- The Rundown AI
- Posts
- Meta takes on ChatGPT
Sign Up | Advertise | Tools | AI University
Welcome, AI enthusiasts.
The open-source AI community has a new champion in Llama 3.
With impressive capabilities, a colossal 400B+ parameter version on the horizon, and integrations across its social empire, Meta clearly means business. Let’s get to it…
In today’s AI rundown:
Meta releases next-gen Llama 3
Microsoft’s VASA-1 brings still photos to life
Personalize the new version of GPT-4 Turbo
GPT-4 nears expert eye doctor analysis
6 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
META
Image source: Midjourney
The Rundown: Meta just introduced Llama 3, the next generation of its open-source LLM — with 8B and 70B parameter versions that surpass open-model competitors across a variety of evaluation benchmarks.
The details:
Meta’s tests show Llama 3 8B and 70B outperforming similarly-sized models like Google's Gemma, Mistral 7B, and Anthropic's Claude 3 Sonnet.
The models were trained on a dataset 7x larger than Llama 2, containing 15T tokens and 4x more code.
The largest 400B+ parameter version is expected to be released in the “coming months“ and competitive with GPT-4.
Llama 3 is embedded across platforms via the Meta AI assistant, available on Facebook, Instagram, WhatsApp, Messenger, and a new meta.ai website.
Why it matters: With impressive benchmarks and a behemoth model waiting to be released, Zuck’s master plan is taking shape. With the average human still having limited exposure to AI, Meta’s broad integrations just freely delivered a top-tier AI model to over 3 billion people across its products.
TOGETHER WITH NETMIND POWER
The Rundown: NetMind Power offers a decentralized AI ecosystem that makes model training and inference accessible and affordable for everyone.
With NetMind’s Power platform, you can:
Train, fine-tune, and deploy your AI models with ease
Access a global network of GPUs without expensive cloud costs
Contribute your idle GPUs to the network and earn rewards
Join us today and get free credits to start your AI project.
MICROSOFT
Image source: Microsoft
The Rundown: Microsoft just unveiled VASA-1, a new AI model capable of generating strikingly lifelike talking head videos from a single still image and an audio clip.
The details:
VASA-1 requires just a single photo and a speech audio file to create a realistic talking video of the subject, with synced mouth movements and emotive animations.
The model can generate nuanced expressions, natural head motions, and even realistic singing performances, going beyond simple lip-syncing.
Users can use input sliders to control aspects of the generated video, such as eye gaze direction, head distance, and emotional tone.
Why it matters: VASA-1 represents a big leap, with potential applications in areas like virtual avatars, gaming, and computer animation. But while just a research demo, the ability to generate insanely realistic deepfakes is clearly here — with significant implications given upcoming elections and broader adoption from bad actors.
AI TRAINING
The Rundown: In this tutorial, you’ll learn how to access the newest version of GPT-4 Turbo, a majorly improved model, and personalize your experience.
Step-by-step:
Go to OpenAI’s Playground and sign in. You will need a ChatGPT Plus subscription
On the left side, click where it says “Chat”.
Click on the models dropdown menu, press “Show more models”, and select “gpt-4-turbo-2024-04-09”.
Enter a prompt where it says “User” and play around with the different parameters, e.g., system prompt, temperature, or maximum length, and personalize your outputs.
PRESENTED BY HIGGSFIELD
The Rundown: Diffuse is a new video creation app that allows users to create personalized, lifelike videos from just a single selfie.
With Diffuse, you can:
Generate dances, emotes or memes with lifelike, realistic movement
Create personalized characters from a single reference image
Bring ideas to life with Prompt Builder, a powerful text-to-video feature
Download Diffuse now from the App Store or Play Store to start creating today.
AI RESEARCH
Image source: Midjourney
The Rundown: A new study from the University of Cambridge found that OpenAI's GPT-4 model performs close to the level of expert ophthalmologists in analyzing eye conditions and suggesting treatments.
The details:
GPT-4, GPT-3.5, PaLM 2, LLaMA, and expert and trainee ophthalmologists were tested on 87 multiple-choice questions.
GPT-4 scored 60/87, higher than the average scores of trainees (59.7) and junior doctors (37) but slightly lower than the expert’s average of 66.4.
While the study is newly published, the testing occurred in mid-2023, meaning LLM capabilities available now likely already exceed the models used.
Why it matters: GPT-4 is already performing near the level of the average expert eye doctor. With additional LLM advances, capabilities will only be pushed higher, massively increasing accessibility to healthcare — especially in settings where access to specialists may be limited (or non-existent).
NEW TOOLS & JOBS
📝 Collato - Smart notetaker transforms notes into actions for busy people
💬 HuggingChat iOS - Chat with open-sourced models and customize your chatbot
👩⚖️ SpeedLegal - Personal AI contract negotiator
🎨 ChatSlide - AII-in-one AI assistant for creating slides and videos from documents
🏥 CompliantChatGPT - AI assistant for HIPAA-compliant healthcare data management
🛍️ Arena Chat - Boost sales with AI chat for your store
💼 Hive - Account Executive
🤝 Cohere - Manager, Business Development
👩💻 Mistral AI - Inference/C++ Backend Engineer
🔍 Anthropic - Recruiter, Research
QUICK HITS
Google announced a restructure to enhance AI product development, creating a new ‘Platform and Devices’ team and merging its Research and DeepMind teams.
Stability AI informed employees that it is laying off 10% of its workforce in order to “right-size” the company, which follows a major leadership change.
Nothing Earbuds released two new headphone models, introducing ChatGPT integration that allows users to interact with the AI via a simple squeeze.
The U.S. Air Force and DARPA announced they have successfully conducted simulated dogfights with AI-controlled aircraft, a significant milestone in AI aviation testing.
JPMorgan Chase CEO Jamie Dimon said he envisions a world where AI significantly enhances banking, with the company embedding the tech across all aspects of the business.
Mentee Robotics introduced Menteebot, a humanoid robot capable of performing complex tasks and learning through natural language commands.
Movie studio A24 is facing backlash after using artwork that appears AI-generated across marketing posters for its new “Civil War” film.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 550k+ AI enthusiasts
Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.
FEEDBACK
How would you rate today's newsletter?Vote below to help us improve the newsletter for you. |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.
Join the conversation