Good Morning! A new study from GitClear suggests GitHub Copilot may introduce more coding mistakes, while Google unveils Lumiere (a Text-to-Video AI), an AI that can generate videos from text. Though developer burnout remains high, 70% still choose to code for fun on weekends, preferring written learning content over video.

— Forrest Knight & Meghanadh Vasireddy

New Research Finds GitHub Copilot May Negatively Impact Code Quality

A recent study by GitClear analyzed the impact of GitHub Copilot on code quality and maintainability. Their analysis of over 150 million changed lines of code between 2020 and 2023 uncovered concerning trends.

Specifically, they found that code churn - the percentage of lines reverted or updated within two weeks - is projected to double in 2024 compared to 2021 levels. This suggests more mistakes are being introduced and subsequently fixed when Copilot is used.

Additionally, the proportion of added and copy/pasted code increased relative to updated, deleted, and moved code. This indicates less refactoring and reuse of existing code, violating the DRY (Don’t Repeat Yourself) principle. The rise in copy/pasted code in particular may cause maintainability issues down the road.

In summary, the findings show signs that AI-generated code from Copilot:

• Introduces more defects needing rapid fixing based on high code churn

• Discourages refactoring and code reuse leading to duplication

• Increases technical debt that will burden future developers

Read More Here

Google Unveils Lumiere, a Text-to-Video AI

Google has announced the development of Lumiere, an AI system that can generate high-quality video from text descriptions. This new technique represents a significant advance in text-to-video generation through the use of a novel "Space-Time U-Net" neural network architecture.

Lumiere is able to produce smooth, realistic video animations in high resolution from brief text prompts. For example, the query "two raccoons reading books together" yields a short clip of just that. The system can also edit existing video by allowing the user to highlight part of the clip and provide modifying text, like "change dress color to red."

Additionally, Lumiere provides creative options such as stylization, cinemagraphs, and referencing the style of other videos. This could enable applications in animation, special effects, content creation, and more.

The key innovation is the Space-Time U-Net, a type of convolutional neural network designed to generate coherent video in a single pass. This overcomes limitations with prior text-to-video models that often produce blurry or disjointed results requiring lengthy processing. The Space-Time U-Net's unified architecture learns spatial semantics and temporal dynamics end-to-end.

Read More Here

Developers experience burnout, but 70% of them code on weekends

A new report from JetBrains on the state of the developer ecosystem in 2023 reveals some interesting insights into developers' experiences with burnout, learning preferences, and weekend coding habits.

Key Findings:

• 73% of developers surveyed have experienced burnout at some point in their careers. Those who have experienced burnout feel tired more often and 51% use health tracking apps or devices.

• 75% of developers have quit a learning program or course at some point, preferring to learn through documentation, APIs, blogs, and forums instead. 53% prefer written content for learning compared to 45% who prefer video.

• 70% of developers report that they code for fun on the weekends, despite high reported burnout rates.

Read More Here

O’Reilly finds 3,600% increase in interest in generative AI in last year

O'Reilly's annual report on technology trends shows staggering increases in interest around generative AI and related topics over the past year.

Interest in generative pretrained transformers (GPT) and generative AI overall grew an astounding 3,600% year-over-year. Other AI topics also saw spikes, including a 900% rise in generative models, 325% in transformers, and 195% in natural language processing.

The report also uncovered growing interest in prompt engineering for fine-tuning AI models, on par now with interest in transformers.

Beyond AI, security is another area that saw major upticks, especially among developers. Network security grew 5%, while governance saw 22% more interest. Application security surged 42%, with DevSecOps specifically rising 30%, indicating shifts towards bake-in security.

On the programming side, stalwart C++ rose 10%, likely due to its use in AI systems. Rust also grew 7.8% in interest year-over-year.

More broadly, developers seem to be working on soft skills like communication, professional development, and project management.

Read More Here

Youtube Spotlight

I built my own 16-Bit CPU in Excel

The video shows Inkbox functional 16-bit CPU, coupled with 128 kilobytes of RAM and a 4K monitor, entirely within an Excel spreadsheet. The process involves building individual components of the CPU, creating an Instruction Set Architecture, designing higher-level CPU functionalities, and even developing a custom Assembly Language called EXCEL-ASM16.

Was this forwarded to you? Sign Up Here