Iscriviti al feed

The more you use artificial intelligence (AI) tools, the more you’re likely to find that a single AI model doesn’t have all the answers you need. Simply put, an AI trained on general knowledge from around the internet isn’t particularly skilled at providing specialized knowledge, regardless of how well it generates techno-babble to emulate a valid answer. InstructLab is an open source project that gives you the power to construct, adapt, or adjust a large language model (LLM) specific to your organization’s requirements. People are excited about InstructLab for more reasons than that though, so we asked our very own Red Hatters using InstructLab why this project had inspired them to turn to open source for their generative AI needs.

1. You can learn the technology

Rodrigo Freire is a Red Hat Chief Architect in Brazil, and he’s interested in InstructLab because it’s given him the chance to learn (and share) about the technology.

I wanted to do AI on the command-line, mostly to get a better understanding of the technology and how it works. When the InstructLab project was announced, I partnered with one of our Specialist Solution Architects to learn more. We came up with a tutorial on how to deploy it on a laptop without an underlying container infrastructure.

With this setup running on a modest laptop, it was interesting to demo it for customers and friends. It also revealed the need for a good GPU for fine-tuning. It’s the perfect model to tinker with on your laptop, but get RHEL AI [Red Hat Enterprise Linux AI] on some serious hardware before you put it in production!

Rodrigo Freire

2. It’s not a monoculture

Nicholas Jayanty, a Principal User Experience Researcher in Red Hat Products and Global Engineering, points out that InstructLab is helping the IT industry avoid the trap of a monoculture.

AI has long been seen as an entirely opaque “black box”. Bringing source control management through a GitOps workflow brings us one step closer to opening that box up. This visibility allows for greater collaboration and accountability.

How we make AI is just as important as what we make. Instructlab has created an inclusive user experience that brings diversity to the way models are improved. This diversity of contributors mitigates the risk of AI being developed in a monocultural environment.

In the opening chapters of The Open Source Way, there’s talk about the importance of lowering barriers to contribution. InstructLab’s user experience hasn’t just opened the doors to contribution, it’s removed them altogether. I never thought I’d be able to contribute to an open source project, given my technical limitations. But I’ve contributed my first skill, an Inclusive IT Terminology skill, thanks to InstructLab. That’s my first open source contribution ever! It’s a user experience design triumph.

3. Your feedback makes a difference

As Nicholas Jayanty points out, feeling ownership of an AI is difficult when there’s literally no way to contribute to it. But then again, just because an LLM offers users the opportunity to contribute doesn’t mean everyone knows how to do that. Jana Gutierrez Kardum, an account planning and sales productivity lead in Dubai, notes that Git isn’t an entirely simple technology, but InstructLab’s developer team has listened to feedback about usability, and has managed to make it easy to contribute nevertheless.

I immediately knew that I wanted to try and contribute to InstructLab despite the fact that I’m not very technical. I booked a couple of hours (well, it was a bit more than a couple) to learn the process. I created an account on Github, and through trial and error I eventually got to the point where I made the model work and produce the training data based on the skill I provided. I must say that the moment when the file was generated, and I was able to feed it back and observe how the model learns, was extremely satisfying.

It’s one thing to use a generative AI model as a user who just consumes whatever’s given to you. But when you’re actually able to contribute something, and then see the model absorbing your input, is a whole different experience, and I recommend everyone to go and try it. For me as a person with very limited (that is, non-existent) experience with Github, the most complicated part was understanding the instructions. To make the journey easier for other non-technical users, I created a step-by-step presentation of all the steps I needed to do in order to make the model work. That was pretty useful, at that point in time. As of July 2024, the official documentation of InstructLab is much simpler and clearer. I recently downloaded InstructLab onto my personal computer, and was able to get it up and running quickly, without any major issue. So don’t be shy and give it a try.

Jana Gutierrez Kardum

4. Your artificial intelligence needs full transparency

Dan Aubin, a Red Hat IT support engineer in Australia, is interested in InstructLab for its inherent transparency.

Being able to test local models without cloud infrastructure and hosting is an empowering way to test and learn about generative AI in an open source way. Transparency and inclusivity are really important to software development, especially in the AI space. There’s a lot of concern over data privacy, so it’s vital to be able to test software locally, and determine whether there are security concerns.

Kush Gupta is a Red Hat solutions architect, and has been working professionally and personally on improving AI. InstructLab makes it possible, and he has the code to prove it.

I’m excited for the continued development and improvement of open source LLMs, and InstructLab is the platform for that. With lots of bigger companies shipping a mix of open and closed source development and models, InstructLab is a first in making the models open, and the mechanisms for improving these models easy to use. The result speaks for itself.

I created an experimental project called Instruction Synthesizer that leverages an LLM to exclusively generate question and answer pairs, formatting its response in the Taxonomy format that InstructLab uses. In other words, you can generate a draft Taxonomy entry based on whatever data you throw at it!

It’s important for people to be able to contribute to AI models in an open source way because certain companies in the AI space are using what’s there in the public, but never contributing innovation back. Without this back and forth, innovation is lost, and many people end up duplicating work. InstructLab flips that paradigm completely, increasing transparency to the fullest and allowing others to innovate on what we have all built together.

Kush Gupta

5. AI is a hybrid platform

InstructLab doesn’t exist in a vacuum. No single AI or LLM is ever going to represent the single best way to summarize, infer and deliver data. Cedric Clyburn is a developer advocate in New York City, and he’s excited about how InstructLab technology can be combined with other approaches to AI.

There’s been buzz among customers and the community lately about when to use fine-tuning (with approaches like InstructLab) and when to use RAG (retrieval-augmented generation). Many developers are curious about the key differences and use cases between those two approaches. InstructLab enhances an LLM by adding skills and knowledge through fine-tuning, using a small set of human examples to generate high-quality synthetic data for training. This improves a model’s overall capabilities, effectively making knowledge “baked-in” for faster inferencing. RAG, on the other hand, supplements an LLM with domain-specific information without retraining, which is ideal for basic model enhancement or accessing up-to-date information. You don’t necessarily have to choose one over the other. They can be complementary.

Using RAG with an InstructLab-tuned model can provide superior results, combining enhanced model capabilities with dynamic knowledge retrieval. This is ideal for enterprise environments where specialized domain knowledge and up-to-date information is crucial.

Cedric Clyburn

Try InstructLab

Learn how to get started with InstructLab, or check out these videos about what InstructLab is and how to use it!


Sull'autore

Seth Kenlon is a Linux geek, open source enthusiast, free culture advocate, and tabletop gamer. Between gigs in the film industry and the tech industry (not necessarily exclusive of one another), he likes to design games and hack on code (also not necessarily exclusive of one another).

Read full bio
UI_Icon-Red_Hat-Close-A-Black-RGB

Ricerca per canale

automation icon

Automazione

Novità sull'automazione IT di tecnologie, team e ambienti

AI icon

Intelligenza artificiale

Aggiornamenti sulle piattaforme che consentono alle aziende di eseguire carichi di lavoro IA ovunque

open hybrid cloud icon

Hybrid cloud open source

Scopri come affrontare il futuro in modo più agile grazie al cloud ibrido

security icon

Sicurezza

Le ultime novità sulle nostre soluzioni per ridurre i rischi nelle tecnologie e negli ambienti

edge icon

Edge computing

Aggiornamenti sulle piattaforme che semplificano l'operatività edge

Infrastructure icon

Infrastruttura

Le ultime novità sulla piattaforma Linux aziendale leader a livello mondiale

application development icon

Applicazioni

Approfondimenti sulle nostre soluzioni alle sfide applicative più difficili

Original series icon

Serie originali

Raccontiamo le interessanti storie di leader e creatori di tecnologie pensate per le aziende