- The Rundown AI
- Posts
- ChatGPT gets a memory!
Welcome, AI enthusiasts.
Ever feel like ChatGPT is a bit… Forgetful? So did OpenAI.
A new memory feature will let ChatGPT learn and remember details across chats — paving the way for more natural, personalized assistants. Let’s explore…
In today’s AI rundown:
OpenAI tests new ChatGPT memory capabilities
Nvidia releases local RTX chatbot for PCs
How to use ChatGPT-V to describe objects on the go
Cohere’s massively multilingual Aya model
6 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
OPENAI
Image source: OpenAI
The Rundown: OpenAI just launched a new experiment to select users, enabling ChatGPT to remember key details from prior chats, aiming to provide more helpful and personalized responses in future conversations.
The details:
Users can control what's remembered, with options to delete specifics, wipe memory, or disable the feature.
ChatGPT can pick up on preferences, contexts, instructions, and other relevant info to improve over time and remember details across chats.
GPTs will also be able to leverage distinct memory capabilities when enabled by their builders.
The experiment is rolling out to a small percentage of both free and plus users this week, with a broader rollout expected at a later date.
Why it matters: Features like memory are key stepping stones towards more natural chatbot interactions that better understand personal contexts while still maintaining user control. Personalized and tailored AI assistants are inching closer every day!
TOGETHER WITH AE STUDIO
The Rundown: Accelerate your company's success with AE Studio's proven approach to building high-impact custom software and AI solutions.
Turbocharge your product development and save hundreds of hours
Have industry experts build your custom software and AI solutions
Hire AE to pinpoint what you should be building and where to find ROI
Book your FREE strategy session with AE Studio today to learn more and see why AE is the go-to partner for startups and enterprises.
NVIDIA
Image source: Nvidia
The Rundown: Nvidia just released Chat with RTX, a free chatbot demo that can run leading open-source models locally on Windows RTX PCs, with the ability to analyze process responses based on a user's own files and data.
The details:
Chat with RTX can query documents, videos, and other local content for fast, relevant answers without needing the internet.
The chatbot keeps data on device rather than relying on cloud services, enhancing privacy.
The app can also integrate data from YouTube videos and playlists by simply providing a URL.
You can download NVIDIA Chat with RTX here if you use a Windows RTX PC or workstation.
Why it matters: Nvidia’s making serious power moves with Chat with RTX — firing shots at closed, online models like ChatGPT while pumping RTX sales and breaking into the consumer chatbot space with local PC integration.
AI TRAINING
Live from San Francisco
The Rundown: With the impressive speed and accuracy of GPT-4 Vision, you can now use your iPhone to get descriptions of places and objects by simply double-taping the back of your phone.
Hat tip to Alvaro Cintas for this creative idea.
Step-by-Step:
Add this GPT-4 Shortcut to your iPhone.
Generate and copy a new OpenAI API key.
Add your Open AI key to the GPT-4 shortcut.
On your iPhone, go to Settings —> Accessibility —> Touch —> Back Tap —> Double Tab —> GPT-4 Vision shortcut.
You are all set! With the camera open, double-tap the back of your phone to receive a description of the object.
PRESENTED BY COVALENT
The Rundown: Covalent provides instant access to serverless GPUs like H100s, A100s, L40s, and RTXs, entirely in Python. No reservations needed – only pay for what you use.
With Covalent, you get:
AI infrastructure with a single Python decorator
Only pay for the resources you consume
No reservations or minimum spend required
Sign up today and get $20 in free credits.
AI RESEARCH
Image source: Cohere
The Rundown: Cohere researchers just introduced Aya, a multilingual AI model that expands capabilities across 101 languages, 50% of which are lower-resourced — significantly expanding coverage while maintaining effectiveness.
The details:
Aya was trained on 2.5 times more data, covering over twice as many languages as prior multilingual AI models.
Despite significantly greater language coverage, Aya outperformed comparable models like mT0, BLOOMZ, and Bactrian-X across benchmarks.
Extensive data weighting and pruning balances representation across higher and lower-resourced languages, expanding beyond data-rich languages.
Why it matters: By greatly expanding language coverage, Aya works towards democratizing access to AI capabilities beyond dialects with the most data for training — while still maintaining high performance. The days of universal translation and global access to AI chatbots are inching closer.
NEW TOOLS & JOBS
🚀 BuildShip- Streamline app development with AI-powered backend workflows and APIs (link)
✨ Chat Ur Data GPT- Boost productivity and unlock data insights with AI-driven assistance (link)
📅 Howie.ai- Say goodbye to scheduling hassles with advanced AI coordination (link)
📝 Superlist- Create to-do lists, capture thoughts or detailed notes, assign tasks to teammates, and more (link)
📊 Ffoward AI- Unlock insights and patterns in user interviews and build a better product roadmap (link)
⚙️ Sheet Copilot- Effortlessly automate your Google Sheets workflow with AI (link)
QUICK HITS
Andrej Karpathy, one of the founding members of OpenAI and top AI researchers worldwide, has departed from OpenAI.
Nous Research introduced Hermes 2 Llama-2 70B, its largest model trained on over 1M synthetic data entries, also featuring advanced prompt formats for enhanced dialogue interaction.
Sam Altman spoke at the World Government Summit, revealing insights on GPT-5 capability expectations, AI regulation, and more.
GitHub announced a new cohort for its Accelerator program, dedicating $40,000 in funding and resources to projects making open-source AI advancements.
Researchers unveiled AutoMathText, a new strategy that improves the mathematical reasoning of AI models by autonomously evaluating and selecting high-quality content.
ElevenLabs is expanding its Speech to Speech feature into 29 different languages, allowing for multilingual custom voices with maintained emotion and consistency.
Dozens of protestors gathered outside of OpenAI’s San Francisco office to object to the development of AGI and the company backtracking on its military affiliation stance.
The University of Pennsylvania became the first Ivy League institution to offer an undergraduate degree in Artificial Intelligence.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 450k+ 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