Well well well young padawan, we meet again. As a prelude to the past, and an ode to the timeline we’ve been tracking, it’s time we talk about the endless possibilities in store for the future before we step back into the equally enthralling voyage of open source in the early 2010s.
The State of the Art
405 billion parameters. A context window of 128K tokens. Till date, the largest open source LLM. Unlike its primarily text-based precursors, Llama-3-405B is definitely one for the history books. Its multimodal nature paired with the months of training over countless GPUs, has definitely given Meta a standing in the race to AGI, beating Open AI’s GPT-4 Omni and Claude Sonnet by a considerable margin, at least from a benchmark standpoint.
Now the notion of “more is better” definitely doesn’t always hold true and even the most novice are aware of the curse of dimensionality. So does Llama’s increased token size really surpass other LLMs in the standing? Short answer, it’s extremely good on paper. According to the Scale AI’s SEAL leaderboard, Llama-3.1-405B ranks second in math and reasoning, fourth in coding, and first in following instructions.
The decoder-only transformer architecture paired with the Fairscale PyTorch Library features open weights, a freely accessible open source model and three tiers of usage. Is there even a catch?
While its training data is not visible for obvious purposes (Yes Zuck, we know your secrets) and a Meta approved license if the threshold of 700 million monthly active users is broken, the playing field is definitely leaning towards Llama 3.1’s favour now.
Another recent groundbreaking update, with humble beginnings in a Dutch attic, is Blender 4.2, captivating artists, animators, and educators alike.
Time for a quick history lesson. Imagine a tool born not just out of necessity, but out of a quest for creative freedom. Back in 1994, Ton Roosendaal set out to tackle the headaches of client demands with what would become Blender. Named after a song, Blender started as an in-house solution at NeoGeo, Roosendaal’s animation studio. Fast forward to 2002, Blender took a bold leap into the open source realm, thanks to a groundbreaking crowdfunding effort that saved it from financial woes. Since then, it's been a whirlwind of growth and innovation, driven by a vibrant global community.
With Blender 4.2, it isn’t just an update; it’s a game-changer. This Long Term Support (LTS) version brings stability and reliability to the forefront, making it perfect for both production environments and educational settings alike, featuring an enhanced user interface, revamped sculpting, enhanced asset management and smoother workflows with geometry nodes.
Beyond their impressive features, Blender and Llama-3.1-405B embody the spirit of open source innovation. It’s not just about what they can do; it’s about the community that fuels their growth.
Now, back to our scheduled program folks.
Cloud Nine for Open Source
Rewinding back to the year 2010, the rising tide of open source software and the booming demand for cloud computing converged, giving birth to OpenStack –a collaborative project between Rackspace Hosting and NASA.
OpenStack has democratized computing power, enabling innovation to everybody. It is an open source cloud operating system that has redefined the creation and management of public and private clouds.
Offering powerful Infrastructure-as-a-Service (IaaS) capabilities that allow organizations to rent IT infrastructure on a pay-as-you-go basis, OpenStack empowers on-demand deployment and management of virtual machines (VMs). Rather than investing in physical hardware, these VMs handle computing, storage, networking, and identity services. Not just that, OpenStack provides administrators with a user-friendly dashboard for seamless control.
Beyond its open source nature, OpenStack's true strength lies in its exceptional compatibility with software solutions as well as other open source technologies. To enhance application availability, OpenStack's interconnected components provide advanced tools for managing processes, handling problems, and overseeing services.
OpenStack was in its “media darling” phase in the 2010s, pushed as an open source alternative to proprietary solutions like AWS and VMware, ripping off the shackles to the pay-as-you-go models.
The formation of the OpenStack Foundation in 2012, backed by industry giants like IBM, HP, and Red Hat, cemented OpenStack's reputation as a formidable cloud infrastructure solution. By 2014, OpenStack had emerged as the leading choice for cloud infrastructure, propelled by a vast community of developers and organizations.
Today, OpenStack stands tall as one of the most active open source projects. While some may argue it no longer remains cutting edge, its influence is pervasive. It serves as the backbone for Kubernetes, AI/ML stacks, and Telco NFV, and organizations across nearly every industry have embraced OpenStack. Over 300 data centers worldwide now boast OpenStack-powered public clouds, underscoring its undeniable impact.
As OpenStack laid the foundation for open source cloud infrastructure, other key players were arriving on the horizon, promising to revolutionize how applications were packaged and deployed— Docker and Kubernetes.
Docker and Kubernetes: The Dynamic Duo
The Advent of Docker
Newer versions. Faster updates. Larger Communities. Insanely smart inventions. The Open Source Revolution was finally taking off. Maybe a bit too much. With the exponential increase in open source software, people had a plethora of technologies playing a weirdly tuned orchestra with right tones but the wrong harmony. That’s when Docker paved an innovative approach for the world to witness.
Docker is an open source tool that uses containerization, much like your run-of-the-mill zip-file, where the contents are packed into one compressed file. In Docker’s case, it packages an application's code and all the things it needs to run, like libraries and settings, into one package. This makes it easier to develop, test, and run software consistently.
The concept of containers and virtualization has been around for some time, from the early Solaris containers, an old-school system from the '90s to Linux Containers (LXC). Docker was released as open source in 2013, playing a crucial role in packing containers to increase portability across environments. While Virtual Machines (VMs) are commonly used to distribute applications and support legacy apps, they still have their drawbacks due to their large size, high resource consumption, and portability issues. Docker serves as a better alternative: containers are small, have minimal overhead, are completely portable, and are designed with an application-centric approach; an all-rounder that knocks the VMs out of the park.
Docker containers give you a consistent runtime environment everywhere, whether on your laptop, a cloud server, or even a smart fridge at this point because of how influential it is. Docker made container deployment safer and easier than ever. Now, giants like Spotify, Netflix, and Uber use Docker to scale up effortlessly, save a ton of cash, and make their services top-notch.
The Captain and the Navigator
Imagine Docker as the sturdy ship that keeps your valuable cargo (applications and their dependencies) safe and sound. Kubernetes, on the other hand, is the expert navigator, charting the course, managing the crew (containers), and ensuring smooth sails even in the stormiest seas. Together, they make a formidable team, ensuring your precious cargo reaches its destination efficiently and reliably.
Oh Captain! My Captain
The birth of Kubernetes can be traced all the way back to the early 2000s, when Google needed a way to efficiently manage its massive fleet of servers. Enter "Borg" – Google’s first-ever container management system, named after the hive-minded Borg aliens from Star Trek. Creative huh. Developed between 2003 and 2004, Borg centralized the management of tasks across Google’s data centres, powering up services like Gmail, Google Docs, and YouTube with unprecedented efficiency.
Fast forward to 2013, and Google introduces "Omega," a sequel to Borg that elevated container management even further. Around the same time, Docker emerged as another crucial player in the tech world. Inspired by their experiences with Borg and Omega, developers set out to create an open source tool originally codenamed "Project Seven of Nine" (in context to Star Trek’s Borg lore). Eventually, this project emerged as Kubernetes – a name that’s easier to pronounce and remember!
Kubernetes isn’t just any old container orchestrator. It’s like the pilot in command of cloud computing – portable, flexible, and completely open source. Think of it as a starship captain expertly managing containers (small, self-contained spaceships for your apps) across vast clusters of servers. Whether you’re running a star-studded website or plotting your next intergalactic app, Kubernetes ensures your ship (or rather, your app) sails smoothly through the cosmos of computing.
Since joining the Cloud Native Computing Foundation (CNCF) in 2016, Kubernetes has grown faster than a warp-speed spaceship. Thousands of developers contribute to its code, and its annual conference, KubeCon, is like Comic-Con for cloud enthusiasts! Experts predict that by 2027, almost every organization on Earth will be running containerized apps thanks to Kubernetes.
So, whether you dream of building the next Google or simply want to understand how the cloud works, Kubernetes is your launchpad to a universe of possibilities.
In retrospect, the early 2010s saw pivotal advancements in open source technology with projects like OpenStack, Docker, and Kubernetes. These innovations not only transformed cloud computing and software deployment but also underscored the collaborative spirit that defines the modern tech landscape. They continue to influence and inspire, demonstrating the enduring impact of open source principles on innovation and industry evolution.