There's a ton going on in the AI space, but if it's associated with meme lord Elon Musk, we naturally pay attention. The billionaire's Artificial Intelligence startup xAI has just officially launched its large language model into the wild wild web.
Grok was unleashed in chatbot form last year, only accessible with a Premium+ subscription on X (formerly Twitter, as you already know yet we somehow still feel obliged to mention). Now, it's available on GitHub under Apache License 2.0. Which allows commercial use, modification and distribution; albeit without liability or warranty.
Developers, researchers, maybe enthusiasts with enough Internet knowledge and a supercomputer can build on Grok-1 and directly impact how the xAI updates future versions of the model. Base model weights and network architecture have been released, but without its training code. Which simply implies users can't see what Grok learnt from ...but to say it's text data from X wouldn't be too much of a stretch.
Created by the team involved in molding OpenAI's ChatGPT (more on that later), one thing Grok had going was access to real-time data on X. While that's live information, it is also a source highly susceptible to inaccuracy.
Grok-1 is currently "not fine-tuned for specific application such as dialogue". Yet, it's modeled after Douglas Adams' Hitchhiker’s Guide to the Galaxy as a cheeky alternative to relatively serious rival models from OpenAI (GPT-4), Meta (LlLaMa 2), Google (Gemini, Gemma2B/7B) and others.
Grok has two modes you can toggle—'fun' and 'regular'. No points for guessing which is default. If that wasn't enough to drive the point home, its site spells out that "Grok is designed to answer questions with a bit of wit and has a rebellious streak, so please don’t use it if you hate humor!"
And if you're wondering how Musk never fails to come up with alien-sounding names and what Grok means, we can answer the latter. It's slang for intuitive understanding, or establishing rapport.
Musk's stance on open sourcing AI is something he has been very verbal about. The most notable company he criticised being OpenAI; which the businessman initially helped fund and previously co-found, but eventually sued for breaching an agreement about keeping it non-profit. The proprietor subsequently revealed emails claiming that Musk was well aware of the plans and was more about wanting it folded into Tesla.
ANYWAY. Making generative AI tools free-for-all in the name of democracy has always been iffy waters. There's always the risk of abuse in the hands of unpredictable, fallible human beings. With fake news already its own monster, Gen AI sometimes feels like the steroids to its hulking mass.
All things considered, it's definitely a buzz-worthy announcement about exciting tech. But as the safest way any common folk can conclude about the consequences—only time will tell.
