TLDR: Build in the real world or simulate on your device. Flatpack connects AI to buttons and data, like GPIO pins connect hardware to sensors, with a free and open-source development environment, packaging, and distribution.
At Flatpack, our mission is clear: we are committed to building trust in AI.
Flatpack democratises AI and ML through micro language models and model compression. Our platform enables users to train custom language models with 100M to 10B parameters. We introduce flatpacks (FPKs) to integrate AI and ML into edge computing, electronic components, and robots.
在 Flatpack，我们的使命十分明确: 致力于建立对 AI 的信任。
Flatpack 通过微语言模型和模型压缩普及 AI 和 ML。我们的平台使用户能够训练具有 100M 到 10B 个参数的自定义语言模型。我们引入了 flatpacks (FPKs)，将 AI 和 ML 集成到边缘计算、电子元件和机器人中。
Edge artificial intelligence
Edge artificial intelligence uses local devices to enhance decision-making near data sources, improving privacy, response times, and security while reducing reliance on cloud connectivity. Benefits include decreased latency, improved scalability, and reduced energy usage.
边缘人工智能使用本地设备来增强数据源附近的决策能力，提高隐私性、响应时间和安全性，同时减少对云连接的依赖，包括降低延迟、提高可扩展性和减少能源消耗等优势。
Micro language models
Micro language models (100M to 10B parameters) provide an efficient alternative to large language models, addressing deployment, reliability, and scalability challenges. Their compact size enables rapid pre-training and fine-tuning, allowing organisations to adapt swiftly to market shifts and specific needs. These models can be readily deployed in edge-adjacent environments such as laptops, microprocessors, and smartphones, offering accessibility and versatility.
微型语言模型（100M 到 10B 个参数）为大型语言模型提供了一种有效的替代方案，解决了部署、可靠性和可扩展性挑战。其紧凑尺寸可实现快速预训练和微调，使组织能够迅速适应市场变化和特定需求。这些模型可轻松部署在笔记本电脑、微处理器和智能手机等边缘相邻环境中，提供可访问性及多功能性。
Python package
On August 17, 2023, version 0.0.1 of our Python package was launched under the Apache License 2.0, marking its initial release to the public.
Hooks and schedules
Flatpack uses hooks to trigger custom actions in response to specific events, while schedules automate AI processes at predetermined intervals.
Open weights
Genuine open source in AI hinges on one crucial element: open weights. Without them, it is like borrowing a brain instead of genuinely owning the knowledge.
“Not your weights, not your brain. /…/ You realize you are renting your brain.” (Andrej Karpathy 2024)
Fortunately, more big tech companies, including IBM and Microsoft (kudos), are releasing open-weight models under permissive licenses, reducing the need to reinvent the wheel and emphasizing the importance of integration and utility in advancing community-driven AI innovation.
IBM Granite (no affiliation): “Granite is open, trusted and targeted. Two ways to think about openness, one, open as open weights. It’s available for public to download. And the second one is open as in, there is less restrictions on how the customers can legally use these models for a range of use cases.” (Maryam Ashoori 2024)
🔗 IBM Granite 3.1 (Apache 2.0)
Microsoft Phi-4 (no affiliation): “A lot of folks had been asking us for weight release. /…/ Well, wait no more. We are releasing today official phi-4 model on HuggingFace! With MIT licence!!” (Shital Shah 2024)
Market size
According to a 2023 McKinsey report, generative AI could add $2.6-$4.4 trillion annually to the global economy, comparable to the UK’s 2021 GDP of $3.1 trillion (McKinsey 2023).
“That would add 15 to 40 percent to the $11 trillion to $17.7 trillion of economic value that we now estimate nongenerative artificial intelligence and analytics could unlock.” (McKinsey 2023).
Did you know that the International Data Corporation (IDC) predicts AI will have a $19.9 trillion global economic impact by 2030, driving 3.5% of GDP and significantly affecting jobs worldwide (IDC 2024)?
In 2024, the top five global industries are projected to generate $22.71 trillion in revenue. These industries are life and health insurance carriers, car and automobile sales, commercial real estate, pension funds, and oil and gas exploration and production (IBISWorld 2024).
“According to the research, in 2030, every new dollar spent on business-related AI solutions and services will generate $4.60 into the global economy, in terms of indirect and induced effects.” (IDC 2024).
“A tide of investment is pouring into generative artificial intelligence. Will it be worth it? /…/ The use cases, or killer apps, that fully justify the intense investment are yet to emerge.” (Goldman Sachs 2024)