This year, IEEE spectrum readers were keenly interested in all things software: what’s happening in the tumultuous world of open-source, why the sheer size of code causes security vulnerabilities, and how we must take seriously the energy costs of inefficient code. The ever-growing presence of artificial intelligence has also made itself known in the world of computing, with the introduction of an LLM-based internet search tool, the search for ways to circumvent AI’s voracious appetite for data in scientific applications, and the transition from coding co-pilots to fully autonomous coders – something , which is still a work in progress.
And if you scroll down to the bottom of our list of the best PC stories of the year, you’ll find a treat in the form of IEEE spectrum original sci-fi shorts.
Andriy Onufriyenko/Getty Images
Artificial intelligence was destined to come out on top in 2024, including in the field of computing. Coding assistants like Github’s Copilot or Amazon’s CodeWhisperer are already changing the way software engineering is done. This raises an obvious concern – are AI coders coming to replace software engineers?
The short answer is, not yet. And not for lack of trying. The AI lab, Cognition, has created a fully autonomous AI software engineer named Devin AI. Devin boasts the ability to design, build and deploy a website, fix bugs in the codebase, and fine-tune LLM all by himself. Open-source alternatives to Devin soon followed. But even at the tasks they claim to solve, these coding autopilots aren’t very good yet. For example, Devin solved only 14 percent of the GitHub problems he was presented with. And the real world of software development is much more interactive and complex, with many teams working together to design, triage, and solve large-scale problems together. So this article assumes that coding assistants with a real human in the loop will be more successful, at least for now.
IEEE spectrum/Getty Images
This cult favorite’s roundup of the most popular programming languages of 2024 yielded some predictable results along with new trends.
To no one’s surprise, Python tops the charts as the most popular language at the time and among IEEE members. Employers have a slightly different preference—they give an advantage to job applicants who know SQL (pronounced “continue”), a database query language. It should be noted that knowledge of SQL alone is not enough and must be paired with a more traditional programming language such as Python or C++. But those who already know the language and want to get a head start on the job market would do well to add SQL to their CV.
In new trends, the lesser-known languages Typescript and Rust have seen significant gains since last year. Both languages have features that implicitly protect the coder against certain types of errors. Typescript requires programmers to declare the type of each variable (float, integer, boolean, or other) before using it, which increases reliability. Rust is memory safe, which means it protects the program from writing data to memory it is not meant to write to, thereby removing some vulnerabilities.
Daniel Zender
In 1995, Niklaus Wirth, a computer science pioneer best known for designing the Pascal language, wrote an article titled “A Plea for Lean Software.” In it, Wirth bemoaned the increasing size of code—the literal number of lines and the amount of space it took up in memory—which he considered both unnecessary and dangerous. After all, the more code you write, the more opportunities you have to introduce a bug or security vulnerability.
Almost thirty years later, after Wirth’s death in January 2024, lifelong technologist Bert Hubert returned to Wirth’s plea, despairing at how catastrophic the state of software excitement had become. In this renewed plea, which reads like a cry from the soul, Hubert explains how dire the situation is: The size of software has become enormous, with applications as simple as garage door openers requiring up to 50 million lines of code to implement. Coders routinely import many external libraries without really knowing what’s in them, which greatly expands the code size and introduces many potential vulnerabilities. Security breaches have become so common that many consider it dangerous to run code themselves and instead resort to software-as-a-service.
To shine a beacon of hope to other desperate software engineers in the choppy seas of huge code, Hubert wrote a sample application called Trifecta that supports online image sharing. Trifecta has minimal dependencies and clocks in at 3 megabytes of code, a fraction of the size of competing solutions. Hopefully the next thirty years will get software bloat under control.
iStock
For decades, Google search reigned supreme until it became a patented eponym like “just Google.” Taking on the web search giant’s dominance has been nearly impossible—until now. Broken startup Perplexity.ai used artificial intelligence tools to challenge Google’s crown. Before the end of 2024, Perplexity had approximately 15 million users, which, full disclosure, includes the author of this summary. That’s still chump change compared to Google’s nearly 5 billion users, but Perplexity offers something that traditional search doesn’t: the power of LLM.
The company, which started in 2022 with 4 employees, came across the idea of AI-powered search on a Slack channel. They combined several artificial intelligence tools, including Search Augmented Generation (RAG) to read web pages relevant to a particular search, Bidirectional Encoder Representation of Transformers (BERT) to rank web pages, and an abbreviated web crawler to index the Internet.
Joshua Sortino/Unsplash
Most AI models are data intensive. Chatbots, for example, are trained on most of the Internet before they can “speak” well. Scientific models of artificial intelligence are no different. In many cases – such as modeling airflow around an airplane wing or a star collapsing into a black hole – generating high-quality training data for AI models is slow and expensive.
One approach is to use AI-generated training data to train another AI model. But even that can be expensive and inaccurate. A team of researchers from the Georgia Institute of Technology, IBM Research, and MIT has developed a solution that reduces the training data needed to achieve the desired accuracy by a factor of 100. Their model, called physics-enhanced deep substitution, combines first principles of physical theory with a neural network to create a model that is better than the sum of its parts.
Elias Stein
When we talk about the power cost of artificial intelligence or computing in general, we usually think of hardware – how efficient are CPUs and GPUs at doing their jobs? But the way we write software can have drastic effects that often go unnoticed. For example, proper website design can reduce 93 percent of emissions generated by page load.
Designing greener software is a win-win: the software itself is more efficient, runs faster, and causes fewer emissions. However, it requires some awareness and some thought in designing and implementing the desired solution. This article highlights the growing green software movement and provides ground rules for building more energy-efficient websites, applications, and AI implementations.
As management consultant Peter Drucker is said to have said, “What gets measured gets improved.” This applies to the energy effects of the software. There are a growing number of tools that measure emissions from websites, code bases, artificial intelligence and more, but experts say access to reliable data remains a challenge and better measurement tools are needed before we can truly decarbonize software.
Moor Studio/iStock
The founder of the open-source web platform WordPress has published a cease-and-desist letter against WordPress hosting service WP Engine, claiming that neither money nor developer hours are being contributed to the project. WP Engine sued the founders in response to copyright infringement.
This legal battle highlights a fundamental question at the heart of the open source model: How does one get paid for work that is given away for free? In the early days, open-source development was done by enthusiasts working on passion projects in their spare time. Now, these passion projects power an estimated 70 to 90 percent of all apps, and big companies make good money from them. This has brought the crisis to a head: Open source project managers are reporting growing levels of dissatisfaction, and poorly maintained projects are creating security vulnerabilities. But there is hope: efforts to convince companies to commit to paying administrators are growing.
Andrew Archer
What would it take to build a computer the size of a planet? In a departure from our traditional reporting model, IEEE Spectrum commissioned science fiction writer Karl Schroeder to present the answer to this question. Contributing editor Charles Choi commented on the story, explaining how the fictional world draws from real science and technology.
Virtual minds floating in a computer made from the planet Mercury organize efforts to terraform other planets. Will it be a brave new world or a solar system of loneliness?
From your articles
Related articles on the web