guglielmoi.bsky.social

Tell me you are using AI to generate your ads without telling me you aren't 🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣

A great article that goes deep into the current #risks with #agentic #AI. While the author speaks from a SaaS standpoint, this analysis applies to any solution that relies on AI agents. A recommended reading for anyone who's planning to venture into this wonderful world. 😈

AI Agents Problems No One Talks About. Memory like a goldfish, goals like a broken GPS, and feedback loops that would make your worst coworker look brilliant.

"Think of them like that brilliant but widely unreliable colleague who generates amazing ideas randomly in meetings. You’d never trust them to close the deal alone. They’re input providers; you should own the output (and outcome)." 👇
ai.gopubby.com/ai-agents-pr...
#aiagent #risks

Today only: Everything half + liveProjects and liveVideos $10 + Annual Pro Subscriptions for $199.99 😍
This includes the MEAP of my latest book
👇
mng.bz/YZra
@manningbooks.bsky.social
#MemorialDayWeekend #sales #AI #DataScience #SoftwareDevelopment

“It’s a lot easier to destroy a thing than to build a thing.”

Domain-Specific Small Language Models Bigger isn’t always better. Train and tune highly focused language models optimized for domain specific tasks.</b> When you need a language model to respond accurately and quickly about a specific fi...

If you want to learn more about #opensource #smalllanguagemodels and #ollama, you can consider reading my book 👇
mng.bz/YZra
@manningbooks.bsky.social

GitHub - ymichael/open-codex: Lightweight coding agent that runs in your terminal Lightweight coding agent that runs in your terminal - ymichael/open-codex

This is why I ❤️ #opensource: shortly after whatever commercial #AI tool is released, an OS option is available. #opencodex CLI is a fork of the original OpenAI Codex CLI with expanded support for multiple providers (including #ollama) and models
#typescript #python #llm
github.com/ymichael/ope...

One of the most important skills for a machine learning engineer or data scientist to have is being able to know when NOT use AI.

Before you even think about trying to use AI to solve a problem, you need to first try to solve the problem without AI.

#datasceince

Ollama's new engine for multimodal models · Ollama Blog Ollama now supports new multimodal models with its new engine.

#Ollama now supports multimodal models via Ollama’s new engine 🥳
#visual #slm #genai
ollama.com/blog/multimo...

Running an #AI #Agent locally and using only #opensource components gives you more control, but still some implementation hurdles can be met. Please be patient, as this is still a work in progress (I have started it last week only). Please give it a star if you like it 🙏 5/5
#python #smolagents

This work follows few principles:
- Everything stays local (a scientist laptop, or a local server). Nothing must leave the user's organisation network.
- #opensource parts only.
- #smalllanguagemodels only.
- User Experience is based on the BYOP (Bring Your Own PDB) and MD expertise principle :)
4/5

I have started a #github repo, where I am building a local #agent with the goal of moving away all the annoying work from the scientists that use #gromacs, so that they could focus on their main tasks, where they can apply in full their domain knowledge and expertise. 3/5
github.com/virtualrambl...

#gromacs has plenty of applications in Material Science, #drugdiscovery and #chemical Reaction Modelling. Although being a very powerful tool, it requires a lot of manual work to prepare an end-to-end simulation and analysis pipeline. 2/5

#Gromacs is a popular #opensource software suite for high-performance #moleculardynamics (MD) and output analysis. MD is a computer simulation technique used to study the physical movements of atoms and molecules over time. 1/5
www.gromacs.org

Back here after a while. Many professional and personal things prevented me to handle multiple social networks, across 2024 in particular. I'm back 🙂

Bluesky is officially “open” now (no more invites). So just a reminder to not engage with trolls and bad actors, there is no algorithm here to promote or spread them, replying to them is what puts them on everybody’s timelines. So block and give no oxygen.

Prompt-Free Diffusion - a Hugging Face Space by shi-labs Discover amazing ML apps made by the community

A #HuggingFace space to play with Prompt-Free #Diffusion is also available: t.co/SEmZ94iVlE
3/3
#Python #opensource #AI

GitHub - SHI-Labs/Prompt-Free-Diffusion: Prompt-Free Diffusion: Taking "Text" out of Text-to-Image D... Prompt-Free Diffusion: Taking &quot;Text&quot; out of Text-to-Image Diffusion Models - GitHub - SHI-Labs/Prompt-Free-Diffusion: Prompt-Free Diffusion: Taking &quot;Text&quot; out of Text-to-Image D...

#SeeCoder is reusable to most public #text2image models as well as adaptive layers like #ControlNet, #LoRA, T2I-Adapter, etc. Inputs to the model are a reference image and a depth map, a canny edge map, a scribble or a openpose.
The official GitHub repo: t.co/RkEDS0RQLD
2/3
#AI #opensource

What if we take text out of #text2image #Diffusion models? Prompt-Free Diffusion relies only on visual inputs to generate new images, handled by Semantic Context Encoder (#SeeCoder) by substituting the commonly used CLIP-based text encoder.
The paper: arxiv.org/abs/2305.16223
1/3
#AI

Prompt engineering means that the human adapts to the computer. Artificial intelligence was once thought of differently, wasn't it?

Actually some can run also on a MacBook Air.

Hugging Face launches open source AI assistant maker to rival OpenAI’s custom GPTs The arrival of Hugging Chat Assistants shows how fast the open source community continues to catch up to closed rivals such as OpenAI.

This is huge: #HuggingFace launches #opensource #AI assistant maker to rival OpenAI’s custom GPTs.
venturebeat.com/ai/hugging-f...

Llama 2 powered by #ONNX:
an optimized version of the #Llama 2 model:
github.com/microsoft/Ll...

#GenerativeAI #LLM