Awakened Intelligence 2025#004
What can you do with a Generative AI model, cool learning resources, and how to run LLMs on your phone - he PoorGPUguy weekly dose of cutting-edge open-source AI news & insights.
Welcome back to our weekly dose of AI! In this newsletter we are going to see some of the amazing thing you can do with Generative AI. I am not using anymore only LLM (Large Language Models) for two main reasons:
I like also Small Language Models (they can run on almost any hardware and perform as good as the Big shots)
Visual Models and more generally Multi-modal models are on hype lately, so we cannot leave them in a corner.
Let’s dive in!
Hugging Face SmolAgents
In the past few weeks, Hugging Face started another amazing project called smolagents. It is the trend of the 2025 (?) and many are claiming that this will be the year of the agents…but 🤔 What are agents?
Any efficient system using AI will need to provide LLMs some kind of access to the real world: for instance the possibility to call a search tool to get external information, or to act on certain programs in order to solve a task. In other words, LLMs should have agency. Agentic programs are the gateway to the outside world for LLMs.
AI Agents are programs where LLM outputs control the workflow.
An agentic system runs in a loop, executing a new action at each step (the action can involve calling some pre-determined tools that are just functions), until its observations make it apparent that a satisfactory state has been reached to solve the given task. Here’s an example of how a multi-step agent can solve a simple math question:
It is a slim and powerful library, with plenty of tutorials.
Go and test them out!
OpenVINO with DeepSeek-R1
𝗗𝗲𝗲𝗽𝗦𝗲𝗲𝗸-𝗥𝟭 is a popular reasoning model, as discussed in the previous newsletter. To use it, even in its smaller format (1.5 Billion parameters), is almost impossible without a NVIDIA GPU. So usually we go for llama.cpp.
But what if you can make full use of your integrated Graphic card, coming with almost every Intel chip?
🚀 Unleash the power of 𝗗𝗲𝗲𝗽𝗦𝗲𝗲𝗸-𝗥𝟭 on your 𝗜𝗻𝘁𝗲𝗹 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗲𝗱 𝗚𝗣𝗨 with just 𝟯 𝗹𝗶𝗻𝗲𝘀 𝗼𝗳 𝗣𝘆𝘁𝗵𝗼𝗻 𝗰𝗼𝗱𝗲! 🤯
OpenVINO™ GenAI is a library of the most popular Generative AI model pipelines, optimized execution methods, and samples that run on top of highly performant OpenVINO Runtime.
This library is friendly to PC and laptop execution, and optimized for resource consumption. It requires no external dependencies to run generative models as it already includes all the core functionality (e.g. tokenization via openvino-tokenizers).
What you need to run a Small Language Model on your Phone
In reality Asghar Ghorbani gitHub repo is all you need, as simple as that. He created two apps, one for Android and one for iOS that you can directly download in the respective App stores.
So it is better to say that, PocketPal is all you need
He was really good at all the details, including that you have out of the box two amazing features:
you can download the LLM (small ones) directly from the app
you can add your own small Language models in GGUF format
First of all download and install the App from your store (in my case Google Play store). After that open it to start getting and configuring the models.
PocketPal AI comes pre-configured with some popular SLMs:
Danube 2 and 3
Phi
Gemma 2
Qwen
Deepseek-R1
Run Language models on your iPhone
Dmitry “Leo” Kuznetsov is a software developer working on open source free GyPTix app that is capable of running on iPhone and other iOS devices as well as macOS. Just to stay on topic, he is working on integrating DeepSeek Janus Pro 1B into GyPTix right now.
What is amazing on this project, totally free, is that DeepSeek Janus Pro 1B is a multi-modal input/output generative model: it means that you can load an image and chat with it, or you can even ask to generate images!!
You can check it out here:
Why Choose Gyptix?
Absolutely FREE – No hidden costs or subscriptions.
100% Private – No personal data collection, ever.
Your Data, Your Control – Nothing is stored or shared.
No Internet Required – Works entirely on your mobile device.
Local AI Engine – Runs independently without network access.
True Offline Mode – Functions even in airplane mode.
Go and check it out! The app is in preview on Apple store, it costs nothing (it is free) and the author does not have any plans to monetize it!
Learning about Generative Artificial Intelligence
Dive Deep into LLMs with Andrej Karpathy's Masterclass
Want to truly understand the magic behind Large Language Models (LLMs)? Many of you may wander how to get started or deepen their existing knowledge, and I have the perfect recommendation: Andrej Karpathy's incredible introductory session.
If you have 3.5 hours to spare, dedicate it to this invaluable resource. Frankly, I haven't found (yet) a more effective way to grasp the complexities of LLMs.
Karpathy's session covers the entire training stack, from the ground up, explaining how these models are developed. He also provides insightful mental models for understanding their "psychology" and how to use them effectively in practical applications. Andrej has the ability to make such a complex topic so accessible and comprehensive. Here's a glimpse of what you'll learn:
Pretraining: starting from the foundations, covering data, tokenization, the inner workings of Transformer neural networks (including I/O), inference, and practical examples like GPT-2 training and Llama 3.1 base inference.
Supervised Finetuning: here he explores the world of conversational data and gains insights into "LLM Psychology," including hallucinations, tool use, knowledge/working memory, self-awareness, the importance of tokens for processing, spelling nuances, and the concept of jagged intelligence.
Reinforcement Learning: Understand the power of practice, with discussions of DeepSeek-R1, AlphaGo, and the crucial role of Reinforcement Learning from Human Feedback (RLHF).
Well, I think with this you have a week plenty of explorations.
Today as a gift I will leave for you an article that explains how to generate images locally from your PC, even if you don’t have a powerful GPU… It totally runs on every computer on the planet!
This is only the start!
Hope you will find all of this useful. I am using Substack only for the newsletter. Here every week I am giving free links to my paid articles on Medium. Follow me and Read my latest articles https://medium.com/@fabio.matricardi
Check out my Substack page, if you missed some posts. And, since it is free, feel free to share it!