Image: © Koshiro/Stock.adobe.com
xAI claims the new multimodal version of Grok exceeds leading AI models in terms of text reading and real-world understanding, but it is still unclear when – or if – Grok will make a mark that can rival models such as ChatGPT.
Elon Musk’s AI-focused start-up is racing to compete with various models and has announced an update for its ChatGPT challenger – Grok.
The company shared details of Grok 1.5 Vision – its first “multimodal model” that will give the chatbot the ability to process visual information including diagrams and other images. The start-up said this version of Grok will be available soon to early testers and existing Grok users.
The new preview comes less than a month since xAI unveiled Grok 1.5, which the start-up claims has superior coding and math capabilities to its predecessor, Grok-1.
How does it rank?
xAI claims Grok 1.5 Vision is able to beat various leading AI models in certain areas, such as mathematics, text reading and real-world understanding.
“We are particularly excited about Grok’s capabilities in understanding our physical world,” xAI said in a blogpost. “Grok outperforms its peers in our new RealWorldQA benchmark that measures real-world spatial understanding.”
The company said this benchmark is designed to evaluate the “real-world spatial understanding capabilities of multimodal models”. The examples include various images with multiple-choice questions, such as an image of a road and a question of the correct route a car can turn in its current lane.
“While many of the examples in the current benchmark are relatively easy for humans, they often pose a challenge for frontier models,” xAI said.
The company has shared an initial – and rather small – release of this benchmark, which contains more than 700 images that each include a question. xAI said it will expand on this benchmark as its multimodal models improve.
“In the coming months, we anticipate to make significant improvements in both capabilities, across various modalities such as images, audio and video,” xAI said.
How has Grok compared so far?
While it is unclear how well the latest model will compare when put under the pressures of public testing, the earlier models have not appeared to make a major impact on the AI sector. The initial test versions have been made available to limited users – mainly premium subscribers of the X social media platform.
xAI previously claimed Grok will answer questions that are rejected by most other AI systems, while Musk has advertised the connection to X as a “massive advantage” over other AI models.
The earlier model of the chatbot was subject to a mix of praise and criticism for its performance – with some being unhappy that its responses don’t align with their political views.
Joseph Thacker, a security researcher with SaaS security company AppOmni, previously said there could be issues with relying on the “immense amount of data” that exists on X, as it contains a variety of harmful content, including false information.
“There’s a lot of toxicity, incorrect information, bias and racism [on X],” Thacker said. “So, while it’s likely to sound very human, there are risks of that surfacing.”
Find out how emerging tech trends are transforming tomorrow with our new podcast, Future Human: The Series. Listen now on Spotify, on Apple or wherever you get your podcasts.
