Equifax bets on Google Cloud Vertex AI to speed up model, scores, data ingestion

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Equifax said it is deploying Google Cloud's Vertex AI platform across its systems as it accelerates data ingestion as well as new product launches.

Speaking on Equifax's first quarter earnings call, CEO Mark Begor said driving AI innovation is key to the company's growth goals via products such as Ignite and Interconnect. Begor said:

"During 2024, we're deploying both Equifax proprietary explainable AI along with Google Vertex AI across Ignite, Interconnect and our global transaction systems.

For Equifax, Vertex AI enables faster and more predictive model development on our Ignite platform. And for our clients, Ignite, which combines data analytics and technology into one cloud-based ecosystem, customers can connect their data with our unique data through our identity resolution process to gain a single holistic view of consumers."

Constellation Research previously covered Equifax's cloud transformation in a customer story. Equifax also was a reference account at Google Cloud Next.

Equifax reported first quarter revenue of $1.39 billion, up 7% from a year ago, with net income of $124.9 million.

Those platforms combined with Equifax's broad reach into datasets means faster ingestion and analytics and new products and services.

Begor said Equifax has access to 100% of the US population through its data sets in a single data fabric. He added that Equifax's cloud infrastructure is processing data 5x faster than its legacy applications could.

Equifax is also planning on completing its cloud transformation, closing data centers and saving $300 million a year in 2024. From there, Equifax will be focused on growing its EFX.AI offerings with higher performing models, scores and data products.

"Completing the cloud transformation also frees up our team to fully focus on growth and expanding innovation, new products and new markets. Our progress towards completing the cloud is gaining momentum with over 70% of our total revenue in the new Equifax Cloud at the end of the quarter," said Begor. "And we're focused on executing the remaining steps to reach 90% with Equifax revenue in the cloud by year-end."

Meta launches Llama 3, wide availability with more versions on deck

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Meta launched its Llama 3 open source large language model and said it will be available on AWS, Databricks, Google Cloud, Microsoft Azure, Snowflake, IBM WatsonX, Nvidia NIM and have support from enterprise hardware platforms.

Yes, we're in the age of weekly LLMs that leapfrog each other, but Llama 3, which will initial come in 8B and 70B parameters with more versions on deck, is of interest to enterprises.

Why? Companies are likely to look to capable open source LLMs and then look to fine tune them with enterprise data. Llama 3 represents a backdoor enterprise play for Meta.

• 77% of CxOs see competitive advantage from AI, says survey
• Foundation model debate: Choices, small vs. large, commoditization
• Middle managers and genAI | Why you'll need a chief AI officer | Enterprise generative AI use cases, applications about to surge 

In a blog post, Meta outlined the Llama 3 effort, which is a big leap over Llama 2. "Improvements in our post-training procedures substantially reduced false refusal rates, improved alignment, and increased diversity in model responses. We also saw greatly improved capabilities like reasoning, code generation, and instruction following making Llama 3 more steerable," said Meta.

Meta added that it evaluated Llama 3 based on prompts covering 12 use cases including brainstorming, advice, coding, creative writing, extraction and summarization to name a few. In this use case testing scenario, Llama 3 70B topped Claude Sonnet, Mistral Medium and Llama 2.

According to Meta, Llama 3 is being deployed across its applications including Facebook, Instagram and WhatsApp. Llama 3 is also available for a spin on the web.

Going forward, Meta said a 400B parameter model is training and it'll "release multiple models with new capabilities including multimodality, the ability to converse in multiple languages, a much longer context window, and stronger overall capabilities."

Infosys acquires In-tech, posts mixed Q4 with genAI progress

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Infosys CEO Salil Parekh said the services provider is landing large deals and "seeing excellent traction with our clients for generative AI work," but its fourth quarter was mixed. Infosys also said it would acquire In-tech, an engineering and R&D services provider catering to the German automotive industry.

The company reported fourth-quarter earnings of $958 million on revenue of $4.56 billion, flat from a year ago. Wall Street was expecting Infosys to report fourth quarter earnings of 17 cents a share. For the year ended March 31, Infosys reported earnings of $3.17 billion on revenue of $18.56 billion.

For fiscal 2025, Infosys projected revenue growth of 1% to 3% with operating margins of 20% to 22%.

Constellation Research CEO Ray Wang said Infosys is facing a bevy of moving parts that the company will have to navigate. "The combination of AI arbitrage, margin compression, and exponential efficiency is having an effect on the overall market.  Constellation expects that the overall service market will be flat to single digit growth for the next year," said Wang.

Constellation ShortList™ Digital Transformation Services (DTX): Global

Indeed, Infosys outlined a bevy of items on its earnings call. During the fourth quarter, Infosys said it had "a rescoping and renegotiation of one of the large contracts in the financial services segment." That renegotiation led to a 1% revenue hit, but 85% of the contract continued as is.

Parekh on a conference call said Infosys said fiscal 2024 brought in $17.7 billion in large deals. These large deals revolve around cost efficiency and consolidation.

Here's a look at what Infosys is facing.

The good

Generative AI remains a highlight for Infosys. Parekh said:

"We're working on projects across software engineering, process optimization, customer support, advisory services and sales and marketing areas. We're working with all market-leading open access and closed large language models.

As an example, in software development, we've generated over 3 million lines of code using one of generative AI large language models. In several situations, we've trained the large language models with client specific data within our projects. We've embedded generative AI in our services and developed playbooks for each of our offerings."

Public and private cloud migrations remain a priority. "We continue to work closely with the major public cloud providers and on private cloud programs for clients. Cloud with data is the foundation for AI and generative AI and Cobalt encompasses all of our cloud capabilities," said Parekh.

Data and automation. Parekh said the acquisition of In-tech played into the data strategy for Infosys. "We see data structuring, access, assimilation critical to make large language models and foundation models to work effectively, and we see good traction in our offering to get enterprises, data ready for AI," he said.

The challenges

Jayesh Sanghrajka, CFO of Infosys, said the company saw 180 basis points of margin compression quarter over quarter due to the renegotiation of the large contract, salary increases, brand building and visa expenses.

Infosys did offset the margin hit somewhat by lower post sales customer support and efficiency efforts called Project Maximus.

On the economy, Sanghrajka said:

"We continue to see macroeconomic effects of high inflation as well as high interest rates. This is leading to cautious spend by clients who are focusing on investing in services like data, digital, AI and cloud."

Industry demand

Sanghrajka said industry demand was mixed with strength in financial services and manufacturing as well as retail.

Financial services: "Financial services firms are actively looking to move workloads to cloud, pipeline and deal wins are strong and we are working with our clients on cost optimization and growth initiatives."

Manufacturing: "There is increased traction in areas like engineering, IoT, supply chain, smart manufacturing and digital transformation. In addition, our differentiated approach to AI is helping us gain mind and market share. Topaz resonates well with the clients. We have a healthy pipeline of large and mega deals."

Retail: "In retail, clients are leveraging GenAI to frame use cases for delivering business value. Large engagements are continuing S/4HANA and along with infra, apps, process and enterprise modernization. Cost takeout remains primary focus."

Communications: Clients remain cautious, and budgets are tight. Sanghrajka said cost takeout, AI and database initiatives may show promise.

Overall, Sanghrajka said Infosys should benefit with large deals recently won as well as AI. "We are witnessing more deals around vendor consolidation and infra managed services. Deal pipeline of large and mega deals is strong due to our sustained efforts and proactive pitches of our cost takeouts and digital transformation, etc., across the subsectors," he said.

SAS launches industry-focused models, Model cards that serve as AI nutrition labels

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

SAS said it will launch a series of AI models that are lightweight and focused on industry use cases. The company also added generative AI features to its Viya platform and unveiled "nutrition labels" for models.

The news, outlined at the company's SAS Innovate conference in Las Vegas, is part of the company's broader investment in AI.

"Once seen as the laggard in the industry, the decades of experience in mastering data by SAS is now very valuable in the packaging of data models for easy consumption for customers for AI.  They are making it easier for customers to put AI to work," said Constellation Research CEO Ray Wang. 

Here's the breakdown of what was announced:

Industry-focused models. SAS said it will roll out a series of models for individual license starting with an AI assistant for warehouse space optimization designed to enable nontechnical users to optimize and plan faster.

The general idea of the industry AI models is to give enterprises something they can deploy quickly with low overhead costs. SAS will target financial, healthcare, manufacturing and public sector AI models.

77% of CxOs see competitive advantage from AI, says survey | Why digital, business transformation projects need new approaches to returns | Why you'll need a chief AI officer | Enterprise generative AI use cases, applications about to surge

SAS' bet is that it can move beyond large language models and drive value with industry-proven AI models for fraud detection, supply chain, document conversation and healthcare payments to name a few.

SAS Viya gets generative AI tools. SAS Viya will get trustworthy genAI tools and introduce a synthetic data generator called SAS Data Maker.

Viya has genAI orchestration tools to integrate external models, Viya Copilot for developer, data science and business productivity, Data Maker, which is in private preview, and genAI features in SAS Customer Intelligence 360.

• Constellation ShortList™ Self-Service Data Science & Machine Learning
• Constellation ShortList™ Multicloud Analytics and Business Intelligence Platforms

Model cards and AI governance. SAS also said that it will offer model cards, which are a nutrition label for AI designed to flag bias and model drift, as well as AI governance services.

SAS said Model cards will be an upcoming feature in Viya in mid-2024.

SAS Viya Workbench goes GA. SAS announced the general availability of SAS Viya Workbench aimed at model developers. Viya Workbench is a self-service compute environment for data prep, analysis and developing models. Viya Workbench will be available by the end of 2024 with SAS and Python initially and R to follow. Yiya has two development environment options Jupyter Notebook/JupyterLab and Visual Studio Code. 

AWS and SAS expand partnership. SAS said it has expanded its hosted managed services to AWS including SAS Viya. SAS' full product suite is available on AWS. 

Constellation Research’s take

Andy Thurai, analyst at Constellation Research, said:

“Viya Copilot is a useful offering that can help users with multiple tasks and is likely to be particularly useful in knowledge gap analysis and data wrangling tasks. Viya Copilot could useful in reducing the manual tasks that take up data scientist time.

SAS Data Maker, in private preview, can help users create synthetic data but only in tabular format. The problem is that AI needs a lot of unstructured data, which is harder to generate. SAS Data Maker may not be much of use to many organizations that need synthetic data for AI today. I hope SAS can eventually get there.

Packaged AI models and Industry-specific lightweight models are interesting. It is notable that more vendors seem to be moving towards smaller models. Google just announced last week about the specialized models’ concept that can run on edge and other lightweight locations.

There is an argument that specialized, lightweight models may underperform without augmentation. These models need to be heavily trained on industry-specific data for them to be useful. Model cards and AI governance advisory services can help enterprises improve AI governance. However, there are smaller startups such as Guardrails AI that offer much more broader offerings in this space.

SAS’s AI initiatives are noteworthy and should interest the company’s customers. But the AI market is moving fast and many other vendors have already announced more advanced and game-changing features.”

Constellation Research analyst Doug Henschen gave his take on Viya Copilot and SAS Data Maker as well as the big picture. 

"Both Viya Copilot and SAS Data Maker are in private preview at this point, so I’d say they are potentially significant. The part that stands out for me is Data Maker, as SAS is one of the few vendors that is talking about and addressing the need for synthetic data generation. Constellation believes data scarcity will limit the accuracy and effectiveness of AI-based systems. SAS is one of the few companies talking about this capability in the context of their generative AI capabilities."

"SAS’s Copilot is similar to what many vendors have announced, and what a few now have available in public preview or even general availability. The capability that is somewhat differentiated is SAS Data Maker. All the big cloud vendors and big companies like IBM support synthetic data generation, but they don’t tend to talk about it in the same context as natural language-based GenAI. It’s a technique in which you feed the AI samples of data that you might have at a relatively small scale and the system will then generate similar, non-privacy-sensitive data sets for use in training on a massive scale." 

SAS has taken a conservative approach, but it has moved more quickly to infuse GenAI into its Customer 360 app, because that’s where competitors including Adobe and Salesforce have been pushing GenAI aggressively. Large customers doing analytics and AI at scale do not switch horses quickly based on this or that hot new feature. In fact, plenty of companies and CXOs are being very cautious about GenAI. What you will see, and what we have already seen over the years, is innovation teams doing experiments with cutting-edge vendors or cutting-edge capabilities provided by cloud vendors, for example. So SAS has to keep up. I see the company’s Generative AI Orchestration announcement as a signal to customers that SAS will enable customers to tap into a proven stable of open models when they are ready to pursue GenAI at scale.   

The pace of innovation is constantly accelerating, particularly in GenAI, so I’d like to see these private-preview announcements move into public preview and general availability as quickly as possible. I’d also like to see more detail on the portfolio of models SAS plans to make available for orchestration and where those models fit with various use cases and industry applications."

77% of CxOs see competitive advantage from AI, says survey

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Seventy-seven percent of CxOs believe AI will give their companies competitive advantage, but 91% of companies will determine that they don't have enough data to achieve a level of precision needed for trust, according to a Constellation Research and Dialpad survey.

The survey is based on responses from more than 1,000 senior executives in the US, Canada, UK, Australia and New Zealand about their AI initiatives.

Key takeaways from the survey include:

• 77% of leaders believe AI will give them competitive advantage.
• 75% of respondents believe AI will have a significant impact on their roles in the next three years.
• 54% are concerned about AI regulation.
• 38% are moderately to extremely concerned about AI.
• 72% of CxOs plan to reskill workers for AI.
• 69% of respondents are already using AI at work.
• 33% of executives say their companies are using two AI solutions, 15% have three and 9.5% using four or more.

The survey also focused on early adopters who are applying AI to analytics, automating work, content creation, forecasting and insights. These CxOs are betting that the Return on Transformation Investments (RTI) with AI will come from efficiency, revenue and growth, compliance and risk and proactive monitoring.

• Why digital, business transformation projects need new approaches to returns
• Data leaders bullish on generative AI, but multiple challenges remain, says Informatica
• Hitachi Vantara Exchange New York: 10 takeaways on generative AI, transformation, data management
• GenAI trickledown economics: Where the enterprise stands today

While security, data leakage, hallucinations with generative AI and cost are key concerns for CxOs, high quality and high-volume data has emerged as a more long-term concern. Ninety-one percent of companies will determine they don't have enough data to achieve a level of precision they trust. For now, however, 66.5% of CxOs believe their team is getting enough data to power AI efforts.

Related: Middle managers and genAI | Why you'll need a chief AI officer | Enterprise generative AI use cases, applications about to surge | CEOs aim genAI at efficiency, automation, says Fortune/Deloitte survey

UnitedHealth sees $1.35 billion to $1.6 billion hit in 2024 due to Change Healthcare cyberattack

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

UnitedHealth Group has tallied up the costs from its Change Healthcare cyberattack including direct response, funding to care providers and lost revenue as the incident sucked out $3 billion of cash flow in the first quarter.

For 2024, UnitedHealth said the tab for the Change Healthcare cyberattack could be as high as $1.6 billion.

In February, UnitedHealth's Change Healthcare unit was hit with a ransomware attack. Since Change Healthcare processes claims and handles other financial processes prescriptions couldn't get filled and physicians ran low on cash.

On March 27, UnitedHealth said Change Healthcare could process medical claims, but its update page notes that some payer processes are being restored through April. The company also said it has provided more than $6 billion in advance funding and interest-free loans to care providers.

Couched in non-GAAP results and pro forma noise, you have to scroll to the bottom of UnitedHealth's first quarter earnings release to get a feel for the Change Healthcare costs. Here's the breakdown:

• $1.22 billion: First quarter net loss for UnitedHealth, but some of that was due to the sale of a subsidiary.
• $279 million: Business disruption impacts to UnitedHealth's Optum unit, which houses Change Healthcare. Business disruption impacts refer to revenue lost during the cyberattack.
• $593 million: Total direct response costs due to the cyberattack. Costs attributed to the Optum unit were $363 million.
• $872 million: Total UnitedHealth costs related to the Change Healthcare attack.
• $1.35 billion to $1.6 billion: Total cyberattack hit for 2024 as projected by UnitedHealth, or $1.15 a share to $1.35 a share.

Adjusted for the Change Healthcare fiasco, UnitedHealth reported earnings of $6.91 per share on revenue of $99.8 billion. Both figures topped Wall Street estimates. UnitedHealth's adjusted figures included the revenue hit to Change Healthcare and excluded direct response costs.

Broadcom CEO Tan says VMware customers can get support extensions amid subscription transition

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Broadcom CEO Hock Tan penned another missive to VMware customers arguing that the software vendor has lowered the price of VMware Cloud Foundation, poured money into research and development, benefited partners and will complete the transition to subscriptions.

Tan also noted that Broadcom is working with extending support contracts for VMware customers struggling with the transition to subscription pricing.

The latest blog from Tan comes as Reuters reported the European Commission has received complaints about Broadcom's VMware pricing changes and the regulator sent requests for information to Broadcom.

Broadcom has had a steady cadence of blogs that appear to be aimed at allaying VMware customer concerns. To date, Nutanix has been the biggest beneficiary of VMware customer angst. It's unclear whether Broadcom's blog barrage is hitting the mark, but the missives collectively acknowledge that VMware customers may be a smidge disgruntled. The rundown:

• April 15. Tan makes case for VMware pricing and offers a few concessions.
• March 14. Tan recaps first 100 days of VMware purchase, noted that "this level of change has understandably created some unease among our customers and partners."
• Feb. 14. VMware acknowledges that changes have been rough on customers, but a long blog post could be classified as mansplaining.
• Dec. 11: VMware outlined product rationalization and subscription plan.

Here's a look at the key points from Tan in order of importance.

Broadcom acknowledges that "fast-moving (VMware) change may require more time." Tan wrote:

"We continue to learn from our customers on how best to prepare them for success by ensuring they always have the transition time and support they need. In particular, the subscription pricing model does involve a change in the timing of customers' expenditures and the balance of those expenditures between capital and operating spending. We heard that fast-moving change may require more time, so we have given support extensions to many customers who came up for renewal while these changes were rolling out. We have always been and remain ready to work with our customers on their specific concerns."

Customers can keep their existing perpetual licenses for vSphere. Tan wrote:

"It is important to emphasize that nothing about the transition to subscription pricing affects our customers’ ability to use their existing perpetual licenses. Customers have the right to continue to use older vSphere versions they have previously licensed, and they can continue to receive maintenance and support by signing up for one of our subscription offerings.

To ensure that customers whose maintenance and support contracts have expired and choose to not continue on one of our subscription offerings are able to use perpetual licenses in a safe and secure fashion, we are announcing free access to zero-day security patches for supported versions of vSphere, and we’ll add other VMware products over time."

VMware is standardizing the pricing metric across cloud providers to per-core licensing to match its end-customer licensing. Tan said this standardization will allow enterprises to seamlessly move VMware Cloud Foundation on-premise to cloud and back if needed.

• VMware mansplains to customers about model changes, subscription pricing, account upheaval
• KKR buys VMware's EUC division from Broadcom for $4 billion
• Dell Technologies to exit commercial agreement with VMware
• Will VMware customers balk as Broadcom transitions them to subscriptions?
• BT150 CXO zeitgeist: AI trust, AI pilots to projects, VMware angst, projects ahead

Akamai adds Nvidia GPUs to cloud network, initially targeting media

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Akamai has added Nvidia GPUs to its distributed cloud network adding a service optimized for processing video content at the edge.

The cloud service, announced at the National Association of Broadcasters' (NAB) conference, is powered by Nvidia RTX 4000 Ada Generation GPUs.

Akamai has been steadily building out its distributed cloud infrastructure for multiple use cases including AI and machine learning workloads that require low latency near data.

• Akamai's IaaS vision: Become your cloud alternative
• Akamai launches Gecko, aims to combine cloud compute, edge networks
• Why generative AI workloads will be distributed locally
• Constellation ShortList™ Digital Performance Management

According to Akamai, the Nvidia RTX 4000 instances can process video frames per second 25x faster than CPU encoding and transcoding. Although the Akamai Nvidia-powered instances are initially aimed at the media industry, the company is also betting on other use cases.

These use cases include:

• Virtual reality and augmented reality content.
• Generative AI and machine learning workloads for training and inference at the edge.
• Data analytics and scientific computing.
• Gaming and graphics rendering.
• High-performance computing.

Akamai said it will continue to add GPU instances optimized by industries.

Foundation model debate: Choices, small vs. large, commoditization

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Foundational model debates--large language models, small language models, orchestration, enterprise data and choices--are surfacing in ongoing enterprise buyer discussions. The challenge: You may need a crystal ball and architecture savvy to avoid previous mistakes such as lock-in.

In recent days, we have seen the following:

• Google Cloud Next outlined how it is deploying Gemini 1.5 Pro across multiple use cases and business roles. The gist is that Google Cloud wants to use Vertex AI to be the model orchestration platform to build a bevy of generative AI agents. Google Cloud's Model Garden serves up more than 130 model choices, but there's no question that Gemini is the star. "We are now building generative AI agents," said Google Cloud CEO Thomas Kurian. "Agents are intelligent entities that take action to help you achieve specific goals."
• AWS highlighted its thoughts on open model choices and leveraging Bedrock combined with Q to be the generative AI platform of choice. AWS has the Titan model, but emphasizes model choices and is positioned to be the Switzerland of LLMs.
• Constellation Research's latest BT150 CXO call revolved around small models vs. large ones. Vendors such as ServiceNow have championed smaller use-case specific models for efficiency and costs.
• Anthropic CEO Dario Amodei said enterprises won't be in a position in the future of choosing a small custom model or a large general one. The correct fit will be a large custom model. LLMs will be customized for biology, finance and other industries.
• Archetype AI raised $13 million in seed funding and launched its Newton physical world foundational model. Newton is built to understand the physical world and data signals from a range of environmental sensors. Newton highlights how LLMs are being aimed at green field opportunities.
• For good measure, Microsoft will likely talk about building customized Copilots and model choices at Build in May. A few sessions on Azure AI Studio have highlighted orchestration.

Dion Hinchcliffe: Enterprises Must Now Cultivate a Capable and Diverse AI Model Garden

Now that's a lot to talk about considering how enterprises need to plow ahead with generative AI, leverage proprietary data and pick a still-in-progress model orchestration layer without being boxed in. The dream is that enterprises will be able to swap models as they improve. The reality is that swapping models may be challenging without the right architecture.

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

Will enterprise software vendors use proprietary models to lock you in? Possibly. There is nothing in enterprise vendor history that would indicate they won't try to lock you in.

The crystal ball says that models are likely to be commoditized at some point. There will be a time where good enough is fine as enterprises toggle between cost, speed and accuracy. Models will be like compute instances where enterprises can simply swap them as needed.

It's too early to say that LLMs will go commodity, but there's no reason to think they won't. Should that commoditization occur platforms that can create, manage and orchestrate models will win. However, there is a boom market for models for the foreseeable future.

AWS Vice President of AI Matt Wood noted that foundation models today are "disproportionately important because things are moving so quickly." Wood said: "It's important early on with these technologies to have that choice, because nobody knows how these models are going to be used and where their sweet spot is."

Wood said that LLMs will be sustainable because they're going to be trained in terms of cost, speed and power efficiency. These models will then be stacked to create advantage.

Will these various models become a commodity?

"I think foundational models are very unlikely to get commoditized because I think that there's just there is so much utility for generative AI. There's so much opportunity," said Wood, who noted that LLMs that initially boil the AI ocean are being split into prices and sizes. "You're starting to see divergence in terms of price per capability. We're talking about task models; I can see industry focus models; I can see vertically focused models; models for RAG. There's just so much utility and that's just the baseline for where we're at today."

He added:

"I doubt these models are going to become commoditized because we haven't yet built a set of criteria that helps customers evaluate models, which is well understood and broadly distributed. If you're choosing a compute instance, you can look at the amount of memory, the number of CPUs, the number of cores and networking. You can make some determination of how that will be useful to you."

In the meantime, your architecture needs to ensure that you aren't boxed in as models leapfrog each other in capabilities. Rapid advances in LLMs mean that you’ll need to hedge your bets.

Constellation Research analyst Holger Mueller noted:

"We are only at the beginning of the foundation model era. The layering of public LLMs that are up to speed on real world developments and logically merged with industry knowledge, SaaS packaging, and functional enterprise domain specific models are  going to be crucial for gen AI success. Bridging the gap from real world aware and fitting an enterprise makes genAI workable and effective."