How would you allocate \(10 million for marketing your prominent tech company if you were limited to a \)10,000 budget? Would you consider investing in a Super Bowl commercial or sponsoring an F1 event?
One potential avenue for investment could be in the development of a relational AI design. Conceptual models have gained significant attention lately, with a shift towards enhancing suppliers’ core products and services rather than traditional marketing strategies.
For instance, Databricks recently introduced DBRX, a cutting-edge relational AI model comparable to OpenAI’s GPT series and Google’s Gemini. DBRX comes in two versions: Base (DBRX Base) and fine-tuned (DBRX Instruct), allowing for customization and utilization with public, proprietary, or commercial data sets.
Naveen Rao, the VP of relational AI at Databricks, highlighted the versatility of DBRX, stating that it was optimized for English but can also operate in multiple languages like French, Spanish, and German.
While Databricks describes DBRX as “open source” akin to models like Meta’s Llama 2 and Mistral’s offerings, there is ongoing debate about the true nature of this classification.
The extensive training process for DBRX, costing around $10 million and spanning two months, has resulted in a model that reportedly surpasses existing open source benchmarks. However, accessing and utilizing DBRX may pose challenges for non-Databricks clients due to hardware requirements, such as a minimum of four Nvidia H100 GPUs.
To address these obstacles, Databricks offers the Mosaic AI Foundation Model as a managed solution, enabling clients to leverage DBRX efficiently. Rao emphasized the platform’s potential for custom model development and highlighted the performance advantages of DBRX compared to competitors like Llama 2.
Despite its advancements, DBRX has limitations, including potential inaccuracies in responses and a lack of multimodal capabilities. Rao acknowledged these constraints and emphasized ongoing efforts to enhance the model’s reliability and safety.
While DBRX presents promising advancements in the field of generative AI, it faces stiff competition from established players like OpenAI’s GPT-4. Databricks aims to continue evolving DBRX and exploring new possibilities in generative AI to stay competitive in the market.
In conclusion, the journey ahead for DBRX is challenging, considering the evolving landscape of generative AI technologies and the need to address concerns around data biases and model accuracy. Databricks remains committed to innovation and aims to deliver enhanced versions of DBRX in the future.
