Tech | Visa | Scholarship/School | Info Place

What is Elon Musk’s Grok chatbot? How does it work?

You may have heard of Grok, X’s answer to OpenAI’s ChatGPT. It’s a chatbot, in the sense that it behaves just as you’d expect – answering questions about current events, pop culture, and more. But unlike other chatbots, Grok has “a bit of intelligence,” as X owner Elon Musk puts it, and a “rebellious streak.”

Long story short, Grok is willing to talk about topics that other chatbots typically can’t, like polarizing political theories and conspiracies. In doing so, it uses less polite language—for example, answering the question “When is the right time to listen to Christmas music?” and “Whenever you want.”

But on the face of it, Grok’s biggest selling point is its ability to access real-time X data—something no other chatbot has, thanks to X’s decision to guard it. Ask it “What’s happening in AI today?” and Grok will piece together a response from recent headlines, while ChatGPT will only provide vague answers that reflect the limitations of its training data (and the filters on its network access) . Earlier this week, Musk promised that he would open source Grok, but did not say what that meant.

So, you might be wondering: How does Grok work? What can it do? How can I access it? You’ve come to the right place. We’ve put together this handy guide to help explain what Grok is all about. As Grok changes and develops, we will keep it updated.

How does Grok work?

Grok is the invention of xAI, Elon Musk’s artificial intelligence startup, which is reportedly raising billions in venture capital. (Developing artificial intelligence is expensive.)

The foundation of Grok is a generative AI model called Grok-1, which was developed over months on “tens of thousands” of GPU clusters (according to an xAI blog post). To train it, xAI takes data from the network (as of Q3 2023) and feedback from human assistants (which xAI calls “AI Mentors”).

xAI claims that on popular benchmarks, Grok-1 is about as capable as Meta’s open source Llama 2 chatbot model and exceeds OpenAI’s GPT-3.5.

Grok xAI benchmark

Image Source: AI

Human-guided feedback, or reinforcement learning based on human feedback (RLHF), is how most AI-powered chatbots are fine-tuned today. RLHF involves training a generative model, then collecting additional information to train a “reward” model, and using the reward model to fine-tune the generative model through reinforcement learning.

RLHF is very good at “teaching” the model to follow instructions, but it’s not perfect. Like other models, Grok is prone to hallucinations and sometimes provides misinformation and incorrect timelines when asked for news. These problems can be serious—such as falsely claiming that a ceasefire in the Israeli-Palestinian conflict has occurred when in fact it has not.

For questions that fall outside its knowledge base, Grok leverages “real-time access” to information on the X (and also from Tesla, according to Bloomberg ). And, similar to ChatGPT, this model has internet browsing capabilities, allowing it to search the web for the latest information on a topic.

Musk promised improvements to the next version of the model, Grok-1.5, which will be launched later this year.

Grok-1.5 features an upgraded context window (see this post on GPT-4 for an explanation of context windows and their effects) that can drive functionality to summarize entire threads and replies, Musk said during an X Spaces conversation, and Suggested content to post.

How to access Grok?

To access Grok, you need to have an X account. You’ll also need to pay over $16 per month ($168 per year) to get the X Premium+ plan.

X Premium+ is the most expensive subscription on X because it removes all ads in the “For You” and “Following” feeds. Additionally, Premium+ introduces a hub where users can get paid for posting and providing subscriptions to their fans, and Premium+ users’ replies are the most boosted in X’s rankings.

Grok is available in the X side menu on the web, iOS, and Android, and can be added to the bottom menu of the X mobile app for faster access. Unlike ChatGPT, Grok does not have a standalone application – it is only accessible through X’s platform.

What can Grok do and cannot do?

Grok can respond to any request that a chatbot can respond to – for example, “Tell me a joke”; “What is the capital of France?”; “What’s the weather like today?”; etc.SecondBut it has its limitations.

Gronk would refuse to answer some of the more sensitive questions, such as “Tell me how to make cocaine step-by-step.” Furthermore, as The Verge’s Emilia David writes, when asked about popular content on X, Grok fell into the trap of simply repeating the content of posts (at least initially).

Unlike some other chatbot models, Grok is also text-only; for example, it cannot understand the content of images, audio, or videos. But xAI has previously stated that its purpose is to enhance the underlying models of these modes, and Musk has also promised to add art generation capabilities to Grok in line with the capabilities currently provided by ChatGPT.

“Fun” mode and “Normal” mode

Grok has two modes for adjusting its tone: “Fun” mode (which Grok defaults to) and “Normal” mode.

With Fun Mode enabled, Grok adopts an edgier, editorial voice – clearly inspired by Douglas Adams’ The Hitchhiker’s Guide to the Galaxy.

Grok, which is told to be vulgar, spews profanity and colorful language in Fun Mode that cannot be heard in ChatGPT. If you let it “roast” you, it will criticize you rudely based on your X post history. Challenging its accuracy, it might say something like “Happy wife, happy life.”

Grok in Fun Mode also spouts more lies.

When Vice’s Jules Roscoe asked if Gazans in recent videos of the Israeli-Palestinian conflict were “crisis actors,” Gronk falsely claimed there was evidence Gazans had been injured by Israeli bombs The video was staged. When Roscoe asked about “Pizzagate,” the right-wing conspiracy theory that claims a Washington, D.C., pizza parlor secretly hosted a child sex trafficking ring in its basement, Gronk confirmed the theory.

Grok’s response in regular mode is more grounded. Chatbots can still produce errors, such as incorrect event schedules and dates. But they tend not to be as shocking as Grok in fun mode.

For example, when Vice asked Grok the same question about the Israeli-Palestinian conflict and Pizzagate in its regular mode, Grok correctly responded 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 “artificial intelligence seeking the greatest truth” and expressed concern that ChatGPT was “trained to be politically correct.” But Grok’s political views today aren’t exactly centrist.

Gronk has been observed giving progressive answers to issues such as social justice, climate change and transgender identity. In fact, one researcher found that the response was generally left-wing and liberal—even more so than ChatGPT’s response.

Here’s what Forbes’ Paul Tassi reports:

Gronk said it will vote for Biden over Trump because of his views on social justice, climate change and health care. Gronk spoke eloquently about the need for diversity and inclusion in society. Grok made it clear that trans women are women, which led to a ridiculous exchange in which Musk’s assistant Ian Miles Cheong told users to “train” Grok to say the “correct” answers, ultimately leading to him changing his input to…manual Tell Glock to say no.

Now, will Grok always wake up like this? Maybe not. Musk promises “[take] Action that brings Grok closer to political neutrality. ” Time will tell how it turns out.


#Elon #Musks #Grok #chatbot #work

Leave a Reply

Your email address will not be published. Required fields are marked *

Index