Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Nvidia released NVLM 1.0, an open-source large language model family that includes a flagship 72B parameter version NVLM-D-72B.

The effort, detailed in a research paper, means Nvidia is also championing frontier open source LLMs. Previously, Meta and its Llama family of LLMs were leading the open-source model wave.

• Meta AI upgraded with Llama 3.2 models as Meta melds AI, AR, spatial computing
• Meta launches Llama 3.1 450B and for Zuckerberg it's personal
• Here's why Meta is spending $35 billion to $40 billion on AI infrastructure, roadmap

According to Nvidia researchers, NVLM 1.0 improves text-only performance after multimodal training. The Nvidia models have an architecture that enhances training efficiency as well as multimodal reasoning.

Nvidia also released the model weights for NVLM 1.0 and will open source the code. The move is notable since proprietary models don't release weights.

It remains to be seen whether LLM giants will follow suit. Regardless, NVLM 1.0 will enable smaller enterprises and researchers to piggyback off Nvidia's research. One thing is certain: LLM innovation is picking up pace. 

• Liquid AI launches non-transformer genAI models: Can it ease the power crunch?
• Intuit embraces LLM choice for multiple use cases
• 13 artificial intelligence takeaways from Constellation Research’s AI Forum
• Google Cloud rolls out new Gemini models, AI agents, customer engagement suite
• OpenAI releases o1-mini, a model optimized STEM reasoning, costs
• Anthropic launches Claude Enterprise with 500K context window, GitHub integration, enterprise security

More on Nvidia:

• Nvidia's uncanny knack for staying ahead
• Nvidia highlights algorithmic research as it moves to FP4
• Nvidia launches NIM Agent Blueprints aims for more turnkey genAI use cases
• Nvidia shows H200 systems generally available, highlights Blackwell MLPerf results
• Nvidia outlines roadmap including Rubin GPU platform, new Arm-based CPU Vera

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

MongoDB said its MongoDB 8.0 is generally available with throughput optimizations and efficiency enhancements.

At its MongoDB.local London event, MongoDB 8.0 went to GA along with other enhancements to the company's Atlas platform. MongoDB 8.0 is available on AWS, Google Cloud and Microsoft Azure through MongoDB Atlas, MongoDB Enterprise Advanced for on-premises users and MongoDB Community Addition.

• MongoDB rebounds in Q2 with better-than-expected results
• MongoDB adds to Atlas platform, scales partnerships, flexibility
• GenAI boom eludes enterprise software...for now
• MongoDB Atlas expands Google Cloud Vertex AI integration, eyes vertical use cases

MongoDB has a popular document database that is being used in generative AI applications looking to tap into unstructured data. MongoDB said its latest database has more than 45 architecture enhancements. Here is a rundown of what MongoDB announced:

• MongoDB 8.0 improves throughput by 32% to query and transform data and reduced memory usage. The database has sped up bulk writes by 56% and concurrent writes during data duplication by 20%. With high volumes of time series data and complex aggregations, MongoDB 8.0 can run 200% faster with lower costs.
• Sharding improvements in MongoDB 8.0 mean it can distribute data up to 50 times faster at 50% lower costs.
• Better controls to optimize performance for high demand and spikes in usage. MongoDB enables customers to set a default maximum time limit for running queries, reject recurring problem queries and run through events like restarts in peak demand.
• MongoDB Queryable Encryption to allow customers to encrypt sensitive application data, store it as randomized encrypted data and run queries on encrypted data for processing without cryptography expertise. Data will remain encrypted until it reaches an authorized user with a decryption key.
• The company enhanced MongoDB Atlas' control plane to scale cluster faster and optimize performance. Atlas customers will see up to 50% quicker scaling times. Auto-scaling will also see 5x improvements.
• A private preview for MongoDB for IntelliJ Plugin, which is a popular developer environment for Java.
• The company announced a public preview for MongoDB Participant for GitHub Copilot, which integrates AI tools in a chat experience in the MongoDB Extension for VS Code.
• MongoDB added support in MongoDB and Ops Manager for multiple Kubernetes clusters. Customers can deploy ReplicaSets, Sharded Clusters and Ops Manager across local or distributed Kubernetes clusters.
• MongoDB Atlas Search and Vector Search are generally available via Atlas CLI and Docker. MongoDB also announced vector quantization for Atlas Vector Search. The move reduces memory by up to 96%.
• Integration between MongoDB and large language model frameworks such as LangChain, LlamaIndex, Microsoft Semantic Kernel, AutoGen, Haystack, Spring AI and ChatGPT Retrieval Plugin.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Cerebras Systems has filed for a US initial public offering, but this alleged Nvidia competitor has multiple risk factors and depends on one UAE-based customer, Group 42 Holding, for 87% of its revenue.

Despite the risk, Wall Street will closely watch Cerebras Systems given a weak IPO market and the need for more competition in the genAI infrastructure market. Cerebras also argued in its IPO filing that it has a more power efficient approach to AI training and inference workloads.

Cerebras Systems launches Cerebras Inference, touts performance gains over Nvidia H100 systems

Here's what you need to know.

• Cerebras' processors are 57 times larger than the leading commercially available GPU. The company argues that its Cerebras Wafer-Scale Engine (CS-3) is the largest ever sold. It has 52 times more compute cores and 88 times more on-chip memory and 7,000 more memory bandwidth than leading GPUs. By keeping operations on the wafer, Cerebras' processors can solve problems faster with less power. Cerebras said its processors are designed for both AI training and inference.

• Wafer size matters. Cerebras is arguing that its approach with massive processors means that it can scale more easily for model training. The approach also works for memory bandwidth requirements for inference. The crux of the Cerebras argument is that individual GPUs are too small, scaling GPUs is inefficient and limited by memory bandwidth.
• The company is a play on on-premises generative AI. Organizations can purchase Cerebras AI supercomputers for on-premise deployments. Cerebras said its AI Supercomputer can scale up to 2,048 CS-3 systems or 256 exaFLOPs. The plan for Cerebras is to court sovereign AI initiatives, governments, cloud service providers and research institutions.

• The largest Cerebras cluster deployed as of Sept. 27 comprised of more than 100 CS systems so there's a lot of headroom to scale further.
• Cerebras relies mostly on Group 42 Holding Ltd., a UAE company, for its 87% of its revenue for the six months ending June 30. Cerebras said that Group 42 (G42) can acquire $335 million worth of Cerebras shares to give it more than a 5% stake.
• But Cerebras is raising capital to expand its customer base. The company, however, has a long way to go. Two customers accounted for 68% and 16% of Cerebras' accounts receivable balance as of June 30.
• Cerebras is losing money. For the six months ended June 30, the company posted a net loss of $66.6 million on revenue of $136.4 million. For the year ended Dec. 31, Cerebras had a net loss of $127.15 million on revenue of $78.74 million.
• The company could be hampered by US export regulations. Specifically, Cerebras must comply with U.S. Export Administration Regulations (EAR), which are administered by the U.S. Department of Commerce’s Bureau of Industry and Security (BIS), as well as economic and trade sanctions, including those administered by the U.S. Department of the Treasury’s Office of Foreign Assets Control (OFAC). Given nearly all of Cerebras' revenue is tied to a customer in UAE those import and export control laws loom large. Cerebras did note that it has obtained a BIS export license to sell its CS-2 systems in UAE.
• TSMC is the lone manufacturer of Cerebras systems. In its IPO filing, Cerebras said: "We worked with TSMC to develop the processes necessary to manufacture the semiconductor wafers needed for our wafer scale engine, which involve many complexities and proprietary technologies. We are currently dependent on TSMC to produce all of the wafers that we use in our products. We have no formalized long-term supply or allocation commitments from TSMC, and TSMC also fabricates wafers for other companies, including certain of our competitors, many of whom are significantly larger than us and purchase considerably more wafers from TSMC than we do."
• Competition is fierce. Not surprisingly, Cerebras' competitive set is well funded. Nvidia, AMD, Intel, Microsoft and Alphabet were cited as competitors for AI workloads.

Related:

• Nvidia outlines roadmap including Rubin GPU platform, new Arm-based CPU Vera
• AMD expands its data center, AI infrastructure push with $4.9 billion purchase of ZT Systems
• DigitalOcean highlights the rise of boutique AI cloud providers
• Equinix, Digital Realty: AI workloads to pick up cloud baton amid data center boom
• The generative AI buildout, overcapacity and what history tells us

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Liquid AI, an MIT spinoff, launched its Liquid Foundation Models (LFM) in three sizes without using the current transformer architecture used by large language models (LLMs) with good performance.

According to Liquid, LFMs have "state-of-the-art performance at every scale" and have a smaller memory footprint and more efficient inference than LLMs. That efficiency could mean that LFMs will use less power and be more sustainable. After all, electricity is one of the biggest limiting factor for AI workloads and one big reason for the renaissance of nuclear power.

• Constellation Energy, Microsoft ink nuclear power pact for AI data center
• Generative AI driving interest in nuclear power for data centers
• DigitalOcean highlights the rise of boutique AI cloud providers
• Equinix, Digital Realty: AI workloads to pick up cloud baton amid data center boom
• The generative AI buildout, overcapacity and what history tells us
• AI infrastructure is the new innovation hotbed with smartphone-like release cadence

Simply put, we need more efficient models than taxing the grid, using too much water and throwing more GPUs at the issue. "Liquid AI shows that leading models don't have to come from deep pocketed large players and can also come from startups. The intellectual race for AI is far from being over," said Constellation Research analyst Holger Mueller. 

LFMs are general-purpose AI models for any sequential data, including video, audio, text, time series and signals. Liquid will hold a launch event October 23 at MIT Kresge in Cambridge to talk about LFMs and applications in consumer electronics and other industries.

LFMs come in three sizes--1.3B, 3.1B and 40.3B mixture of experts (MoE)--and are available on Liquid Playground, Lambda, Perplexity Labs and Cerebras Inference. Liquid AI said its stack is being optimized for Nvidia, AMD, Qualcomm, Cerebras and Apple hardware.

The smallest model from Liquid AI is built for resource-constrained environments with the 3.1B model focused on edge deployments.

Although it is early Liquid AI plans to "build private, edge, and on-premise AI solutions for enterprises of any size." The company added that it will target industries including financial services, biotech and consumer electronics.

Liquid AI has a bevy of benchmarks comparing its LFMs to LLMs and the company noted that the models are a work in progress. LFMs are good at general and expert knowledge, math and logical reasoning, long-context tasks and English. LFMs aren't good at zero-shot code tasks, precise numerical calculations, time-sensitive information and human preference optimization.

13 artificial intelligence takeaways from Constellation Research’s AI Forum

A few key points about LFMs and potential efficiency gains.

• Transformer models' memory usage surges for long inputs so they do not do edge deployments well. LFMs can handle long inputs without affecting generation speed or the amount of memory required.
• Training LFMs require less compute compared to GPT foundation models.
• The lower memory footprint means lower costs at inference time.
• LFMs can be optimized for platforms.
• LFMs could become more of an option as enterprises start to hot swap models based on use case. If LFMs do become an option more efficiency and lower costs would favor their increased adoption.

Bottom line: Liquid AI's LFMs may steer the conversation more toward efficiency over brute strength when it comes to generative AI. Should genAI become more efficient it could upend the current economic pecking order where all the spoils go to infrastructure players--Nvidia, Micron Technology, Broadcom and others.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Intuit is operating on one data and AI platform that enables it to select up to 10 large language models for various consumer and business use cases via its Generative AI Operating System (GenOS).

The ability to select multiple large language models (LLMs) gives Intuit the ability to leverage genAI for use cases with a few clicks and build in redundancy. Intuit is looking to leverage a unified data and AI platform to solve customer problems and bring in human experts when needed.

Speaking at Intuit's Investor Day, CTO Alex Balazs, said GenOS is transformative to the company's platform plans. Intuit made a big bet five years ago on one data platform and AI as a way to expand its total addressable market. Now Intuit TurboTax, Credit Karma, QuickBooks and Mailchimp run on a unified platform as well as GenOS, which powers Intuit Assist.

"Developers leveraging our platform and GenOS now have access to more than 10 LLMs," said Balazs. "They're able to easily select the right large language model that solves the specific customer use case. GenOS also allows us to seamlessly switch between LLMs to provide resiliency so the customer has a smooth experience."

• Enterprises leading with AI plan next genAI, agentic AI phases
• 13 artificial intelligence takeaways from Constellation Research’s AI Forum
• Intuit sees 'green shoots' from its generative AI strategy
• Intuit’s bets on data, AI, AWS pay off ahead of generative AI transformation (PDF)

Intuit also recently announced AI Workbench, a dedicated development environment for AI native experiences, said Balazs. Earlier this month, Intuit outlined enhancements to GenOS including AI Workbench as well as updates to GenStudio, GenRuntime and GenUX.

According to Intuit, GenOS AI Workbench includes an LLM Leaderboard for use cases, prompt management, an automated evaluation service for LLMs and traceability for prompt workflows. Other updates include:

• An LLM sandbox in GenStudio that includes Anthropic Claude via Amazon Bedrock, Gemini from Google Cloud, Llama from Meta AI and Mistral AI to complement custom LLMs and OpenAI GPT models via Microsoft Azure.
• GenRuntime, a layer that includes GenOrchestrator to plan, execute and retrieve knowledge and tools for agentic workflows.
• GenSRF (security, risk and fraud) that has guardrails for genAI deployments.
• GenUX, which includes more than 140 new UX components, widgets and patterns for developments.

The company's strategy highlights how enterprises leading in genAI are building platforms that are able to hot swap models as they advance. Intuit has built its infrastructure on Amazon Web Services and said last year at re:Invent that GenOS uses Amazon Bedrock as well as multiple services including Sagemaker.

Intuit's selection of models in GenOS is a subset of what's available on Amazon Bedrock. For instance, Meta has more than 10 Llama models available on Amazon Bedrock with providers such as A121, Anthropic, Cohere and Mistral offering more than a handful of foundational models.

This model choice is also increasingly being offered by software as a service providers, which are packaging a selection of models that can be used to build AI agents. Demonstrations of Salesforce's Agentforce platform highlighted the ability to select models to build agents.

Model selection is a key cog in what Balazs calls Intuit's durable advantage--its data and AI platform and ability to enable machine learning, natural language processing and LLM development to embed fintech throughout its environment.

Balazs said Intuit's model agnostic approach allows the company to future proof its platform as it chases its five big bets: Revolutionize speed to benefit, connect people to experts, unlock smart money decisions, be the center of small business growth and disrupt the mid-market.

"We're going to continue to innovate and look ahead and determine the best way to serve our customers, especially as the AI landscape continues to rapidly evolve. Almost every day, there's some type of announcement of some new AI capability."

Indeed, Intuit demonstrated the use of digital avatars as a way to provide guidance and insights to customers. Balazs said that avatars will help people retain information and learn. The goal would be to couple LLMs, genAI and avatars to deliver human-like experiences that seamlessly hand off to human experts.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Agentic AI is going to hit sprawl quickly, boardrooms are being reconstituted over fears of being left behind, genAI is still mostly an experiment with fuzzy returns and old-school issues like change management still determine whether companies successfully move from pilot to production.

Those are a few of the takeaways from Constellation Research's AI Forum. Here's a look at everything we learned at the AI Forum in New York.

The board of directors is driving the AI conversation. AI is clearly a boardroom issue, said Betsy Atkins, CEO of Bajacorp. "What boards have figured out is that if they don't lean in and adopt AI and technology they're going to be left behind," said Atkins.

Boards are also being reconstituted for AI. "I see boards shifting in terms of cohorts," said Atkins, who said enterprises are creating boardrooms that can look at technology as well as new business models to differentiate.

However, the board also wants ROI. Atkins said that enterprises are looking at use cases with quick ROI because boards now realize how expensive AI can be.

Change management is more important than technical capability in production generative AI deployments, Michael Park, SVP, Global Head of AI GTM at ServiceNow. Park added: "I think getting the data structure ready and the instance ready is the easy part. That's just the tech, and there's hard work that needs to be done around it. The challenge that we're seeing right now is the organizational change management and getting people to see what's possible. Change management has been the biggest struggle. The tech is real."

Agentic AI. "There's no doubt that agentic AI is the future," said Park. He said there will be two domains of AI agents. One will augment a human being to supercharge capabilities. And another domain will be an aggregated set of agents that work on behalf of a unit. "I think every job is going to be affected in some ways and transform productivity for employees and customer experiences," said Park.

Attendees at the AI Forum generally agreed that the agentic AI wave is real, but doubted the technology has quite caught up with production use cases yet. That skepticism sure hasn’t stopped vendors from talking about agentic AI though.

In recent weeks, Salesforce, Workday, Microsoft, HubSpot, ServiceNow, Google Cloud and Oracle all talked about AI agents and likely overloaded CxOs who have spent the last 18 months trying to move genAI from pilot to production. Other genAI front runners—Rocket, Intuit, JPMorgan Chase--have mostly taken the DIY approach and are now evolving strategies.

Agent orchestration will be needed quickly because overload will be here soon. Boomi CEO Steve Lucas said that the number of AI agents will outnumber the number of people in your business in less than three years. "The digital imperative is how do I work with agents? The number of agents will outnumber the number of humans in less than three years," said Lucas. Fun fact: Constellation Research analyst Holger Mueller thinks Lucas prediction is way conservative.

This post first appeared in the Constellation Insight newsletter, which features bespoke content weekly and is brought to you by Hitachi Vantara.

Healthcare is expected to be the most transformed industry by AI. Multiple attendees and panelists at AI Forum noted that healthcare will see the most transformational impact from AI. Anand Iyer Chief AI Officer at Welldoc, said data and AI can transform outcomes and be more preventative. Iyer said: "You can actually figure out what cocktail of exercise, food, stress, reduction, and all of these vectors that drive somebody's own health. You can figure out the exact cocktail that works for Person X, in a way that fits into their life flow and their clinician's workflow."

There may be a catch with AI health transformation. AI will bring costs down initially but may end up being more expensive due to the level of personalization.

Trust in AI will take time. Scott Gnau, Vice President of Data Platforms Intersystems, said every technology wave requires time to earn trust. Gnau noted building trust in a technology can take years, but AI has a chance to earning user trust quickly. "One or two bad answers can set generative AI back, but the addition of provenance and governance helps," said Gnau. "One of the game changers is that AI can actually be used to explain the provenance of the answer. I think we have a unique opportunity to accelerate trust."

Generative AI is still in the science experiment phase. Gnau added that there's a degree of FOMO with deploying AI. "Is generative AI or a large language model the right tool for every problem out there? Absolutely not. There are things that you've built that run your business today that are good so don't suck all of the budget away and let them crumble," said Gnau. "Make those systems better with AI and use the right tool for the right processes."

The AI playbook isn't fully baked. In a pop-up survey at the AI Forum, 35 CxOs indicated that they are trying a little bit of everything when it comes to AI (good thing they could give multiple answers). Respondents indicated that they were using multiple approaches to build AI capabilities. The majority (79%) said they were developing home-grown AI services on hyperscale cloud services and 48% were also using open-source frameworks and large language models. Many of these efforts included AI embedded in packaged applications that they already used such as Salesforce, Adobe, Oracle, SAP etc.

14 takeaways from genAI initiatives midway through 2024

Data quality remains the biggest hurdle in generative AI deployments. "Data quality is the biggest roadblock to realizing generative AI's full value. You need a data driven strategy combined with a model driven strategy and then you can iterate quickly," said Michelle Bonat, Chief AI Officer of AI Squared. But without a focus on data quality, your models won't be good enough to use.

The role of the Chief AI Officer. Chief AI Officers will need to know a lot of business functions and technology much like CIOs and CTOs, but to lead AI strategy you'll need to know the technologies. "I think it's necessary to have someone with a good knowledge of AI," said Phong Nguyen, Chief AI Officer of FPT Software, which is based in Vietnam. "You need to have the deep technical skills and understand what AI can bring."

Minerva Tantoco, CEO of City Strategies LLC, agreed. "When something is relatively new with a lot of potential it really does require a strong alignment with the goals of the organization," she said. "Once you set a strategy it becomes the fabric of the enterprise. But in the beginning, you want the chief AI officer to have a really strong background in AI. This is a transformational role."

AI leaders need to be trilingual. Tantoco said AI leaders need to be trilingual in technology, business and governance and compliance. "At this stage, you need to collaborate across multiple disciplines while leaning into the strong technical background," she said.

David Trice, CEO of inZspire AI, said AI leaders have multiple roles to juggle. First, enterprises need to drive AI or they'll fall behind. Trice echoed Tantaco's sentiment that AI leaders need to bring multiple threads together. "Product, data and AI innovation need to be at the table with legal, compliance and security," said Trice.

Human led AI or vice versa? Chris Nicholas, President and CEO Sam's Club, said artificial intelligence is enabling the company to "take 100 million tasks out of our clubs" even though it has more associates. Yet he has a clear view on who leads the AI charge: Humans. AI is about freeing humans from the mundane to solve customer problems.

AI and human rights. But just in case AI does kill jobs it's worth pondering a human rights update for age we're entering. Will there be a reskilling safety net and the ability for humans to pursue their passions for a living? A workshop on AI and human rights surfaced a lot of thoughts about the right to work as well as the right to opt out of AI and what's likely to become augmented humanity. One prevailing thought was that we are working towards using AI to augment human intelligence. In the future, that pecking order will be reversed and human intelligence will augment AI.

More from AI Forum:

• AI projects remain work in progress, pilots, say CxOs
• Sam's Club CEO Nicholas on AI, frictionless commerce, focus on members
• Boomi CEO Lucas: AI agents will outnumber your human employees soon

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

AMD is starting to land hyperscale deals for its AMD Instinct MI300X AI accelerators with ROCm open software and the wins portend more competition for Nvidia.

The chipmaker said that Oracle Cloud Infrastructure (OCI) chose AMD Instinct MI300X and ROCm for its latest OCI Computer Supercluster instance. The OCI Supercluster with AMD MI300X supports up to 16,384 GPUs in a single cluster.  OCI outlined AMD MI300X performance in a June blog. 

Oracle's cloud, AI plans are a master class in co-opetition

AMD's OCI win came a day after Vultr, a privately held cloud computing platform, said it will use MI300X and ROCm. Vultr focuses on AI workloads. At a recent Goldman Sachs investor conference, AMD CEO Lisa Su said the company is rolling out MI300X at scale. Su said:

"We launched MI300X in December. It has had just tremendous customer traction and customers have been really excited about it. We have several large hyperscalers, including Microsoft, Meta, Oracle, which have adopted MI300 as well as all of our OEM and ODM partners."

Su said AMD's biggest efforts have been on the software side with ROCm and working with large language models (LLMs). AMD has also built out its AI business with the acquisition of Silo AI and ZT Systems. AMD will follow up the MI300X with the MI325 in the fourth quarter and then the MI350 series and MI400. AMD will hold an event Oct. 10 to highlight its upcoming AI roadmap.

For now, AMD's AI processors are just getting traction, but enterprises will be happy to have a counterweight to Nvidia and some additional competition.

• AMD: Strong Q2, raised Q3 guidance, data center revenue doubles
• AMD outlines new GPUs, Instinct accelerators with annual cadence
• AMD sees AI inference, training workloads increasing in enterprise
• Microsoft Build 2024: Azure gets AMD MI300X Accelerators, Cobalt preview, OpenAI GPT-4o

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

IBM has installed its second-gen IBM Quantum Heron processors in its Poughkeepsie data center as it builds out its quantum infrastructure.

Big Blue launched its latest Heron quantum processor and IBM Quantum System Two late last year. Now those systems have been deployed in IBM's Quantum Data Center, the company has more than a dozen quantum computers in its fleet, which is available via IBM Cloud.

Constellation Research analyst Holger Mueller said the move to install the latest Heron quantum processor gives IBM's quantum efforts a big boost. Mueller said:

"IBM is showing the commercial viability with Heron, with its second system available for quantum use cases. This is great news for the quantum community as it can run new use cases on the largest set of available qubits. All eyes are now on how IBM will be connecting these Heron systems together.

Heron is the Lego block that IBM wants to use to put a massive quantum system together built on racks. A lot can go wrong, as we know with traditional racks - but IBM has HPC experience. It's good to see Heron - the quantum Lego block - is working."

IBM said its Heron-based quantum system offers a 16-fold improvement in performance and a 25-fold increase in speed over previous IBM systems. The company said its two Heron-based systems in addition to its other quantum systems mean its IBM Quantum Data Center can run quantum circuits better than classical systems simulating them.

Quantum computing has been developing quietly as the tech industry has been focused on generative AI. Ultimately, generative AI and AI could converge for computing breakthroughs.

Recent quantum computing developments include:

• Microsoft claims hybrid quantum breakthrough with Quantinuum, partners with Atom Computing
• Quantum Brilliance, Oak Ridge National Laboratory aims to meld quantum computing, HPC
• IonQ's quantum computing bets: Quantum for LLM training, chemistry and enterprise use cases
• Classiq CEO Minerbi on the intersection of quantum computing, HPC and use cases
• IBM launches latest Heron quantum processor, IBM Quantum System Two
• Quantinuum raises $300 million, valued at $5 billion
• Will IonQ make quantum computing enterprise relevant in 2025?
• Why your quantum computing vendors are going to look familiar

Big Blue said its quantum data center can push new algorithms forward to reach quantum advantage. IBM also touted its Qiskit software to program quantum computers.

IBM said it will continue to expand its IBM Quantum Data Center as it executes on its roadmap. The Poughkeepsie location is the global hub for IBM's Quantum Network, but it is expanding with a second quantum facility in Ehningen, Germany.

• Quantum Computing Platforms
• Quantum Computing Software Platforms
• Quantum Full Stack Players

Here's a look at IBM's quantum roadmap (click to expand).