6 Emerging ISV Trends (in Generative AI) from H1 2024
At Zinnov , we've been closely monitoring the rapid advancements in Generative AI, recognizing its transformative impact across various industries. Based on our observations in H1, 2024, here are 6 key trends which are shaping the future of Gen AI Technology:
?1. Private LLMs in Regulated Industries
An increasing number of leading tech companies are forming alliances with Gen AI-native builders to develop private, on-premises LLMs. This trend is prominent in highly regulated sectors such as financial services and healthcare, where data privacy is of paramount importance.
惠普企业服务 selected Aleph Alpha’s LLM Luminous as the foundation for its AI private cloud service, HPE Green Lake for LLMs; 英特尔 partnered with Hugging Face to drive improved performance for machine learning tasks on Intel hardware. It also announced plan to create open platform for enterprise AI in collaboration with Anyscale, Articul8, DataStax, Domino Data, Hugging Face, KX Systems, MariaDB, MinIO, Qdrant, RedHat, Redis, SAP, VMware, Yellowbrick, and Zilliz; Dell Technologies partnered with Hugging Face to make the open-source Gen AI models available on-premises and optimized for Dell infrastructure.
2. AI Innovations at the Edge and development of Smaller AI Models:
The focus on creating smaller, efficient AI models facilitates enhanced on-device processing capabilities, crucial for edge computing applications where quick data processing is essential. This trend is increasingly applied in mobile devices and IoT applications, where rapid data processing and low latency are critical. 苹果 developed ReaLM, a small language model, which is less complex than large language models and runs faster at on-device tasks than larger models; 谷歌 announced an expansion of its Gemma family models. It released Gemma 1.1, Code Gemma language models and PaliGemma, an open-vision model designed to deliver fine-tuned performance on a wide range of vision-language tasks; 微软 developed a family of Phi model which includes Phi-2, a 2.7 Bn-parameter Transformer-based language model, Phi-3, a 3.8 Bn-parameter language model
3. Strategic Mergers and Acquisitions to tackle data integration and governance
Data Compatibility and Governance: To tackle the challenges of data integration and governance, firms are strategically acquiring Startup/ ISVs that are particularly focused on enhancing data observability within AI applications, crucial for maintaining transparent and efficient operational workflows. Databricks acquired Lilac AI to help enterprises improve the quality of data for generative AI applications and language models; Snowflake acquired TruEra to enhance AI observability within the Snowflake AI Data Cloud; Databricks acquired Tabular to provide data compatibility between Delta Lake and Apache Iceberg through UniForm; OneSix acquired Strong Analytics to deliver data engineering, ML, and AI services; DataStax acquired Logspace, creator of the open source package Langflow, to help developers build generative AI applications faster
4. Rise of Serverless Vector Databases for Cost Efficiency and Performance:
领英推荐
The shift towards serverless vector databases is driven by the need to manage large volumes of data efficiently, without the overhead of managing physical servers. These architectures are increasingly used in applications requiring real-time data processing, such as financial trading platforms and real-time health monitoring systems. Pinecone launched Pinecone Serverless, a new enhanced serverless architecture to power its services. This new architecture offers up to 98%? cost reduction; Elastic launched Search AI Lake, a cloud-native architecture optimized for real-time, low-latency applications including search, RAG, observability, & security. This application also powers the Elastic Cloud Serverless offering, which removes operational overhead to automatically scale & manage workloads; Amazon Web Services (AWS) announced the general availability of the vector engine for Amazon OpenSearch Serverless with new features. The new update enables to store, update, and search billions of vector embeddings with thousands of dimensions in milliseconds
5. Enhancements in Enterprise AI Platforms
Integration of LLMs: Enterprises are increasingly embedding LLMs into their platforms, using extensive historical data to enhance both predictive capabilities and real-time decision-making. In customer service, integrated LLMs are used to power chatbots and virtual assistants that provide timely and contextually relevant support. Dataminr developed 50+ Dataminr-unique LLMs and FMs, to enhance its predictive AI and generative AI capabilities Uniphore introduced the X-platform, a multimodal solution that combines various AI technologies to improve customer experiences and sales interactions. This platform optimizes operations across voice, text, and video channels
6. Development of Actionable and GUI Agents:
There's a growing focus on developing sophisticated AI agents that enhance productivity through better task automation and decision support. These agents are being deployed in customer relationship management and enterprise resource planning systems to streamline operations and provide enhanced analytical insights. Orby AI raised $30 Mn in funding in a Series A round to accelerate the development and commercialization of its Gen AI process automation platform powered by a large action model; H Company has raised $220 Mn in funding in a seed round from investors including UiPath, AWS, Samsung to develop the foundational model and advanced AI agents
These trends suggest a future where AI not only supports but leads innovation and operational strategies.
Happy to hear your thoughts.