Databricks Acquires MosaicML to Democratize AI with Accessible Tools for Building, Owning, and Protecting Generative AI Models
Last month, data processing giant Databricks announced a deal to acquire generative AI platform MosaicML for approximately $1.3 billion.
The acquisition aims to enable enterprises to "build, own, and protect world-class generative AI models while maintaining control over their data." Databricks CEO Ali Ghodsi emphasized the goal of democratizing AI, making the "Lakehouse the best place to build generative AI and LLMs (large language models)."
MosaicML is renowned for its advanced MPT massive-scale language model and offers enterprises an economical and efficient way to quickly build and train models using their own data. As a company, MosaicML also claims to provide competitively priced services that are on par with open-source leaders such as LLaMA and Falcon.
MosaicML's Offerings
Founded in 2016 by Stanford University electrical engineer Naveen Rao and Harvard graduate Hanlin Tang, MosaicML has raised a total of $37 million in funding. Its latest funding round took place on January 1, 2023.
MosaicML provides developers with various options to leverage its platform, including APIs for easy integration with frontend applications and customization of models using their own data.
Developers can also use MosaicML's tools to train custom models from scratch and utilize third-party tools like LangChain on top of their custom models, providing flexibility and ownership of complete models.
Like its competitors, MosaicML offers its technology as a rental service, but it sets itself apart by providing customers with code. This allows customers to run the code on their own hardware, ensuring the confidentiality of their data. This approach attracts enterprise customers who prioritize data privacy, as the value of AI systems largely depends on the training data used.
The company emphasizes the importance of open-source models, especially in industries dealing with sensitive data, where local customization and control over models are crucial. MosaicML focuses on serving specific industries such as healthcare and banking, offering the ability to securely interpret and summarize large amounts of data.
The company claims to make its technology accessible to all organizations at a significantly lower price, up to 15 times cheaper than its competitors. Important clients such as AI2, Generally Intelligent, Hippocratic AI, Replit, and Scatter Labs utilize MosaicML for various generative AI use cases.
MosaicML's MPT-30B LLM is a model with 300 billion parameters, and the company claims its quality surpasses OpenAI's GPT-3 while requiring fewer parameters, making it easier to run on local hardware and more cost-effective.
MosaicML utilizes a mechanism called "FlashAttention" for faster inference and training speeds, making it more efficient than Falcon and LLaMA. Additionally, the design of MPT-30B optimizes its performance on deep learning GPUs, adapting to the limitations of actual hardware.
Furthermore, MosaicML claims that MPT-30B performs on par with LLaMA and Falcon in terms of performance. The model requires fewer computational resources during training while providing similar results, especially in encoding tasks.
Databricks' Motivation
Through several strategic moves, Databricks has found a strong position in the market. The launch of LakehouseIQ, the acquisition of MosaicML, and the development of Unity Catalog put Databricks in a favorable position to maintain its market position and compete for incremental market share.
Over time, MosaicML's platform will be integrated and expanded to provide a unified platform where customers can "build, own, and protect their generative AI models." For Databricks, acquiring MosaicML is a strategic move aimed at providing enterprises with user-friendly and cost-effective tools to easily build large language models (LLMs) using their own data.
By integrating this process into Databricks' broader toolchain and workflows, the company aims to reduce the costs associated with training and running LLMs. This strategy recognizes the market demand for cost-effective, finely-tuned professional LLMs for specific tasks. While general-purpose LLMs will continue to exist, Databricks sees an opportunity to meet the demand for customized solutions. Both Snowflake and Databricks are actively committed to providing enterprise-grade governance and intellectual property protection as part of their professional LLMs products.
The combination of Databricks' Lakehouse platform with MosaicML's technology will provide customers with a "simple, fast way to retain control, security, and ownership of their valuable data at a low cost." MosaicML claims that its automatically optimized model training speeds are 2 to 7 times faster than standard methods.
Source: https://analyticsindiamag.com/why-databricks-acquired-mosaicml/
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.