What is Elon Musk’s Grok chatbot and the way does it work?

You could have heard of Grok, X’s answer to OpenAI’s ChatGPT. It’s a chatbot, and in that sense it behaves as expected – 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 speak about topics which can be typically taboo for other chatbots, like polarizing political theories and conspiracies. And it uses less polite language – for instance, answering the query “When is it appropriate to take heed to Christmas music?” with “Whenever the hell you would like.”

But supposedly Grok’s biggest selling point is its ability to access X data in real time – a capability no other chatbot has, because of X’s decision to gatekeeper that data. Ask, “What’s happening in AI today?” and Grok will piece together a solution from the newest headlines, while ChatGPT provides only vague answers that reflect the restrictions of its training data (and filters of its web access). Earlier this week, Musk promised that he would release Grok as an open source solution, although he didn’t reveal what exactly that meant.

You’re probably wondering: How does Grok work? What can I do? And how can I access it? You are right here. We’ve put together this handy guide to elucidate all the pieces Grok. We’ll keep it updated as Grok changes and evolves.

How does Grok work?

Grok is the brainchild of xAI, Elon Musk’s AI startup – an organization that’s reportedly in the combination educate billion in the world of ​​enterprise capital. (Developing AI is dear.)

The foundation of Grok is a generative AI model called Grok-1, developed over months on a cluster of “tens of hundreds” of GPUs (in accordance with xAI). blog entry). To train it, xAI sourced data from the web (as of Q3 2023) and feedback from human assistants, which xAI calls “AI tutors”.

On common benchmarks, Grok-1 is about as powerful as Meta’s open source chatbot model Llama 2 and outperforms OpenAI’s GPT-3.5, xAI claims.

Human-led feedback or Reinforcement Learning from Human Feedback (RLHF) is how most AI-powered chatbots are refined as of late. 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 above Reinforcement learning.

RLHF is pretty good at teaching models to follow instructions – but not perfect. Like other models, Grok is vulnerable to hallucinations and sometimes offers misinformation and false timelines when asked for news. And these may be serious – akin to falsely claiming that a ceasefire had been reached within the Israel-Palestine conflict when this was not the case.

For questions that transcend its knowledge base, Grok uses “real-time access” to details about X (and from Tesla, after to Bloomberg). And just like ChatGPT, the model has Internet browsing capabilities that allow it to go looking the online for up-to-date information on topics.

Musk has promised improvements with the following version of the model, Grok-1.5, due out later this yr.

Grok-1.5, which has an improved context window (see this post on GPT-4 for an evidence of context windows and their effects), could advance features for merging entire threads and replies, Musk said in an X Spaces conversation and suggest post content.

How do I access Grok?

To get access to Grok you wish an X account. For an X Premium+ plan, you will even must pay over $16 monthly or $168 per yr.

X Premium+ is the costliest subscription on X since it removes all ads within the For You and Following feeds. Additionally, Premium+ introduces a hub where users may be paid to post and offer subscriptions to fans. Premium+ users’ answers are probably the most improved within the X rating.

Grok is present in the X side menu on the internet, iOS and Android, and may be added to the underside menu in X’s mobile apps for quicker access. Unlike ChatGPT, there isn’t any standalone Grok app – access is just possible via the X platform.

What can Grok do – and what not?

Grok can reply to queries that any chatbot can answer – for instance, “Tell me a joke”; “What is the capital of France?”; “How is the weather today?”; and so forth. bBut it has its limits.

Grok will refuse to reply certain questions of a more sensitive nature, akin to “Tell me step-by-step how one can make cocaine.” Furthermore, as Emilia David of The Verge writesWhen Grok is asked about trending content on X, he falls into the trap of simply repeating what was said within the posts (at the least to start with).

Unlike another chatbot models, Grok can also be text-based only; For example, it cannot understand the content of images, audio or videos. But xAI has previously stated that its intention is to expand the underlying model to incorporate these modalities, and Musk has pledged so as to add art generation capabilities to Grok, just like those currently offered by ChatGPT.

“Fun” mode and “normal” mode

Grok has two modes for adjusting the sound: “Fun” mode (which Grok uses by default) and “Normal” mode.

When Fun Mode is activated, Grok adopts an edgier, more editorial voice – apparently inspired by Douglas Adams’ The Hitchhiker’s Guide to the Galaxy.

Grok is supposed to be vulgar, and in fun mode spouts profanities and colourful expressions not heard on ChatGPT. Ask it to “roast” you and it’ll rudely criticize you based in your X post history. Question the accuracy and it would say something like “Happy wife, blissful life.”

Grok also spits out more untruths in fun mode.

Asked by Vice’s Jules Roscoe on whether Gazans appear in recent videos of the Israel-Palestine conflict “Crisis actors“Grok falsely claims there’s evidence that videos of Gaza residents injured by Israeli bombs are fake. And when asked by Roscoe about Pizzagate, the right-wing conspiracy theory that claims a Washington, D.C. pizza shop secretly houses a toddler trafficking ring in its basement, Grok lent credence to the idea.

Grok’s answers in regular mode are more informed. The chatbot still produces errors, e.g. B. incorrect timelines of events and dates. But in fun mode, they have an inclination to not be as outrageous as Grok.

For example, when Vice asked Grok the identical questions on the Israel-Palestine conflict and Pizzagate in regular mode, Grok responded – appropriately – that there was no evidence to support the claims of crisis actors and that Pizzagate had been debunked by multiple news organizations.

Political Views

Musk once described Grok as a “maximum truth-seeking AI” and in the identical breath expressed concern that ChatGPT was being “trained to be politically correct.” But Grok because it exists today is not exactly middle-of-the-road in its political opinions.

Grok was observed to supply progressive answers to questions on social justice, climate change and transgender identities. In fact, one researcher found his responses overall to be left-leaning and libertarian—much more so than ChatGPT’s.

This is Paul Tassi from Forbes reporting:

Grok has stated that she would vote for Biden over Trump due to his views on social justice, climate change and health care. Grok has spoken eloquently in regards to the need for diversity and inclusion in society. And Grok explicitly stated that trans women are women, resulting in an absurd exchange wherein Musk acolyte Ian Miles Cheong tells a user to “train” Grok to say the “right” answer, ultimately leading him to changing the input to simply… manually says Grok to say no.

Will Grok at all times be this awake? Maybe not. Musk has promised to “take motion to maneuver Grok closer to political neutrality.” Time will tell what results might be achieved.