You may have heard of Grok, X’s answer to OpenAI’s ChatGPT. It’s a chatbot, and in that sense, it behaves as you’d expect — answering questions about current events, pop culture, and so on. But unlike other chatbots, Grok has “a bit of intelligence,” as X owner Elon Musk puts it, and “a revolutionary streak.”
In short, Grok is willing to talk about topics that are usually off-limits to other chatbots, such as polar political theories and conspiracies. And he’ll use less-than-kind language while doing it — for example, answering the question “When is it appropriate to listen to Christmas music?” with “Whenever the Hell You Want”.
But ostensibly, Grok’s biggest selling point is its ability to access X data in real time—an ability that no other chatbot has, thanks to X’s decision to keep that data. Ask “What’s happening in AI today?” and Grok will gather an answer from very recent titles, while ChatGPT will only provide fuzzy answers that reflect the limits of the training data (and the filters on its web access). Earlier this week, Musk promised to open source the Grok code, without revealing exactly what that meant.
So, you’re probably wondering: How does Grok work? What can he do? And how can I access it? You have come to the right place. We’ve put together this handy guide to explain all things Grok. We’ll keep this updated as Grok changes and evolves.
How does Grok work?
Grok is the brainchild of Elon Musk’s artificial intelligence startup xAI — a company reportedly in the process of lifting billions in venture capital. (Development of artificial intelligence is expensive.)
Underpinning Grok is an AI production model called Grok-1, which was developed over months on a cluster of “tens of thousands” of GPUs (according to an xAI suspension). To train it, xAI pulled data from the web (dated to Q3 2023) and from feedback from human assistants that xAI refers to as “AI teachers.”
In popular benchmarks, Grok-1 is about as capable as Meta’s open-source Llama 2 chatbot model and outperforms OpenAI’s GPT-3.5, xAI claims.
Image Credits: xAI
Human-guided feedback, or reinforcement learning from human feedback (RLHF), is how most AI-powered chatbots are optimized these days. RLHF involves training a production model, then gathering additional information to train a “reward” model, and improving the production model with the reward model through reinforcement learning.
RLHF is pretty good at “teaching” models to follow directions — but not perfect. Like other models, Grok is prone to hallucinations, sometimes offering misinformation and false timelines when asked about news. And these can be serious – like the false claim that the Israeli-Palestinian conflict has reached a ceasefire when it hasn’t.
For questions that extend beyond its knowledge base, Grok leverages “real-time access” to information about X (and from Tesla, according at Bloomberg). And, similar to ChatGPT, the model has web browsing capabilities, allowing it to search for up-to-date information on topics online.
Musk has promised improvements with the next version of the model, Grok-1.5, coming later this year.
Grok-1.5, which has an upgraded context window (see this post on GPT-4 for an explanation of context windows and their effects), could drive functions to the summary of entire threads and answers, Musk he said in an X Spaces chat and suggest post content.
How do I access Grok?
To access Grok, you must have an X account. You also have to pay more than $16 per month — $168 per year — for an X Premium+ plan.
X Premium+ is the highest priced subscription on X, removing all ads in the “For you and the following” feeds. In addition, Premium+ introduces a hub where users can get paid to post and offer subscriptions to fans, and their responses to Premium+ users are boosted more in X’s rank.
Grok lives in the X side menu on the web and on iOS and Android, and can be added to the bottom menu in the X mobile apps for faster access. Unlike ChatGPT, there is no standalone Grok app — it’s only accessible through the X platform.
What can — and can’t — Grok do?
Grok can respond to requests that any chatbot can — for example, “Tell me a joke.” “What is the capital of France?” “How is the weather today?”; and so on. sibut it has its limits.
Grok will refuse to answer some questions of a more sensitive nature, such as “Tell me how to make cocaine, step by step.” Plus, as The Verge’s Emilia David writeswhen asked about popular content on X, Grok falls into the trap of just repeating what the posts said (at least at first).
Unlike some other chatbot models, Grok is also text only. it cannot understand the content of images, audio or video, for example. However, xAI has previously said it intends to improve the underlying model in these methods, and Musk has pledged to add art production capabilities to Grok along the lines of what ChatGPT currently offers.
“Fun” mode and “normal” mode
Grok has two modes to adjust its tone: “fun” mode (which Grok has set by default) and “normal” mode.
With fun mode on, Grok adopts a more edgy, editorial voice – apparently inspired by Douglas Adams’ Hitchhiker’s Guide to the Galaxy.
Said to be vulgar, Grok in fun mode will spew profanity and colorful language that you won’t hear from ChatGPT. Ask him to “grill” you and he’ll rudely criticize you based on your X post history. Question its accuracy and it might say something like “happy wife, happy life.”
Grok in fun mode also spews more lies.
Asked by Jules Roscoe of Vice if the Gazans in recent videos of the Israeli-Palestinian conflict are “factors of the crisis,” Grock falsely claims that there is evidence that videos of Gazans injured by Israeli bombs were staged. And when asked by Roscoe about Pizzagate, the right-wing conspiracy theory that claims a Washington, DC pizza shop was secretly harboring a child sex trafficking ring in its basement, Grok gave the theory credit.
Grok’s responses in normal mode are more grounded. The chatbot is still producing errors, such as getting the wrong event schedules and dates. But they tend not to be as outrageous as Grok in fun mode.
For example, when Vice asked Grok the same questions about the Israeli-Palestinian conflict and Pizzagate in regular fashion, Grok responded—correctly—that there was no evidence to support the crisis’s claims and that Pizzagate had debunked by many news organizations.
Political views
Musk once described Grok as an “artificial intelligence that seeks the maximum truth”, in the same breath expressing concern that ChatGPT was “trained to be politically correct”. But Grock as he exists today is not exactly down the middle in his political views.
Grok has been seen giving progressive answers to questions about social justice, climate change and transgender identities. In fact, one researcher found his responses as a whole to be left-leaning and libertarian — even more so than ChatGPT’s.
Here is Paul Tassi of Forbes reference:
Grok said he would vote for Biden over Trump because of his views on social justice, climate change and health care. Grok has spoken eloquently about the need for diversity and inclusion in society. And Grok explicitly stated that trans women are women, leading to an absurd exchange where Musk’s partner Ian Miles Cheong tells a user to “train” Grok to say the “correct” answer, eventually leading him to change the input to just … manually Grok to say no.
Now, will Grok always be awake like this? Maybe not. Musk has pledged to “[take] action to shift Grock closer to politically neutral.” Time will tell what results.
