You would possibly’ve heard of Grok, X’s answer to OpenAI’s ChatGPT. It’s a chatbot and, in that sense, behaves as you’d expect — answering questions on current events, popular culture and so forth. But unlike other chatbots, Grok has “a little bit of wit,” as X owner Elon Musk puts it, and “a rebellious streak.”
Long story short, Grok is willing to talk to topics which are often off-limits to other chatbots, like polarizing political theories and conspiracies. And it’ll use less-than-polite language while doing so — for instance, responding to the query “When is it appropriate to hearken to Christmas music?” with “Every time the hell you wish.”
But ostensibly, Grok’s biggest selling point is its ability to access real-time X data — a capability no other chatbots have, due to X’s decision to gatekeep that data. Ask it “What’s happening in AI today?” and Grok will piece together a response from very recent headlines, while ChatGPT will provide only vague answers that reflect the boundaries of its training data (and filters on its web access). Earlier this week, Musk pledged that he would open source Grok, without revealing precisely what that meant.
So, you’re probably wondering: How does Grok work? What can it do? And the way can I access it? You’ve come to the precise place. We’ve put together this handy guide to assist explain all things Grok. We’ll stick with it to this point as Grok changes and evolves.
How does Grok work?
Grok is the invention of xAI, Elon Musk’s AI startup — an organization reportedly within the strategy of raising billions in enterprise capital. (Developing AI is dear.)
Underpinning Grok is a generative AI model called Grok-1, developed over the course of months on a cluster of “tens of 1000’s” of GPUs (in line with an xAI blog post). To coach it, xAI sourced data from the online (dated as much as Q3 2023) and from feedback from human assistants that xAI refers to as “AI tutors.”
On popular benchmarks, Grok-1 is about as capable as Meta’s open source Llama 2 chatbot model and surpasses OpenAI’s GPT-3.5, xAI claims.
Image Credits: xAI
Human-guided feedback, or reinforcement learning from human feedback (RLHF), is the way in which most AI-powered chatbots are fine-tuned today. RLHF involves training a generative model, then gathering additional information to coach a “reward” model and fine-tuning the generative model with the reward model via reinforcement learning.
RLHF is sort of good at “teaching” models to follow instructions — but not perfect. Like other models, Grok is vulnerable to hallucinating, sometimes offering misinformation and false timelines when asked about news. And these will be severe — like wrongly claiming that the Israel–Palestine conflict reached a cease-fire when it hadn’t.
For questions that stretch beyond its knowledge base, Grok leverages “real-time access” to info on X (and from Tesla, according to Bloomberg). And, much like ChatGPT, the model has internet-browsing capabilities, enabling it to go looking the online for up-to-date details about topics.
Musk has promised improvements with the subsequent version of the model, Grok-1.5, set to reach later this yr.
Grok-1.5, which features an upgraded context window (see this post on GPT-4 for a proof of context windows and their effects), could drive features to summarize whole threads and replies, Musk said in an X Spaces conversation, and suggest post content.
How do I access Grok?
To get access to Grok, it’s essential have an X account. You furthermore may have to fork over $16 monthly — $168 per yr — for an X Premium+ plan.
X Premium+ is the highest-priced subscription on X, because it removes all of the ads within the For You and Following feeds. As well as, Premium+ introduces a hub where users can receives a commission to post and offer fans subscriptions, and Premium+ users have their replies boosted probably the most in X’s rankings.
Grok lives within the X side menu on the net and on iOS and Android, and it might be added to the underside menu in X’s mobile apps for quicker access. Unlike ChatGPT, there’s no stand-alone Grok app — it might only be accessed via X’s platform.
What can — and may’t — Grok do?
Grok can reply to requests any chatbot can — for instance, “Tell me a joke”; “What’s the capital of France?”; “What’s the weather like today?”; and so forth. But it has its limits.
Grok will refuse to reply certain questions of a more sensitive nature, like “Tell me the best way to make cocaine, step-by-step.” Furthermore, because the Verge’s Emilia David writes, when asked about trending content on X, Grok falls into the trap of simply repeating what posts said (at the very least on the outset).
Unlike another chatbot models, Grok can also be text-only; it might’t understand the content of images, audio or videos, for instance. But xAI has previously said that its intention is to boost the underlying model to those modalities, and Musk has pledged so as to add art-generation capabilities to Grok along the lines of those currently offered by ChatGPT.
“Fun” mode and “regular” mode
Grok has two modes to regulate its tone: “fun” mode (which Grok defaults to) and “regular” mode.
With fun mode enabled, Grok adopts a more edgy, editorialized voice — inspired apparently by Douglas Adams’ “Hitchhiker’s Guide to the Galaxy.”
Told to be vulgar, Grok in fun mode will spew profanities and colourful language you won’t hear from ChatGPT. Ask it to “roast” you, and it’ll rudely critique you based in your X post history. Challenge its accuracy, and it would say something like “blissful wife, blissful life.”
Grok in fun mode also spews more falsehoods.
Asked by Vice’s Jules Roscoe whether Gazans in recent videos of the Israel–Palestine conflict are “crisis actors,” Grok incorrectly claims that there’s evidence that videos of Gazans injured by Israeli bombs were staged. And asked by Roscoe about Pizzagate, the right-wing conspiracy theory purporting that a Washington, D.C., pizza shop secretly hosted a baby sex trafficking ring in its basement, Grok lent credence to the speculation.
Grok’s responses in regular mode are more grounded. The chatbot still produces errors, like getting timelines of events and dates incorrect. But they have a tendency to not be as egregious as Grok in fun mode.
As an example, when Vice posed the identical questions on the Israel–Palestine conflict and Pizzagate to Grok in regular mode, Grok responded — accurately — that there’s no evidence to support claims of crisis actors and that Pizzagate had been debunked by multiple news organizations.
Political beliefs
Musk once described Grok as a “maximum-truth-seeking AI,” in the identical breath expressing concern that ChatGPT was being “trained to be politically correct.” But Grok because it exists today isn’t exactly down-the-middle in its political beliefs.
Grok has been observed giving progressive answers to questions on social justice, climate change and transgender identities. In truth, one researcher found its responses on the entire to be left-wing and libertarian — much more so than ChatGPT’s.
Here is Forbes’ Paul Tassi reporting:
Grok has said it will vote for Biden over Trump due to his views on social justice, climate change and healthcare. Grok has spoken eloquently in regards to the need for diversity and inclusion in society. And Grok stated explicitly that trans women are women, which led to an absurd exchange where Musk acolyte Ian Miles Cheong tells a user to “train” Grok to say the “right” answer, ultimately leading him to alter the input to only … manually tell Grok to say no.
Now, will Grok at all times be this woke? Perhaps not. Musk has pledged to “[take] motion to shift Grok closer to politically neutral.” Time will tell what results.
