- Starter AI
- Posts
- Llama 3 has arrived, Microsoft’s VASA-1, LLMs for social experiments
Llama 3 has arrived, Microsoft’s VASA-1, LLMs for social experiments
Behold! The power of the Llama.
Sthefania
April 22, 2024
In partnership with
Hello, Starters!
As more models get upgraded, it gets difficult to forecast which one will be the one to finally win the AI race. However, it seems that Meta is pretty confident in their efforts to get first place.
Here’s what you’ll find today:
Llama 3 is now available
Microsoft introduces VASA-1
Social experiments can also be conducted via LLMs
Hugging Face’s new benchmark for health tasks
AI keeps you in touch with Salvador Dalí
And more.
🦙 Llama 3 is now available (2 min)
Meta has unveiled its long-awaited Llama 3, which already stands out for being among the most powerful open-source models and the largest released by the company. The model has two versions, one with 70B and the other with 8B parameters. In several benchmarks, the 8B version performed better than Google's Gemma 7B and Mistral 7B.
Currently, the company is training another version with 400B parameters, but there is no estimated release date. Llama 3 is the model in charge of powering Meta AI, the new chatbot that will be deployed across Meta's offerings.
🖼️ Microsoft introduces VASA-1 (1 min)
Microsoft is succeeding in converting static images into fluid videos with VASA-1. This framework utilises Visual Affective Skills to transform a photo and an audio clip into virtual AI avatars capable of talking and displaying expressions and movements that match the provided audio.
The method outperforms similar models by delivering high-quality videos while maintaining the realism of the portrait. By its nature, this AI model is not available to the public and is only presented as a research demonstration to prevent its misuse by bad actors.
MIT and Harvard researchers have developed a novel technique that enables large language models to autonomously generate and evaluate social science hypotheses. By utilising Structural Causal Models (SCMs), the LLM-driven agents simulate human behaviour and analyse it without human intervention.
In essence, the language model acts as both researcher and research subject, designing and conducting experiments. This setup allows real researchers to intervene and modify decisions at any stage.
🤗Hugging Face has created a new leaderboard to measure the performance of GenAI models on a series of medical-related tasks. This benchmark aims to ensure the safe deployment of medical models before they go live as vital tools.
🦞The Dalí Museum in Florida trained a large language model to answer visitors' questions through a copy of the artist's famous "Lobster Telephone." The data about Salvador Dalí has been sourced from OpenAI's GPT-4, and voice samples have been obtained from archival interviews, resulting in a very convincing dupe.
85% of all AI Projects Fail, but AE Studio Delivers
If you have a big idea and think AI should be part of it, meet AE.
We’re a development, data science and design studio working with founders and execs on custom software solutions. We turn AI/ML ideas into realities–from chatbots to NLP and more.
Tell us about your visionary concept or work challenge and we’ll make it real. The secret to our success is treating your project as if it were our own startup.
⚡️Quick links
What did you think of today's newsletter? |
📄 Social experiments can also be conducted via LLMs (1 min)