Heurist’s Guide to AI: Beginner’s Series Part 1

5 min readJan 11, 2024


Welcome to our first blog article in the Beginner’s Guide to AI Series.

We know that AI can seem complex, but we’re here to break it down in simple terms. In this series, you’ll learn not just the fundamentals of AI but also get insights into what we at Heurist are doing in this exciting field. Whether you’re new to AI or just looking to refresh your knowledge, these articles are for you.

Artificial Intelligence can be thought of as a way for computers to mimic human intelligence. Imagine a machine that can learn, adapt, and make decisions based on the information it receives — that’s AI in a nutshell. Unlike traditional software that follows strict rules, AI evolves, learning from data to improve over time.

What exactly is an “AI model”? Think of it as the brain behind AI’s decision-making ability. Just like the knowledge of the human brain is shaped and accumulated through experiences and learning, an AI model is shaped through data. This model isn’t a static entity; it’s dynamic, constantly evolving and improving as it processes more data. There are many types of tasks that AI models can perform, for example, recognizing speech and converting it to text, detecting objects in images, making product recommendations based on user behavior, predicting stock market trends, or diagnosing diseases from medical images. Recently, the rise of generative AI has drawn people’s attention more than ever. Generative AI is about creating new content, whether it’s writing a poem, composing music, or generating images. It’s like giving a computer an imaginative mind, enabling it to generate original, creative work based on its training.

In the world of AI, two critical processes occur: training and inference. Understanding training and inference is crucial in appreciating AI’s capabilities. Training is like teaching a child through examples. The AI model, during its training phase, is exposed to vast amounts of data, learning patterns and associations. This process requires significant computational power and data, shaping the model’s ability to understand and respond to future inputs.

Inference is where the true magic of AI comes alive. It’s the model applying its “knowledge” to make decisions or predictions on new data. Imagine asking a chatbot a question; the chatbot uses inference to comprehend your query and respond appropriately.

Traning and Inference. Image Source: https://digitalchutney.blog/ml-training-inference/

In essence, while training shapes the AI’s brain, inference brings its intelligence to the real world, interacting with users and providing solutions. At Heurist, we focus on inference because it requires much less computational power, making it ideal for hosting on consumer-level hardware in a decentralized network.

ChatGPT and Beyond: A New Era of Large Language Models (LLMs)

Imagine having a conversation with someone who knows a little bit about everything. That’s the experience of interacting with a Large Language Model (LLM) like ChatGPT. These models are trained on vast swaths of text data, enabling them to understand and generate human language-like text. They can write stories, answer questions, and even program, making them incredibly versatile.

However, the world of LLM is not just limited to well-known models like ChatGPT. There’s a growing buzz around open-source alternatives. These models are not only cost-effective but also demonstrate the potential to match or even surpass the performance of renowned models like ChatGPT 3.5, and in certain aspects, are on par with GPT-4.

LLM development timeline, with models below the line being closed-source while models above the line being open-source. (Source)

It’s important to understand the difference between closed-source and open-source AI models, especially in the context of trust, fairness, and cultural sensitivity. Closed-source AI models, often owned by tech giants like Microsoft and Google, do not provide public access to their underlying algorithms, model weights, or training data. Dominated by a few powerful entities, these models centralize power in the AI sector, potentially stifling innovation and competition.

The industry is thus scrutinizing closed-source AI models. Recently, The New York Times sued OpenAI for misusing copyrighted work during ChatGPT training and even demanded that ChatGPT should be deleted. This lawsuit shows why we need to be careful about the transparency, trust, and accountability issues in AI.

More people are turning to open-source AI models. They are transparent, meaning that you can discover every bit of data about them. They encourage innovation because anyone can use and improve them in a permissionless way. They’re also more flexible, fitting different cultures around the world.

Unleashing Creativity with AI: The Power of Image Generation Models

Imagine having a tool that acts like a digital artist, capable of translating your words into vivid, detailed images. That’s exactly what image generation models such as Stable Diffusion do. You provide a description of a scene, character, or object, and the model uses its training to generate a corresponding image.

For artists and designers, this technology opens up incredible new possibilities. It allows them to experiment with visual ideas quickly, without needing to sketch out every detail manually. An artist might use it to generate initial concepts or explore different styles and compositions before finalizing their artwork. This can significantly speed up the creative process and inspire new, unexplored avenues of artistic expression.

Designers in fields like fashion, interior, and graphic design can use these models to visualize their ideas in a more tangible form. For example, an interior designer could describe various styles and elements they envision for a space, and the AI would generate possible looks. This makes the process of conceptualization and client communication much more dynamic and immersive.

In the communication industry, image generation models are changing the way content is created. For advertisers, marketers, and media professionals, these tools offer a fast and cost-effective way to produce visual content. Instead of coordinating photoshoots or spending hours on graphic design, they can simply describe the image they need, and the AI model generates it. This not only speeds up the content creation process but also allows for a higher degree of customization and creativity.

A remake of an old commercial entirely generated by AI. The original commercial required 30 people and took about a month to produce, whereas the AI-driven remake was accomplished by one individual in less than a day. Created with Pika (https://pika.art/)

Conclusion and Our Vision

At Heurist, we are at the forefront of empowering AI with blockchain technology. By supporting inference for both Large Language Models (LLMs) and image generation models like Stable Diffusion, we allow for innovative applications at reduced costs, and a new level of transparency in AI operations. We ensure that decentralization is not just a buzzword but a fundamental principle in our protocol. We will continue to publish more blog articles like this to bring our community deeper into the world of decentralized AI.

Our Twitter: https://twitter.com/heurist_ai

Discord: https://discord.gg/XQ2Mw3pcKd

Contact Email: team@heurist.xyz




Decentralizing AI model hosting and inference on ZK Layer-2