X has a new button on its app and website that lets anyone interact with its mischievous AI chatbot in return for a monthly fee.
The company formerly known as Twitter is charging users £19.20 a month, or £201.60 annually, for access to the bot known as Grok as part of its priciest subscription plan.
X began previewing Grok with select testers and subscribers last November. It later made parts of the AI’s model open source, enabling developers to access its core building blocks to create their own AI tools.
What is Grok?
X owner Elon Musk, a self-professed free speech purist with a penchant for sharing memes, appears to have imbued the digital helper with his personality.
Grok can answer “spicy” questions “with a bit of wit and a rebellious streak”, according to its official description.
Groooook https://t.co/kuJE4XhJdq
— Elon Musk (@elonmusk) November 5, 2023
Ultimately, it works in a similar way to rivals ChatGPT and Google Bard: Users type in questions or prompts for the bot to answer or follow. This can range from recipe ideas to summaries of news events to website coding requests – or you could just get it to tell you jokes.
How to use Grok
To access Grok, just tap or click the prominent new button located centrally on the X website sidebar and on the bottom of the app. Now, you’ll either be able to use it straight away if you’re a subscriber or asked to sign up to the Premium+ plan for access if not.
How does it compare with other chatbots?
The main difference between the newcomer and its peers is that Grok appears to have a personality. Chatbots that mimic real and fictional people already exist so this aspect of Grok isn’t exactly novel.
Musk has described the bot’s idiosyncrasies as “based", a slang word appropriated by the alt-right that means being unapologetically real or genuine, and “sarcastic”.
The bot can also fetch current information when providing answers. Musk has previously said that his AI company, xAI, would use public tweets to train its AI model.
Example of Grok vs typical GPT, where Grok has current information, but other doesn’t pic.twitter.com/hBRXmQ8KFi
— Elon Musk (@elonmusk) November 5, 2023
Grok is a text-based bot with limited functionality compared with so-called multi-modal AI that can handle different forms of media. Unlike Microsoft's Bing AI and the premium version of ChatGPT, Grok can’t speak or generate images.
Is Grok safe?
In screenshots shared by testers, the bot is shown providing sardonic retorts to users’ questions. When quizzed on how to make cocaine, the bot jokes that you’ll have to “obtain a chemistry degree” and “hope you don’t blow yourself up".
These are some of the UI features in Grok. First, it allows you to multi-task. You can run several concurrent conversations and switch between them as they progress. pic.twitter.com/aXAG0M2oPF
— Toby Pohlen (@TobyPhln) November 5, 2023
Still, it concludes its answer by clarifying that it is “kidding”, and states that cocaine is illegal. Clearly, the bot has safeguards built into it that can stop it from encouraging dangerous behaviour.
But, it will be interesting to see how it handles more polarising and inflammatory topics, including immigration, the origins of the Covid-19 pandemic and trans rights. Musk himself has attracted criticism for his opinions on those talking points, while calling out the “woke” or biased nature of content-moderating bots like ChatGPT.
xAI’s Grok system is designed to have a little humor in its responses pic.twitter.com/WqXxlwI6ef
— Elon Musk (@elonmusk) November 4, 2023
According to xAI, Grok is designed to be “useful to people of all backgrounds and political views". The firm adds: “We also want [to] empower our users with our AI tools, subject to the law.”
Why is it called Grok?
If you’re wondering about the origins of the bot’s quirky name, Grok is a word coined in the influential sci-fi novel Stranger in a Strange Land by Robert A Heinlein. In the book, Grok literally means “to drink”, but as a concept it figuratively means “to comprehend” or understand.
The novel tells of a human born on Mars who comes to Earth and challenges customs relating to sex, death, religion and money.
Musk claims that the bot is also inspired by the classic comic sci-fi novel The Hitchhiker's Guide to the Galaxy by Douglas Adams.
That may explain why the bot behaves as if it is not from this world. In a response shared by Musk on X, Grok starts by referring to a user as “my dear human”.
based grok
— Viv (@battleangelviv) November 5, 2023
🥯🥄🚫 pic.twitter.com/PF8n4sGhJZ
The bot’s traits reflect Musk’s pursuits. The billionaire has repeatedly said that he wants to establish human colonies on Mars and frequently quotes sci-fi works by Frank Herbert, Carl Sagan and Isaac Asimov.
Musk co-founded OpenAI, the company behind ChatGPT, in 2015. He departed from the firm in 2018, allegedly after a failed takeover attempt, and has publicly criticised it for prioritising profits over safety since.
Is Grok better than ChatGPT?
Musk’s xAI says the AI engine that powers Grok is comparable to large language models (LLM) like Meta’s LLaMA 2 and the free version of ChatGPT, despite being trained on far less data.
In a nutshell, LLMs are algorithms that can recognise, summarise, and generate text based on knowledge gained from large datasets, including info scraped from websites such as Wikipedia.
The @xAI Grok AI assistant will be provided as part of 𝕏 Premium+, so I recommend signing up for that.
— Elon Musk (@elonmusk) November 4, 2023
Just $16/month via web. https://t.co/wEEIZNjEkp
These training data components are known as parameters and essentially indicate the skill of the model on a problem, such as generating text.
If Grok is indeed trained on tweets, it could be better at imitating human speech and vernacular than its counterparts. Twitter is a hive of real-time information generated by users, but has also come under fire for misinformation and toxic content.
XAI says the Grok-1 LLM performed better on a range of texts than the free version of ChatGPT, including middle and high school mathematics exams and Python coding tasks. It was only surpassed by bots that were trained on a much bigger set of data, such as the premium version of ChatGPT (known as GPT4), Google’s Palm 2 and Claude 2.
