Augmented Development
A practice where human programmers use artificial intelligence tools as co-workers to write, test, and fix code.
Max Sanna
AI Technologies
A practical overview of Max Sanna artificial intelligence: what it is, where it helps, which models are often recommended, and how to use AI responsibly, fairly, and without bias in real projects.
Artificial intelligence
Recommended Models
AI Art
Responsible AI
General Information About Max Sanna AI Technologies
Artificial intelligence is a broad field focused on building systems that can perform tasks normally requiring human intelligence. Max Sanna uses AI for various use cases, including buyt not limiting:
Natural Language Processing
Language models are used to summarize, explain, draft, translate, and answer questions.
Image Generative AI
Text-to-image models are used to generate and edit images.
Sound and Speech AI
Audio models are used to generate speech and perfrom various actions on music.
Video Generative AI
Text-to-video models are used to generate video content.
Computer Vision
Vision models understand images, detect objects, and classify visual patterns.
Max Sanna Recommended AI Models
Max Sanna uses open source open weight models as most secure from privacy point of view and fair bias controllable ones.
General LLM: Qwen, DeepSeek, Llama Family Models
Strong all-purpose reasoning, summarization, instruction following, and mixed-content workflows.
Agentic Coding: Local Qwen Coder Family Models
Useful for code generation, refactoring, test writing, and architecture explanation.
AI Image Generation: SDXL, Qwen Image, Flux Family Models
Good for concept art, product mockups, and visual ideation with prompt iteration.
AI Video Generation: WAN Family Models
Good for short concept video generation.
AI Sound and Speech: MMAudio and UVR Models
Speech generation, music vocal remover and other audio manipulations.
CLIP, VAE, Text Encoders, Embeddings, etc.
Auxiliary models enabling semantic retrieval, text-to-image context-awarness, etc.
Qwen Family: Qwen3.8 27B; Qwen3.6 27B, 35B-A3B MoE; Qwen3.5 0.8B, 2B, 4B, 9B, 27B, 35B-A3B MoE
Qwen3 Coder: Qwen3 Coder-Next 80B-A3B MoE; Qwen3 Coder 30B-A3B
Qwen3-VL 2B-Instruct/Thinking, 4B-Instruct/Thinking, 8B-Instruct/Thinking, 30B-A3B-Instruct/Thinking
DeepSeek-R1, R1-0528-Qwen3-8B, Distill Llama 3 8B, Distill Qwen 2.5 1.5B, Distill Qwen 2.5 7B, Distill Qwen 2.5 14B, Distill Qwen 2.5 32B
Nomic Embed Text V2 475M-A305M (nomic-embed-text-v2-moe) open-source, multilingual text embedding model (by Nomic AI).
CLIP-L (Contrastive Language-Image Pre-training - Large), CLIP ViT-L/14 vision-language text encoder model (by OpenAI)
OpenCLIP ViT-bigG/14 open-source CLIP model (by the LAION community)
FLUX.1 [dev] 12B open-weight text-to-image model (by Black Forest Labs)
FLUX.1 Krea [dev] 12B open-weights text-to-image model (by Black Forest Labs and Krea AI)
FLUX.1 Kontext [dev] 12B open-weight model (by Black Forest Labs)
Z-Image-Turbo 6B open-sourcetext-to-image model (by Alibaba’s Tongyi-MAI team)
Wan2.2-TI2V-5B, Wan2.2-T2V-A14B, Wan2.2-I2V-A14B open-source MoE video generation models (by Alibaba)
Qwen-Image-2512 (by Qwen team [Tongyi Qianwen team])
Qwen-Image-Edit-2511 (by Qwen team [Tongyi Qianwen team])
Stable Diffusion XL (SDXL) open-source text-to-image generation model (by Stability AI)
- Base, DreamShaperXL, JuggernautXL, animaPencilXL, etc.
Variational Autoencoders (VAE) (by Stability AI and Black Forest Labs)
MMAudio open-source model (by Sony AI and the University of Illinois)
T5-XXL text encoder models (by Google Research)
Max Sanna AI Generative Art
Some examples of Max Sanna Generative AI usage.
On the Edge of Hallucination
The Hole
Feather Waves
Max Sanna AI Generative Short Videos
Some examples of Max Sanna Generative AI usage.
AI Short #1: Painted Tiger
AI Short #2: Splashes
AI Short #3: Out of the picture
Artificial Intelligence Fair, Unbiased, Responsible Use
Responsible AI means reducing harm, improving transparency, and preserving human dignity in outcomes.
Max Sanna uses AI technologies with a clear sense of responsibility, ensuring that innovation supports people, protects privacy and contributes to sustainable progress. AI can help improve efficiency, reduce repetitive work and support better decisions, but it must be applied with transparency, fairness and appropriate human oversight. Max Sanna believes that technology should enhance human judgment rather than replace accountability. In practice, this means using data carefully, respecting confidentiality, checking outputs for accuracy and bias, and choosing AI solutions that are secure, reliable and proportionate to the task. Max Sanna also recognizes that digital technologies have an environmental footprint, so Max Sanna aims to use AI thoughtfully and efficiently, focusing on real value rather than unnecessary complexity. By combining innovation with ethics, safety and sustainability, Max Sanna seeks to make AI a positive force for business, society and the natural environment.
Fair, Unbiased, Responsible AI Use Checklist:
- Test for bias across different user groups and languages before production release.
- Do not use AI outputs as final truth without domain review and fact verification.
- Explain where AI is used and label generated content when appropriate.
- Collect only necessary data and avoid using sensitive data without explicit legal basis.
- Build escalation paths so users can challenge automated decisions.
- Continuously monitor quality drift, safety issues, and misuse patterns.
Common Artificial Intelligence Limitations and Risks
AI can be powerful, but it can also produce plausible errors or incomplete reasoning.
Hallucinations
Models may confidently generate incorrect facts, references, or calculations.
Bias Propagation
Training data imbalance can lead to unfair output behavior if not actively managed.
Prompt Sensitivity
Small wording changes can produce major output differences and inconsistency.
Artificial Intelligence Privacy and Security
Treat AI inputs and outputs with the same care as any other sensitive business data.
- Mask or remove personal identifiers before sending prompts to external systems.
- Prefer regional hosting and enterprise controls for regulated workflows.
- Store logs with retention limits and access controls.
- Run red-team style tests for prompt injection and data leakage scenarios.
AI Compliance: European, International, and USA Acts
Max Sanna tries to stick to Artificial Intelligence compliance requirements differ by jurisdiction and use case. The acts below are commonly referenced in governance programs.
European Union
AI Act: Regulation (EU) 2024/1689 (risk-based AI rules across the EU).
GDPR: Regulation (EU) 2016/679 (personal data protection and automated decision safeguards).
Digital Services Act: Regulation (EU) 2022/2065 (platform transparency and risk obligations).
Data Act: Regulation (EU) 2023/2854 (data access and sharing obligations relevant to AI ecosystems).
International Frameworks
Council of Europe Framework Convention on AI (2024): treaty framework on AI, human rights, democracy, and rule of law.
UNESCO Recommendation on the Ethics of AI (2021): global ethical governance guidance.
OECD AI Principles (2019): internationally used principles for trustworthy AI policy.
ISO/IEC 42001:2023: AI management system standard used in compliance and audit programs.
United States
Colorado AI Act (SB24-205): state-level duties for high-risk AI developers and deployers.
NYC Local Law 144: requirements for bias audits of automated employment decision tools.
FTC Act Section 5: enforcement against unfair or deceptive AI-related practices.
CCPA/CPRA (California): consumer privacy rights affecting AI data processing and profiling.
Illinois BIPA: biometric data obligations that can affect AI vision and identity systems.
Useful Artificial Intelligence Links
Here is a list of some useful Artificial Intelligence links.
Disclaimer
Notice: This web site utilize AI-generated content, which could occasionally contain AI hallucinations.