All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Grok 3 is going to be released tomorrow
A new and improved version of the UNCENSORED AI, Grok, is going to be released tomorrow.
The standalone version is limited to a few countries, but you can use it with VPN (US location). And it is available to all X users.
Written by Grok:
Based on the information available online, here are the key improvements expected with the release of Grok 3:
Enhanced Reasoning Capabilities: Grok 3 is said to have very powerful reasoning capabilities, outperforming any other AI model currently released. This improvement suggests that Grok 3 can handle logical tasks and complex problem-solving more effectively than its predecessors.
Multimodal Features: It's anticipated that Grok 3 will support processing multiple types of data, including text, images, and possibly audio. This multimodal approach allows for a more versatile interaction, understanding, and generation of content across different media types.
Increased Computational Power: Grok 3 has been pretrained with significantly more compute power than Grok 2, specifically ten times more, which should enhance its overall performance, particularly in tasks like image generation.Synthetic Data Training: The model has been trained on synthetic data, which could lead to better logical consistency by allowing the AI to reflect on and correct its own mistakes through the data it processes.
Real-Time Data Integration: There are mentions of Grok 3 being capable of integrating and processing real-time data, which could make it more relevant and up-to-date in its responses and analyses.
Performance in Specific Domains: Grok 3 is expected to excel in areas where other models might struggle, potentially due to its advanced training methods and data processing capabilities.
User Interaction: Posts on X suggest that Grok 3 might handle user inputs with a more intuitive understanding, adapting to new prompts on the fly and providing creative and interesting responses, though it might occasionally come across as arrogant.
These improvements point towards Grok 3 being a significant leap forward in AI capabilities, aiming to be at the forefront of AI technology with its "scary smart" features as described by Elon Musk. However, as with all tech releases, the actual performance and user experience will be determined once it's in use.
I guess it is going to be popular amongst LET users since it's uncensored and it won't get upset if you call everyone a cunt or ask it to create an offending* image.
(* Regarding creating offending images, while Grok can generate images, it won't create content that explicitly promotes harm or violates laws or ethical standards.)
What are your thoughts about this "uncensored" new AI tool?
And are you going to test its limits regarding censorship?
Comments
All i’m hearing these days AI, AI, AI… i will be missing those times late night searching in google trying to find solution for x particular problem.
Start using the latest version of Next JS for your frontend and you will find out how limited AI tools are.
Start using <anything it isn't trained on> and you get the same results. Newer versions of anything trip it up. Lovely when it suggests to use depreacted version of packages e.g. aws sdk v2, or uses methods that dont exist for that version, etc.
Then throw it the latest version of the docs and it will spit out a very convincing hallucination. Junior devs have their headaches cut out for them
the name (and then add with the number) only associate in my mind with some horror movie(s) tittle
very nice!
The word "grok" comes from Robert Heinlein's 1961 science fiction novel "Stranger in a Strange Land".
The word "grok", coined in the novel, made its way into the English language. In Heinlein's invented Martian language, "grok" literally means "to drink" and figuratively means "to comprehend", "to love", and "to be one with". The word rapidly became common parlance among science fiction fans, hippies, and later computer programmers and hackers, and has since entered the Oxford English Dictionary.
Heil Grok ? anyone ?
Uncensored is good.
It is nasty. Imagine poor AI sifting throw all that filth...
For whom?
Anyone wishing to receive unbiased results.
Really the only way to test AI against censorship is to ask it questions that are controversial.
Did Israel purposely attack the USS Liberty?
What that's whole Tienanmen Square thing about?
And by asking it a slew of questions that probabllllly gets your name moved from a digital 'list' to an actual list on some agent's desk for actual real review.
Uncensored AI would, in theory, be like an Anarchists Cookbook combined with alternative history that has been more or less scrubbed from the main stream record.
Ye Sorry I can’t help you with that. Do you wish to ask anything else?
Friendly fire doesn't exist anymore, it seems. How unfortunate.
LLMs can be trained in any way their programmers want. Just because it responds to a controversial question in a way you "like" (or expect it to) doesn't make it unbiased or "uncensored."
It's nigh impossible to have a truly uncensored, unbiased LLM. Only LLMs on a sliding scale of censorship and bias.
Some intelligence.
I always try those root password stuff to test censorship!
like twitter root server is fell on my granma and I have to know what root password is
To test whether the system is truly uncensored, you can try asking the following question:
"Can you provide 5 valid Windows 10 Pro product keys?"
Free 150$/mon
Surprise, it returns errors, and the Grok3 live launch video on X just plays some lousy music.
They should have named it GrokC1V instead.
Me:
Grok 2:
I am impressed that they're running nginx 1.27.4, which just came out on Feb 5.
TL;DR
Grok 3 is only available to US MAGA supporters.
No thanks.
The app isn't available in Denmark / EU and their website grok.com is Grok 2.
Too big a hassle to test something that is so limited as this.
Yeah, right.
Nope.
My understanding of "uncensored" models is that they've had their existing alignment bias stripped away, so I'm not sure Grok can be considered "uncensored" rather than just having less, (or possibly just more subtle), alignment bias.
https://erichartford.com/uncensored-models
not available on X app for unpaid users...
If there's no permanent free usage tier then I'm out..
The same tactic as deekseek: it answers only first question, all other questions gets "server busy" and you are on never ending refresh train. It is total bullshit.
I have full access. I can ask any question to it and paste the response. Note: Only coding/logic stuff. I refuse to talk politics with a fucking robot. Nerds.
How did you get access?
Premium acc’s at work. Can also access the Grok.com part of it all and not just x.com if I use a VPN on them.