[12th May 2024] Interesting Things I Learnt This Week

1. How LLMs Work, Explained Without Math: This post by @miguelgrinberg explains LLMs working in simpler terms as well as defining some of the important LLM terminology. 

My Take: Now that all the initial hype around LLMs have kind of subsided, educating about the actual tech is taking a front seat. Its a good starting read for anyone trying to understand LLMs. I kind of loved the statement by the author "I wouldn't trust any workflow in which the LLM produces output that goes straight to end users without verification by a human.", which is pretty much on point here. 

 

2. 37 tips from a Senior Frontend Developer:  Ndeye Fatou Diop got inspired by  @abbeyperini's post and compiled his own list of tips. 

 My Take: These are some very sane tips which all software engineers should follow. These are not taught in schools but are valid none the less. All engineers learn these the hard way, but thanks to Ndeye you can learn is from his post there. 


3. Glance App: Glance is a a self-hosted dashboard. Users can add widgets to display things like RSS feeds, weather, and stock prices. Glance is lightweight and mobile-friendly.

My Take:  This is a very interesting app. It uses yaml for all configuration and is meant to be light weight with most of it written in go. It has a simple docker image which can be hosted anywhere, maybe even your own laptop. I think after iGoogle moved to graveyard, this is one of the best replacements I have seen for it.   


4. Google Removes RISC-V Support From Android: Google removed RISC-V support from the Android kernel. Google cited a lack of resources and the fragmented nature of the RISC-V ISA as reasons for dropping support. Others argue that Google is strategically blocking RISC-V to favor ARM processors. The future of RISC-V on Android is uncertain as of now.

 My Take:Google seems to be shifting focus towards AI and revenue-generating projects. While supporting RISC-V might have been strategic earlier, it appears to be a lower priority now. Some speculate this could be due to the ongoing US-China rivalry in technology. Regardless, Google's current actions suggest less emphasis on RISC-V compared to China. This could lead to China developing its own Linux version for RISC-V, or perhaps even the first widespread KDE phone.

 

5. Using an LLM for text compression: This article explores using a large language model (LLM) for text compression. The author achieves promising results by combining an LLM with arithmetic coding and a pre-processor for zlib. The experiment uses two LLMs, LLaMA-2-7B and LLaMA-3-8B, both quantized with a specific setting. The core idea is to provide the LLM with recent context and ask it to predict the 128 most likely following tokens. The longest predicted token is then chosen for compression. If none of the predictions exactly match the remaining text, the next character is encoded. The resulting sequence of integers is further compressed using ULEB128. The study shows that LLaMA-3 outperforms LLaMA-2 in most cases. The compression achieved is generally around 2 times better than gzip and at least 1.5 times better than Brotli.

My Take: This is an impressive piece of work! It reminds me of theories about DNA containing the entire blueprint for our bodies, and how it manages to compress that information so efficiently. The author did a fantastic job writing this up, but there's one key detail missing: the speed of encoding and decoding this information. The blog post doesn't mention it, but hopefully we can get it since the code is available on GitHub at https://github.com/arodland/llamacomp.

 

6.  Giant Compressed-Air Battery: This is an article about a company called Hydrostor that is building compressed-air energy storage facilities. The technology uses air, water, and a cavern to store energy for eight hours or more. The company believes this is a good solution for long-duration energy storage needs. They are building their first large-scale plant in Australia. Compressed air storage uses off-the-shelf equipment and could be cost-competitive with other options. However, the technology is still being proven.

 My Take: This is a not a novel idea, but I'm unconvinced it's an efficient way to store energy. As they said, it's unproven technology. I'd be happy to be wrong, though.

 

Comments

Post a Comment

Popular posts from this blog

[24th November 2024] Interesting Things I Learnt This Week

Image
1. AI Makes Tech Debt More Expensive - The core argument is that generative AI significantly widens the gap between codebases with low and high tech debt. Companies with clean code can leverage generative AI to write code much faster, while companies with messy legacy codebases struggle to adopt these new tools. This widens the gap between the two codebases, making tech debt even more expensive for companies with outdated systems. My Take : Legacy systems inevitably accumulate technical debt. A common analogy for this is the state of urinals over time: missed targets lead to increasing mess, keeps pushing people away from urinals and leading to more mess in turn. AI-powered coding tools could potentially accelerate the accumulation of tech debt. However, skilled engineers can leverage AI to reduce and even eliminate it. Ultimately, AI is just a tool, and human judgment remains crucial for making informed decisions. I hope AI will be used responsibly to drive positive outcomes. 2. Make...
Read more

[17th November 2024] Interesting Things I Learnt This Week

Image
 1. Booleans are a trap : Booleans, while seemingly simple, can introduce complexity and limitations into domain modeling. When used to represent complex states, booleans can lead to unclear code and potential errors. To avoid these pitfalls, consider using enums or enum sets. These approaches provide a more expressive and maintainable way to model complex states in your applications. My Take : I wish I had learned this earlier in life. When working with Google Docs, I realized the limitations of booleans. Expanding boolean-based behaviors often led to adding more fields, making the system complex. Enums would have provided a simpler solution. The problem wasn't just the booleans themselves, but also the need to pass them through multiple layers of the stack, often requiring additional parameters. Another important aspect of enums is the inclusion of an " enum not specified " state. This state explicitly indicates that an enum value hasn't been set, which can be usef...
Read more

[1st December 2024] Interesting Things I Learnt This Week

Image
1. C++ Pitfalls - this article, explains some of the pitfalls you can fall into when programming in C++ and shares their experience with them. This includes Operator precedence, Arithmetic conversion rules, Right bit-shift, Implicit conversions and Order of evaluation.  My Take : We've all read about JavaScript's quirks, leading to unexpected outputs. But let's remember that such behaviors aren't unique to JavaScript. Even languages like C and C++ have their share of ambiguities and compiler-specific behaviors. While it's tempting to dive deep into these intricacies, I've found that focusing on clear, consistent, and unambiguous code is often more practical. However, understanding these nuances is essential for any experienced programmer. It helps us avoid pitfalls and write robust, reliable code. Remember, mastering a language isn't just about writing "Hello World." It involves delving into its intricacies, including its counterintuitive and ambi...
Read more