TruthGPT = Grōk is here, and it tells the truth 👀
Plus: Leaked new features in ChatGPT 🔥
Hi folks!👋🏻 This is The Prompt!
Here’s what we have for you today:
Elon Musk announced Grōk, a funny AI that always tells the truth.
Leaked ChatGPT features reveal custom agent training capability
Midjourney released Style Tuner to save styles for AI generated photos
Runway's latest upgrade can create lifelike videos
What the executive order means for openness in AI
and more! Let’s get it!
Elon Musk announces Grōk: like GPT-3.5 with real data & humor
Elon Musk and Twitter (X) announced Grōk, an AI that’s designed to answer questions with a bit of wit and has a rebellious streak.
How’s it different from ChatGPT?
Grok has real-time access to info on 𝕏 - huge advantage, because it doesn’t have a knowledge cut off!
It will also answer spicy questions that are rejected by most other AI systems. Their goal is to design AI tools that are useful to people of all backgrounds and political views.
So basically, the rumors for TruthGPT turned out to be true.
Early results and performance?
The engine powering Grok is Grok-1, their frontier LLM, which was developed over the last four months. Over this short period of time they achieved capabilities similar to LLaMa 2 70B, and slightly better than GPT3.5.
It has 25,000 character context window and API, Image recognition, and Audio recognition is planned to be released soon.
How to get access?
Currently only available for a small set of X premium users in the US, but you can get on the waitlist.
Extra spicy take 🌶️ (and we agree!)
$44 billion was a steal for Twitter
-bot removal to cleanup training data
-increase api price to block others from training
-bookmark button and private likes for better signal
-change sign out experience + rate limits so no scraping
-Grok, first LLm on up to date live data
— Nick Dobos (@NickADobos)
Nov 4, 2023
🚨 What else is going on
Leaked photos from the agent sandbox “Magic Maker” in ChatGPT
Phind's new code generation model beats GPT-4 at coding tasks while running 5x faster at 100 tokens/second.
Midjourney released “Style Tuner” where you can type /tune and your prompt, pick from the sample images MJ creates, and get a code for that particular style to style future images. For example: This code “--style raw-btvMSX6949oRdtpi“ will give a foggy look to your photos created in MJ:
Runway (popular tool for video generation) released new upgrade for their Gen-2 video maker, and the videos are remarkably good. Here’s a video generated with this simple prompt “a lion/black panther in a jungle“
✍🏼 Prompt of the Day
A man sitting in a cafe, cinematic, drinking a coffee, smoke --style 74ws4A0iOlre
What'd you think of today's edition?