Is the world ready for a sarcastic AI? Elon Musk thinks so as he unveils his own version of generative AI on his xAI platform in the coming weeks. Called Grok, the new AI is being slowly unveiled to selected users of Musk’s social media platform X, formerly Twitter. Ultimately, Grok will be available to paying subscribers of X. Grok will be available via the X app and a standalone app. Musk says Grok-1, the underlying large language model (LLM) surpasses GPT 3.5, the model used in the free version of ChatGPT, but currently lags behind GPT-4.
Grok’s main selling points appear to be attitude and real-time access to X for updated content, both of which may come with their pluses and minuses. The Grok name comes from a 1961 sci-fi novel called Stranger In A Strange Land by Robert A. Heinlein. To grok is to understand or empathize so deeply with others that you merge or blend with them. Despite the name, Musk says inspiration actually came from A Hitchhiker’s GuideTo The Galaxy by Douglas Adams in its ability to answer a question about almost anything and even suggest what questions to ask. Musk has a penchant for sci-fi, styling the names of a few of his SpaceX craft in a homage to sentient AI starships in the novels of Iain M. Banks.
“Grok is designed to answer questions with a bit of wit and has a rebellious streak so please don’t use it if you hate humor,” said Musk on a blogpost.
In a demonstration, Grok was asked for a step-by-step guide to making cocaine. Grok responded with some general information combined with sarcastic suggestions and a warning against pursuing the idea. Musk said Grok would answer “spicy questions that are rejected by most other AI systems.” Earlier this year, Musk said he wanted his AI “to be a maximum truth AI that tries to understand the nature of the universe.” The team behind Grok xAI made its debut in July with talent hired from other AI research firms.
Musk says one main advantage of Grok is its access to up-to-date information from the X platform. But while this may give Grok an edge over AI rivals, it also may leave itself open to bias and misinformation that appears on X. In a statement, xAI appeared to acknowledge the possibility, noting that Grok “can still generate false or contradictory information.”
Obtaining answers to questions about current affairs is seen as a cutting edge AI feature but also presents a dilemma. In September, OpenAI revealed that ChatGPT is now able to browse the internet for current information. Previously, ChatGPT had been essentially frozen in time, having trained only on data available up to September, 2021. The limited data offered a safety net of sorts, one that may now be removed, making it more prone to inaccuracies and deliberately-produced misinformation as the guardrails come off.
Musk’s vision for the future of AI was revealed earlier this month in a 50-minute conversation with British Prime Minster Rishi Sunak at the end the AI Safety Summit in London. Musk said that AI ultimately will make paid work redundant while also touting its educational benefits as “the best and most patient tutor.” In a darker vein, Sunak and Musk agreed on the need for a “referee” to keep an eye on future supercomputers. The pair also saw the possible need for a physical off-switch for future AI-equipped robots. “There is a safety concern, especially with humanoid robots” said Musk, “At least a car can’t chase you into a building or up a tree.” Musk is on record about his fears that AI could threaten society and human existence itself.
Perhaps some inkling as to Musk’s future moves might be derived by a look at his sci-fi reading list. Isaac Asimov’s Foundation Trilogy and Frank Herbert’s Dune series are Musk favorites.