Google and Meta's Ground Breaking AI
and the fall of ChatGPT?
Jelani Abdus-Salaam
February 20, 2024
Issue:
This AI Is Faster Than ChatGPT
AI’ers, welcome back! Did you know on Feb 19, 1990, Adobe released Photoshop 1.0.
Crazy to think that in less than 40 years AI is now generating images, editing images, and producing creatives from our wildest dreams.
Here’s What’s On The Menu For Today
✨ Gemini Is Coming For ChatGPT
🤖 Meta Announces New Training Model
📰 News Hot Off The Press
💰️ Newly Added AI Tools
Today’s Featured Tool
🖥️ RenderNet
RenderNet is an AI-powered platform designed to generate perfect poses and consistent characters for various creative projects. It caters to a wide range of artistic needs, from generating detailed character portraits to creating dynamic scenes.
Not sponsored or affiliated
Google Is Ready For War
✨ Gemini Is Coming For ChatGPT
Does anyone use Gemini? This is a question that we get often.
What was once called Google Bard, Gemini is Google’s answer to ChatGPT.
They launched their model a bit later than ChatGPT and at first, it was nowhere near as good…
Until recently.
Danilo3009 said it best.
Finally, Google is a real player in the AI future, I was skeptical when Bard was released, then somehow satisfied with Gemini 1.0, but Gemini 1.5 is very exciting, well done.
Gemini has announced tons of new updates in its 1.5 version.
Here are some of the most notable things we believe make it a strong contender to compete with ChatGPT.
1/ Context Length: 1 Million tokens
The model can process up to 1 million tokens to understand vast information from text, images, and videos.
2/ Large PDF Upload:
1.5 Pro can analyze, classify, and summarize extensive content effortlessly.
For example, a 402-page transcript from Apollo 11 mission to the moon took 30s to analyze and answer the input of “ Find 3 comedic moments. List quotes from this transcript.
3/ Multimodal Prompt
Gemini 1.5 Pro was presented with a drawing and asked, "What moment is this?"
So if a user uploaded an image, , Gemini would be able to analyze and report back with accuracy.
4/ Analyzing Long 44-minute Videos:
The Model accurately analyzes silent Buster Keaton movies, identifying plot points, events, and small details.
This used over 690,000 tokens. Which is still below the 1 million token limit. Insanity.
5/ Complex Code Base
The Model processed 100,633 lines of codes of Three.js.
6/ Translation
Model tested its translation skill on the Machine Translation from One Boo (MTOB) benchmark.
It achieved results akin to a person who studied English to Kalamang translation using a grammar manual. Kalamang is a language spoken by fewer than 200 people worldwide.
Unfortunately, 1.5 is not available to the public yet, but this announcement marks a huge stride in Google cementing its dominance in the AI space.
If you want to try out Gemini 1.0 head to aistudio.google.com
Video = AI Superpowers?
🤖 Meta Announces New Training Model
Lech was spot on in 2022 when he said “The next big thing in advancing AI capabilities: models acquiring a vast amount of knowledge about the physical world by “watching videos.
Two years later Meta announced their new model V-JEPA, a method for teaching machines to understand and model the physical world by watching videos.
The models can understand and predict what is going on in a video, even with limited information.
It learns by predicting missing or obscured parts of a video in its internal feature space. Unlike generative approaches that fill in missing pixels, this flexible approach enables up to 6x improvements in training and sample efficiency.
If you’re an AI geek like we are, check out more details about the project here.
Hot Off The Press
📰 Latest News
New Additions
Tools Added To The Database
UltraAI - an all-encompassing AI command center for your product. As a comprehensive platform, it offers a range of features designed to enhance and optimize your Language Learning Machine (LLM) operations.
BasedGPT - a chatbot designed for interactive conversations with users. It is built to leverage the power of machine learning to deliver engaging interactions that feel fluid and authentic.
SylphAI - an artificial intelligence tool designed to aid users in conducting people searches, helping them to connect more efficiently with others.
Searchbase - an advanced search tool that is designed to effortlessly integrate with existing databases and provide AI-enhanced search capabilities.
Hiccup App - an AI-powered application designed as a personal digital ally for managing thoughts, ideas, and memories. It serves as a private notebook and diary, a reminder for forgotten things, and a space for crafting ideas.
Visioncast - an application available for download on devices such as the iPhone, iPad, and iPod touch. It offers a unique experience, allowing users to manifest specific outcomes or events.
Pact Monster - an AI tool designed to enhance meeting productivity by consolidating various meeting-related aspects in one platform.
Startup Guru - a GPT designed to provide support and guidance in building a startup.
AdGen AI - Generate 100+ ad variations from a single URL in minutes, focusing your time on strategy and effective A/B testing.
Adsby - an AI-powered platform that simplifies the entire Google Ads process.
Sharpe Labs - an institutional-grade crypto super app, serving as a unified frontend for intelligence, tracking, and investing in digital assets.
Zenfetch - transforms your web content into an AI-powered personal search engine and assistant, making it easy to save and converse with your articles, videos, and PDFs.
Powder - Transform lengthy streams into short, shareable social clips with AI.
Circleback - helps teams get the most out of every meeting by automatically writing meticulous notes and taking action on insights that matter.
See more of the newly added tools here.
That’s it for today.
Until next time, stay safe.
-Best of AI Team
Note: This issue was written in collaboration with Perplexity, Grammarly, and MidJourney, and edited by a human. Welcome to the future.
What did you think of this issue? |