Sulagna Saha recently explored the performance of OpenAI’s new SearchGPT prototype, identifying both promising features and existing issues. The preliminary testing indicates that while the search tool showcases impressive capabilities in generating knowledgeable and contextually accurate responses, it also faces challenges with response relevance and accuracy under certain conditions. As this technology continues to develop, these insights could be instrumental in refining the capabilities and applications of AI-driven search tools in complex information environments. Read more in this article from Techstrong’s AI site, inspired by the recent Google Cloud presentation at AI Field Day 5.
Techstrong Gang – September 17, 2024
Stephen Foskett was a guest on the Techstrong Gang on September 17, where he discussed his observations from AI Field Day 5. Stephen provided an overview of AI Field Day, emphasizing its focus on practical applications of AI, such as productizing large language models (LLMs) and the development of AI agents that can automate processes, moving past the limited scope of chatbots. He also discussed advances in networking from Cisco and Arista, the radical new Ethernet fabric solution from Enfabrica, and the agentic AI platform created by Integrail.
Cisco Aims to Operationalize GenAI in Enterprises With AI-Ready FlexPod with NVIDIA NIM CVD
Cisco, in collaboration with NVIDIA, is paving the way for the integration of generative AI in enterprise environments with the introduction of the AI-Ready FlexPod. This technology leverages NVIDIA’s NIM CVD to enhance computational capabilities, aiming to streamline the deployment and management of AI applications. This initiative represents a significant step in operationalizing AI technologies, offering scalability and efficiency improvements for enterprise IT systems. Read more in this Techstrong AI article by Sulagna Saha following Cisco’s presentation at AI Field Day 5.
Hyperscale AI networks with Arista at AI Field Day
Arista Networks highlighted the need for enhanced network infrastructures at AI Field Day, focusing on the challenges of supporting complex AI models with higher bandwidth and lower power consumption. They introduced innovative solutions like Distributed Etherlink Switches to manage congestion and low-power optics to boost bandwidth, aimed at large-scale AI networks prevalent in heavy GPU environments. Additionally, Arista showcased their Cloud Vision management system, which leverages AI to simplify network operations, proving their dedication to evolving network technologies in accordance with the growing demands of AI workloads. Read more in this LinkedIn Pulse article by Alastair Cooke, who attended the Arista presentation at AI Field Day 5.
Arista Highlights the Power of AI
At AI Field Day 5, Arista Networks underscored the significant hurdles in deploying GPU clusters for AI training, highlighting issues like the immense power consumption driven by high data transfer demands ranging from 800Gb/s to 1.6 terabits for GPU communications. Arista suggested a shift towards energy-efficient linear pluggable optics to mitigate power consumption issues and discussed management of heat generation, noting the partial benefits of liquid cooling solutions but emphasizing that these do not fully address the broader environmental impacts. The company pointed out the urgent need for efficiency improvements in the computing industry to navigate upcoming operational and environmental challenges. Read more in this LinkedIn Pulse article by Justin Warren, who attended the Arista presentation at AI Field Day 5.
Enfabrica SuperNIC for massive clusters at AI Field Day
At AI Field Day, Enfabrica introduced its Accelerated Compute Fabric SuperNIC (ACF-S), aimed at enhancing connectivity in large compute clusters. The ACF-S features innovative integration with hardware accelerators like GPUs, supporting high bandwidths with a multi-layer crossbar switch capable of handling 5Tbps for PCIe and 3.2Tbps for Ethernet. This technology is set to boost load balancing, fault tolerance, and scalability, crucial for training complex language models, and is expected to transition into a more adaptable PCIe card format. Read more in this LinkedIn Pulse article by Alastair Cooke, who attended the Enfabrica presentation at AI Field Day 5.
Integrail Just Launched: No-code AI Agents – Your Sales and Marketing Assistant
Integrail has launched an innovative no-code platform for creating AI agents, designed to boost sales and marketing efficiency, as announced at AI Field Day 5. This user-friendly platform supports integration with leading large language models like those from OpenAI, Meta, and Google, and hooks into key business tools such as MS365 and Salesforce. Additionally, it leverages Retrieval-Augmented Generation for enhanced accuracy and offers the capability to assess and select from up to 25 different models to best meet individual business requirements, though it has also raised concerns regarding copyright and data storage issues in light of the EU AI Act. Read more in this LinkedIn Pulse article by Liselotte Foverskov, who attended the Integrail presentation at AI Field Day 5.
Enfabrica MegaNIC, a solution to GPU backend networking
Enfabrica has introduced the MegaNIC, a cutting-edge network interface card tailored for backend GPU applications, designed to address network congestion effectively. The device utilizes their new ACF-S ASIC, supporting advanced features like PCIe lane switching, Ethernet connectivity, and top-of-rack routing, all enhanced by software-defined networking to improve GPU interconnectivity. With its “Millennium” chip offering an 8 Tbps bandwidth and capable of connecting up to four GPUs via high-speed Ethernet links, the MegaNIC aims to boost performance and reduce power use in large-scale GPU superclusters. Read more in this article by Ray Lucchesi following Enfabrica’s presentation at AI Field Day 5.
Cisco Ethernet and validated designs at AI Field Day
At AI Field Day, Cisco demonstrated its latest advancements in Ethernet technology and validated designs, underscoring the importance of sturdy network infrastructure for AI applications. Their presentation focused on the improved performance, scalability, and security of their Ethernet solutions, designed to help organizations efficiently utilize AI. This initiative exhibited Cisco’s dedication to innovation in networking and artificial intelligence, ensuring enterprises have the reliable network performance needed to optimize their AI investments. Read more in this LinkedIn Pulse article by Alastair Cooke.
AIFD5 Recap
In this LinkedIn video, Mitch Ashley discussed how major tech vendors like VMware and Elastic are incorporating AI, ML, and GenAI into their business strategies. He highlighted the unique approaches these companies take in integrating advanced technologies into their operational frameworks and products. Ashley’s analysis underscores the rapid pace of technological innovation fueled by advancements in artificial intelligence.
VMware Private AI at AI Field Day
At AI Field Day, VMware focused on private AI solutions, delivering a secure, controlled deployment environments for AI models. VMware’s infrastructure and cloud technologies are leveraged to ensure compliance and security while enabling businesses to capitalize on AI advancements. The presentation underlined the tailored benefits and potential industry-specific solutions offered by VMware’s private AI initiatives. Read more in this LinkedIn Pulse article by Alastair Cooke.
Elastic brings Vectors to Semantic Search at AI Field Day 5
At AI Field Day 5, Elastic introduced an enhancement to its semantic search capabilities by integrating vector search. This new feature not only supports the creation and storage of vector data but also facilitates the use of multiple embedding models, improving memory efficiency through techniques like quantization. These advancements significantly enhance the precision and relevance of search results, positioning Elastic as a crucial player in the development and testing of Large Language Model applications and related technologies. Read more in this LinkedIn Pulse article by Alastair Cooke.
Keysight
Justin Warren comments on the Keysight Technologies presentation at AI Field Day 5. Coming from a legacy of innovation tracing back to HP/Agilent, and enhanced by acquiring Ixia, Keysight has unveiled a new AI data center test platform dedicated to benchmarking AI network performance and optimizing GPU clusters. This role is critical as it addresses reducing latency and optimizing expensive AI systems, especially during model training, with the capability to handle network troubleshooting at speeds up to 800 Gb/s. Read more in this LinkedIn Pulse article.
Cisco Says AI Networking Is Just Another Workload
During Cisco’s presentation at AI Field Day 5 the company positioned AI networking as a regular workload appearing in enterprise IT environments. The company highlighted the importance of a flexible network structure capable of handling varied workloads, leveraging their robust cloud networking foundation seen in products like UCS and Nexus to manage the demands of AI technologies. Despite the challenges associated with GPU clusters, Cisco remains optimistic about using its existing software-defined networking solutions to support AI applications, with the pragmatic adoption by early users shaping the future integration of AI in networking strategies. Read more in this LinkedIn Pulse article by Justin Warren.
AIFD5 Preview
Justin Warren shares his thoughts on this week’s AI Field Day event, highlighting the presentations by Elastic and VMware. Watch for more of Justin’s posts on PivotNine and LinkedIn.
Keysight at AI Field Day
At AI Field Day, Keysight highlighted its AI infrastructure testing solutions which leverage the capabilities from its Ixia acquisition. The company’s innovative test platform helps in optimizing network performance by emulating AI workloads, addressing common issues like data starvation and network congestion which contribute to 20% of AI job failures. This platform, which has evolved to include AresOnce hardware for traffic generation, also offers in-depth customizable settings and analytics through Jupyter Notebooks, enabling precise network tuning and enhanced performance. Read more in this LinkedIn Pulse article by AI Field Day delegate Alastair Cooke of The Futurum Group.
Integrail at AI Field Day 5
At AI Field Day 5, Integrail presented its innovative no-code platform, Integrail Studio, which simplifies the creation of specialized AI agents capable of collaborating to tackle complex challenges. The platform incorporates a user-friendly drag-and-drop interface, allowing the integration of large language models with business data, enhanced by Retrieval Augmented Generation for increased precision. Furthermore, Integrail facilitates advanced agent capabilities, including customizable workflows, output chaining, and the secure management of proprietary data, supported by benchmarking tools to maintain efficient large language model performance. Read more in this LinkedIn Pulse article by AI Field Day delegate Alastair Cooke of The Futurum Group.
AI Solves All Our Problems
Although AI can be quite useful, it seems that the promise of generative AI has lead to irrational exuberance on the topic. This episode of the Tech Field Day podcast, recorded ahead of AI Field Day, features Justin Warren, Alastair Cooke, Frederic van Haren, and Stephen Foskett considering the promises made about AI. Generative AI was so impressive that it escaped from the lab, being pushed into production before it was ready for use. We are still living with the repercussions of this decision on a daily basis, with AI assistants appearing everywhere. Many customers are already frustrated by these systems, leading to a rapid push-back against the universal use of LLM chatbots. One problem the widespread mis-use of AI has solved already is the search for a driver of computer hardware and software sales, though this already seems to be wearing off. But once we take stock of the huge variety of tools being created, it is likely that we will have many useful new technologies to apply.
The Shifting Focus of AI at AI Field Day 5
AI Field Day 5 is coming on September 11-13, reflecting the shift in enterprise AI from large-scale models to practical applications, including vector databases and agentic AI. The event will gather industry leaders like Cisco, VMware, Keysight, and Arista to explore real-world applications, AI-ready infrastructures, and private AI solutions. Attendees will also hear from companies like Integrail, Elastic, and Enfabrica to learn about next-generation AI capabilities. Tune in live on LinkedIn or Techstrong TV!
AI is Not a Fad
The current hype about building massive generative AI models with massive hardware investment is just one aspect of AI. This episode of the Tech Field Day podcast features Frederic Van Haren, Karen Lopez, Marian Newsome, and host Stephen Foskett taking a different perspective on the larger world of AI. Our last episode suggested that AI as it is currently being hyped is a fad, but the bigger world of AI is absolutely real. Large language models are maturing rapidly and even generative AI is getting better by the month, but we are rapidly seeing the reality of the use cases for this technology. All neural networks use patterns in historical data to infer results, so any AI engine could hallucinate. But traditional AI is much less susceptible to errors than the much-hyped generative AI models that are capturing the headlines today. AI is a tool that augments our knowledge and decision making, but it doesn’t replace human intelligence. There is a whole world of AI applications that are productive, responsible, and practical, and these are most certainly not a fad.
