Grok-2, the artificial intelligence chatbot built into and trained on content from X is now out in beta and it is a huge step up from its predecessor, putting it in the upper ranks of leading AI chatbot tools alongside ChatGPT, Claude and Google Gemini.
Soon after its release, Grok-2 found its way into the top five of the LMSys chatbot arena leaderboard. These are human-rated evaluations of leading LLMs and is usually dominated by Google, OpenAI and Anthropic, so it was a big win for Grok-maker xAI.
With the new release Grok-2 was also given a makeover with a look closer to other chatbot interfaces, as well as the ability to generate images using Flux, the AI image generation model from Black Forest Labs that is close to the quality of industry leader Midjourney.
I’ve been playing with Grok-2 for the past few days and have found it as responsive as ChatGPT but with a better sense of humor and the ability to respond to real-time events thanks to X.
Putting Grok-2 to the test
Woah, another exciting update from Chatbot Arena❤️🔥The results for @xAI’s sus-column-r (Grok 2 early version) are now public**!With over 12,000 community votes, sus-column-r has secured the #3 spot on the overall leaderboard, even matching GPT-4o! It excels in Coding (#2),… https://t.co/gqSWSwYN0z pic.twitter.com/j9UYDBYNt4August 14, 2024
Grok-2-mini is available to anyone with a Premium subscription to X. When you first open Grok you'll see the typical "ask" box, a row of suggested ideas and then trending topics from X that Grok can explain or even answer questions on.
1. The ego check
My first prompt was “who is Ryan Morrison” and there is nothing more humbling than searching for your own name when someone notable shares that moniker (in my case the video games attorney, Ryan Morrison). But by adding “AI journalist” I got a solid result.
It pulled in pieces from my bio on X, my bio on Tom’s Guide and other information I’ve posted on the X platform or that others have posted about me — almost universally about AI.
Grok also displayed X posts about or claiming to reference me but only one met that criteria, the others were from random people with the first name Ryan who mentioned AI.
I decided to try the ‘who is’ test with my much better-known boss, Mark Spoonauer, Tom’s Guide Global Editor-in-Chief. It gave me a bullet-list breakdown of his career, editorial philosophy, and X posts. Grok also offered completely random, unrelated X posts.
2. The coding test
I asked Claude to create a simple text adventure game using Python called The Enchanted Forest. In the prompt, I gave it some specifics including a puzzle to solve, player class and commands for movement and taking items.
The prompt: “Create a Python text adventure game called 'The Enchanted Forest' with at least 3 interconnected locations, 2 items to collect, and a simple puzzle to solve. Include a player class with an inventory, a room class, and commands for movement, taking items, checking inventory, and quitting. The game should have a clear win condition. Provide the complete, runnable Python code.”
The code worked well, creating a simple text adventure game I could play in the Terminal on my MacBook.
When I asked Grok-2 to create a version with a user interface rather than just running it in the terminal it threw a series of errors in the amended code. It also couldn’t fix its own errors. Its coding is roughly level with GPT-3.5.
3. Trending topics
One of the most powerful features of Grok, including the new version, is its ability to analyze trending topics and pull in content from across the X platform.
This makes it particularly powerful when it comes to news stories. You can ask it about any current story and it can pull information from X posts and put it into context using its own training data, including X content.
I asked for information on the release of Luma Labs Dream Machine 1.5 and it was able to give me a one paragraph summary as well as X posts showing some example. I followed this up by asking for more examples and specifics from the new model.
It not only gave me a bullet-list breakdown of the new features but also several X posts showing content generated using v1.5 from a range of X users.
Final thoughts
I've previously said that Grok, because of its integration within X, is an incredibly powerful AI search tool. This was more about the X integration than the model itself but with Grok 2 that changes. The new model is on par with ChatGPT or Claude in terms of responsiveness and is much more open and less likely to refuse a request.
With the addition of Flux, Grok-2 also gains the ability to generate images for the first time and it can even do this in connection with a news story, adding a fascinating dimension to examining what is happening in the world.
With version 2 Grok is now not only a real competitor to the bigger AI chatbot platforms, but adds to the argument that for an app to be a true 'everything app' it must have a viable AI integration, including with live data access to bring everything together.
