- Ben's Bites
- Posts
- Will Open Source win?
Will Open Source win?
PLUS: Big Open Source efforts and job safety from an economic view.
Hey folks, today we’ve got RedPajama releasing a 1.2 trillion-word dataset model, MiniGPT-4 with multi-model functionality (tell me what’s in this image) and an interesting economic case for AI not stealing our jobs (phew).
The Ben’s Bites x Every Hackathon kicks off today; build something related to AI. It’s that straightforward! Join the discord to take part, Dan and I will be hosting on the Stage at 12 am ET.
p.s. today’s links were found on Ben’s Bites News.
From our sponsor:
Discover BetterLegal Assistant - the AI-powered Chrome extension for decoding legal jargon.
Effortlessly understand contracts, terms of use, and more. For a fun twist, activate Pirate Mode and navigate the legal seas with a swashbuckling touch! Make legal documents accessible to all.
🤌 Our Picks
Highlights if you've only got 2 minutes
1/
You’ve gotta love the determination of open-source to go toe-to-toe with Big Tech - it reminds me of all that GameStop stuff a few years back! And RedPajama is the latest initiative looking to drive this forward with bold goals. They have a 3-step plan to rival the big dogs with fully open-source, reproducible, leading language models: 1) Produce a dataset, 2) Train a suite of base models, and 3) Implement instruction-tuning.
Today they announce the completion of stage 1 - a 1.2-trillion-word dataset mimicking that used to train LLaMA. With promises of fully-trained base models released in a matter of weeks and apparent progress on instruction tuning with OpenChatKit, I can feel the trembling of something big in the pipeline.
2/
Believe it or not, someone’s actually managed to deliver something faster than OpenAI. These researchers at King Abdullah University have unveiled a chatbot feature that OpenAI has been promising for a while now: the ability to answer questions about specific images.
Called MiniGPT-4, it’s built on top of something called BLIP-2 as well as a LLaMA-derived language model. (You can’t read anything AI-related these days without being reminded of those long necks and strange faces!) For those of you wanting to get really stuck in, they plan to release a 7B parameter model which might be just the one for you
3/
Time to get a little bit philosophical and a tad economics-ish on you guys - this is a great blog post on the impact of AI on jobs! The Not Boring post uses two real-life economic quirks/principles/whatever to make the argument that the supply of AI will unlock a ‘latent’ demand for services, whereas increasing abilities simply means that people will demand far higher speed and quality of services.
Within the framework of the whole impact-on-jobs debate: added demand means jobs would be far less likely to get displaced as the required output will jump so much. Designers will have to design much better and much faster; people will sue far more people as legal docs become easier to draw up; and consultants will be expected to consult much more efficiently.
🛠️ Cool Tools
Product launches, updates and demos
PromptBot - Simplify the process of creating detailed, powerful prompts for GPT-based bots & APIs. (link)
AutoGPT by SamurAI - Let AI agents operate autonomously to complete tasks for you in your browser. (link)
Llama Lab - A repo dedicated to building cutting-edge projects using LlamaIndex. (link)
ThinkGPT - Agent techniques to augment your LLM and push it beyond its limits. (link)
WebscrapeAI - Collect data from the web without the hassle of manual scraping. (link)
Single Prompt AI - A collection of single-purpose AI interfaces that do one thing and one thing only. (link)
Trivia - Automatically create questions on any topic for a quick trivia. (link)
Kidgeni - AI art generator for kids. (link)
Chat with a kinda fucking rude Clippy. (link)
🤓 Miscellaneous
News, podcasts, videos, blogs etc
Developers are looking for creative ways to build AI-powered chatbot assistants. (link)
The monkey's paw, Norbert Wiener, and the alignment problem. (link)
How much can I do with AI with a single sentence? (link)
Superchat’s new AI chatbot lets you message historical and fictional characters via ChatGPT. (link)
Adobe brings Firefly to its video tools. (link)
The AI-powered VC - Yohei Nakajima, creator of Baby AGI and co-founder of Untapped Capital. (link)
Community is the future of AI - Blog post from Stack Overflow’s CEO + user opinions on it. (link)
Exploring the goods and bads of AutoGPT. (link)
DINOv2 - A self-supervised vision transformer model from Meta. (blog, project page)
AI-powered medical knowledge - Revolutionising care for rare conditions. (link)
Accelerating Hugging Face transformers with AWS Inferentia2. (link)
Synthetic data from diffusion models improves imagenet classification. (link)
LLaVA - Visual instruction tuning. (link)
DeepMind developed an AI program that created a 3D mapping of all 200 million proteins known to science. (link)
The AI revolution: Google's artificial intelligence developers on what's next in the field. (link)
Low-code LLM - Visual programming over LLMs. (link)
🎓 Learn
How-to’s and resources
👋 Too many links?! I created a database for all links mentioned in these emails. Refer 1 friend using this link and I'll send over the link database.
📰 Unclassifieds
Short, sponsored links
How to build an AI chatbot by Dan Shipper. 10% off checkout with coupon BENSBITES.
Wonderchat - Build ChatGPT-powered chatbots in 5 minutes from a link to your website or PDF files.
Tap into Driven Data's largest community (100k people) to hire your ideal data scientist. (link)
Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 80k subscribers?
Advertise
🖼 AI Images of the day
Funny memes and pics from around the web
Send this to 1 AI-curious friend and receive my AI project tracker database! Use this link.
How was today's email? |
Reply