40% of the world’s job will be replaced by robots capable of automating tasks or what consultant would like to call it the boring tasks.

An interesting report named “What jobs will be lost to AI and robotics?” by Takamitsu Sawa on The Japan Times shares the same view on the progress in computerization that make a lot of works disappear, while it’s also pointed out the relevance of increased jobs in the area that robot can’t replace human (yet).

In consultant business world, they used to describe AI in the manner of moving people to more valued work since it’s believed that AI will create millions of more jobs than it will destroy. While we can keep our mind wondering on the fact that there is so many Grab, Uber drivers in Asia and the day that every car is running in self-driving mode, let our mind think further on the programmer jobs.

Question: Can AI learn to code?

Answer in 2019: Sure, they can code better than us. Even it still takes time to build a strong AI (Artificial General Intelligence) but the fact that AI continues to beat human in some areas and consider the problem of physical limitation of human (age, disease, emotion) AI is considered better than our human in Speed, Accuracy, Less-biased and able to work 24/7.

Thanks to the growth of the internet and digital technologies, knowledge is double every month or even every 12 hours that sure exceed the limit of the human brain. It’s from the first day that scientist begins to train the computer, they have already expected the day that singularity happens.

The term singularity describes the moment when machines and technologies will advance beyond human’s abilities. And that is the time we must change ourselves.

In 2015, a Stanford Computer Science Ph.D. named Andrej Karpathy – now is the Director of AI at Tesla used Recurrent Neural Network and took Github’s Linux repository to train a model that is able to produce human-like code with syntactic errors. From 2018, much deep learning AI models are competing in the area used to belong with human developers – writing code, and they do it well. Deep coding is no longer the myth, from programming assistant BAYOU that created with the funding from US Department of Defense by Rice University researchers group to DeepCoder that writing code by looting other programmer code created by Microsoft and Cambridge University.

But why AI not (yet) replace Software programmers?

Thanks to the fact that the growth of AI also leads to the growing of demand for jobs like data engineering, data scientist & AI researcher who still needs to use code to train, build and optimize of the computer model. Now when we are waiting for the potential capability of Artificial General Intelligence, software engineering is required much more than just programming skills, while I also doubt on how AI impact to low-paid & low-skill programmers.

Technology empowers to transform our business, transform our self and we don’t have to fear. There are some benefits that AI can now help the developers to write better code, increase their productivity by leaving the time of doing repeatable boring tasks to learn and acquire new skills that boots up their professional career.

For example, GitHub in 2018 announces the used of many AI models to help easier developer tasks as well as increase their enterprise service competitive features. Now if you haven’t tried it yet, go to https://experiments.github.com/ and try semantic code search which allows you to search code by the meaning of natural language, not keyword matching or comment lookup.

An example of GitHub Semantic Code Search.

From our R&D lab, productivity improvement is also one of our key focus so we have tried to leverage the Generative Adversarial Network to explore the use of AI in migration old legacy applications to websites. If the computer can convert an image to code, the workload of migration such as desktop application to web application can share to a computer cluster instead of huge effort of reverse engineering:

We also explore our partner innovations to help our customer make the software better and faster since agility and speed is now the most matter topic of every enterprise product development. Microsoft AI team give a nice demonstration of an AI model that transforms any hands-draw design into HTML code with AI:

A demonstration of Sketch2Code from Microsoft AI Lab.

Thinking that after we did our workshop with end-user groups, our prototypes are already there and the team can start building the features and delivery in days, the idea is so exciting. While still many hard works have to be done in behind, many tasks will need to be done to make the AI accuracy rate reach production level, there are no doubt that AI will make an impact to the programming world.

To conclude, AI brings to software industry many opportunities but AI also creates the challenge that every software engineering must transform themselves to do more valued work or being replaced by computers.

References

  • CBSNews.com. (2019, Jan). Venture capitalist: AI will displace 40 percent of the world’s jobs in as soon as 15 years. Retrieved from CBS News: https://www.cbsnews.com/news/venture-capitalist-kai-fu-lee-ai-will-displace-40-percent-of-worlds-jobs-in-as-soon-as-15-years-60-minutes
  • Osborne, M. A., & Frey, C. B. (2013). THE FUTURE OF EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS TO COMPUTERISATION? Oxford.
  • Sawa, T. (2018, 10). What jobs will be lost to AI and robotics? Retrieved from The Japan Time: https://www.japantimes.co.jp/opinion/2018/10/12/commentary/japan-commentary/jobs-will-lost-ai-robotics/#.XVuwW6r7SzV

Khoa Tran – FHO.STU

Related posts: