- The Artificially Intelligent Enterprise
- AI Rewind: Elon Musk's AI Entry Grok, a Crazy Irreverent Chatbot
AI Rewind: Elon Musk's AI Entry Grok, a Crazy Irreverent Chatbot
This week’s most popular posts
When it comes to taking on big challenges, the one person I’d not bet against is Elon Musk.
I recently read his biography, which was a fascinating look into how he works. He’s a wild man juggling Tesla, Twitter, SpaceX, The Boring Company, and now XAI. He’s a one-man wrecking crew.
I recently started playing with Grok, his foray into chatbots to challenge ChatGPT and Google Bard. It’s a remarkably different experience.
Grok is powered by a large language model called Grok-1, which has performed better in some aspects than other AI chatbots like OpenAI’s GPT-3.5.
One of the main advantages of Grok is its real-time access to X (formerly Twitter) posts, which allows it to provide up-to-date information on current events. This gives it an edge over other chatbots that don't have access to this data.
Another yet-to-be-seen possibility is that Musk could license the vision data collected by Tesla’s self-driving cars. This would give the Grok LLM capabilities that are unique over many others but more aptly applied to navigation systems and other real-world applications.
I think this is where it will fit into my toolbox, asking for updates on current events.
I have access to Grok via Twitter. It is designed to answer questions with some wit and humor and is described as having a "rebellious streak" in being willing to answer questions that other AI bots decline. It has a fun mode where you can easily have it insult you, and it is even prompted to be vulgar. Or it can be used in regular mode, where you get more serious responses.
Key features of GrokAI:
Can access real-time information from X (formerly Twitter) to provide up-to-date responses
Aims to surpass the capabilities of models like GPT-4.0 and LLaMa 2
Currently in beta, rolled out exclusively to X Premium+ subscribers
Issues have been reported, such as referencing OpenAI, which the company says was due to training on web data containing ChatGPT outputs.
Overall, GrokAI is positioned as a next-generation chatbot that, for now, is just a companion to Twitter, but the big outcome will be that it can be trained on all the Twitter/X data in real-time, which might make it very smart, or it might start a dumpster fire, the jury is still out.
There’s actually a precedent for this. Early on, in 2016, to be exact, Microsoft used Twitter to train a chatbot. This article at The Verge says it all, Twitter taught Microsoft’s AI chatbot to be a racist asshole in less than a day.