A week ago Elon Musk he had taken the opportunity to reiterate that artificial intelligence represents one of the greatest threats to humanityechoing the fears of some experts and technology executives that systems could emerge that escape human control and make decisions that endanger our existence.
Today, confirming the predictions, the visionary entrepreneur and CEO of Tesla and SpaceX has an artificial intelligence chatbot unveiled created by his new artificial intelligence company, xAI. Grok is ready to challenge established systems like ChatGPTand to bring a unique blend of humor and wit to the world of generative AI. Modeled on the cult sci-fi comedy by British author Douglas Adams The Hitchhiker’s Guide to the GalaxyGrok promises to respond even to the spiciest questions that other AIs might escape.
Grok’s personality: a hint of rebellion and wit
Described as having “a little spirit” e “a rebellious streak“, Grok is more than just generative artificial intelligence. Musk himself shared a screenshot on X (formerly Twitter) in which he asked Grok for one recipe for cocaine. The four steps in the answer include “get a degree in chemistry” and “set up a clandestine laboratory in a remote location.” However, the chatbot adds at the end that he is joking and urges people not to try to produce cocaine because it is illegal and risky.
Powered by a broad language model called Grok-1Musk’s AI chatbot can, in real time, access information via the X platform, a great advantage over other systems, which allows it to provide users with up-to-date answers. For the South African tycoon, it is about best artificial intelligence “currently existing”.
In a first round of tests based on middle school math problems and coding activities with Python, Grok It outperformed all other models in its class calculation, including GPT-3.5, the one used in the free version of ChatGPT, and Inflection-1. However, it is lagged behind GPT-4the most powerful ChatGPT model because it is trained with significantly more data and computing resources.
The limits of Musk’s chatbot and who can use it
While Grok’s abilities are impressive, xAI remains transparent regarding the potential limitations of this technology. Like any Large Language Model (LLM), the startup explains, Grok can still generate false or contradictory information. And users need to be aware of this.
Attualmente, Grok it is in its early beta phase, with only two months of training behind him. It is available to a select group of users in the US for testing before a wider release.
Interested users can sign up for a waiting list for a chance to use the bot. Musk also revealed that Grok will eventually be integrated in the X Premium+ subscription servicewhich costs $16 a month under an “early access program.”
Grok’s origins and mission are intertwined with xAI
The name “Grok” comes from the 1961 science fiction novel di Robert A. Heinlein Stranger in a Strange Land. It is a Martian term with no direct terrestrial translation, but critics have established that they understand it to mean “understand deeply and intuitively“. A concept in line with xAI’s mission to build artificial intelligence that promotes collective understanding of the universe.
xAI launched in July with a team made up of former employees of organizations such as OpenAI, DeepMind and others. The company’s mandate is to create AI that addresses people of all backgrounds and political opinions. Musk has previously expressed concern about AI systems becoming overly “politically correct” and aims to offer an alternative. Grok is said to be a means to test the AI approach “in public”.
to know more: Artificial Intelligence: what it is and what it can do for us