As the spring tech show season comes to an end, large language models and generative AI have taken over keynotes, announcements and partnerships.
Audiences have been captivated, and stakeholders' minds have wandered to all the ways business and marketing could transform with this technology. But there have also been a number of smoke-and-mirrors demos, as well as simplistic implementations with API-centric ChatGPT integrations.
I've done a few detailed recaps on companies, including Qlik, Tableau and ThoughtSpot, and HPE. And although it's difficult to keep up with the amount and pace of announcements, a few key themes have emerged:
- Addressing skills gaps remains top of mind. And it's forcing businesses to prioritize partnerships with technology vendors to help fill those gaps.
- Time to value and speed of ramp-up garners vendor attention. The speed with which customers can get started is a priority, whether delivering a pre-trained model, enabling developers to build apps faster, making a readymade tool available to support emerging use cases, or ensuring rightsized infrastructure is delivered to maintain cost efficiency and performance.
- Open source is a growing priority. The robust open source AI ecosystem is delivering greater transparency, accountability, innovation and accessibility to a wide range of users. This drives greater adoption as more and more use cases get exposed to the world.
- Privacy is a major concern. As organizations scramble to maintain compliance and ensure data is not exposed to the outside world, the rise of private and -- in many cases -- smaller LLMs is driving the need for more guidance from experts and strategic partners.
- Vendor partnerships are more strategic than ever. Most vendors cannot do this alone. They will need AI-specific expertise from a software vendor in the open source community, such as Hugging Face; infrastructure ties for high-powered compute from Nvidia; domain expertise in a specific industry, such as Google Cloud's Med-PaLM for life science and healthcare; or access to strategic guidance through AI innovation centers, such as AWS.
With all that in mind, here's a list of recent generative AI and LLM announcements from nine major enterprise tech vendors, in alphabetical order, with my thoughts on each. Are there announcements from each of these vendors that are missing? Yes. Will there be more? Absolutely. Expect a summer and fall of more announcements, integrations and partnerships on the generative AI front.
1. AWS announces tools for building with generative AI, including Generative AI Innovation Center
AWS has announced several new tools and services to make it easier for developers to build generative AI applications.
Amazon Bedrock provides a foundation for building generative AI applications by making foundation models from AI21 Labs, Anthropic, Stability AI and Amazon accessible via an API. Amazon CodeWhisperer is an AI coding assistant that generates code suggestions in real time. Amazon also announced general availability of Amazon EC2 Trn1n instances powered by AWS Trainium chips and Amazon EC2 Inf2 instances powered by AWS Inferentia2 chips.
Additionally, AWS recently announced the Generative AI Innovation Center, a new program to help customers successfully build and deploy generative AI systems. The company is investing $100 million in the program, which will connect AWS AI and machine learning experts with customers around the globe to help them envision, design and launch new generative AI products, services and processes.
We know AWS is a leader in public cloud, and with that comes a massive customer list. I'm certain the company will be picking its customers' brains to identify the tools and use cases that matter most. AWS is partnering with some big names in the foundation model space, and I'm looking forward to seeing those partnerships expand, including more details about the depth of each integration.
A number that sticks out to me is the $100 million dollar investment in Amazon's Generative AI Innovation Center. Many will say, "Why such a small investment?" But it's important to crawl before you run. I expect that number to grow quickly as customers rapidly adopt this technology and look to AWS for help and guidance. With AWS re:Invent coming in the fall, I expect a huge splash as AI -- including LLMs and other generative AI -- takes center stage.
2. Databricks continues its generative AI push with open source and acquisitions
Databricks has been very active in the generative AI space lately. The company released free data for training AI models; announced Dolly 2.0, a free and open source ChatGPT-style AI model; worked with Hugging Face to integrate Apache Spark for faster AI model building; and acquired Okera to enhance data governance in LLMs.
And that was all before the Data + AI Summit. At Databricks' recent event, the company announced its acquisition of MosaicML to enable customers to securely develop and train LLMs in their own secure Databricks environment using their own data. Further, it announced new Lakehouse AI innovations that lets customers easily and efficiently develop generative AI applications, including LLMs, directly within the Databricks Lakehouse Platform.
Databricks continues to impress with its ability to execute here. I would argue the company is right up there with Google, Nvidia and Microsoft in the competitive race to enable customers to explore generative AI in a controlled environment.
In my opinion, Databricks' success is fueled by its consistent ties to open source technologies, as well as its quick recognition that the largest LLMs aren't necessary for many customers, especially in private deployments or domain-specific areas. And right now, Databricks' focus -- which I agree with -- is on letting customers build rightsized models for their specific use cases. A great example was shared at Databricks' recent event: You do not need the entire internet if your use case is dealing with health insurance policies.
3. Dell Technologies announces Project Helix for on-premises generative AI
Dell Technologies and Nvidia have partnered to create Project Helix, which aims to make secure, on-premises generative AI more accessible to businesses.
The partnership will provide a complete hardware and software stack based on Dell and Nvidia infrastructure and software via Dell Technologies Validated Designs to help businesses build and deploy generative AI applications. Several features are designed to ensure data privacy and security, such as encryption, access controls and audit logs. Project Helix is currently expected to become available in July 2023.
Dell Technologies is transforming every day as the company continues to prioritize generative AI throughout the whole business. The announcement is the right move, and this decision makes a lot of sense as an extension of Dell's existing partnership with Nvidia.
And, with so many other generative announcements focused on the public cloud, Dell's on-premises focus is necessary and differentiable, especially as organizations look to the future. Although the cost of generative AI doesn't appear to be holding organizations back just yet, organizations will hit a cost tipping point in cloud environments, whether due to the size and complexity of underlying models or the lack of resources available to customers in cloud environments. This is where I believe Dell has a great play with their edge story.
4. Google Cloud's multipronged generative AI approach
Several recent announcements span a range of areas within Google Cloud's portfolio. Vertex AI supports generative AI, including Model Garden, which enables customers to search, discover and interact with several models from Google and partners. Additionally, Generative AI Studio gives customers tools to tune and customize models, while Generative AI App Builder delivers enterprise search and chatbots to businesses. Duet AI aims to help developers with code completion, reviews and inspections, as well as supporting organizational collaboration within Google Workspace.
Google's new foundation models are also available to customers, including Codey for code generation, Imagen for text-to-image generation and Chirp for speech-to-text generation. Domain-specific models, such as Med-PaLM in life sciences and healthcare or Sec-PaLM in cybersecurity, are enabling enterprises and governments to build specialized AI tools with industry-specific data. And Google has a consulting arm to help businesses learn how generative AI can revolutionize their business while reducing risk and maintaining responsible development and deployment.
Looking past Bard -- and not focusing on the Bard vs. ChatGPT debate -- I believe Google Cloud is the leader right now given the strength of its generative AI offerings today.
Google currently offers tools for those in data and security roles, line-of-business leaders, developers and end users. The company has focused significantly on ramp-up time, and it's paying off: Google has one of the broadest partner ecosystems and a large presence in the open source community. And even if customers aren't using Google Cloud directly, odds are that a smaller generative AI partner is using Google Cloud to build and train its models, which are then consumed by customers.
With all that said, the biggest metric for me is customers, and Google has hundreds of customers experimenting with and using its robust set of entry points and tools. To maintain this leadership position, Google Cloud must stay ahead of the competition in understanding the use cases that matter most. And while I believe Google has the largest generative AI enterprise customer base today, the company must continue to look outward to understand what use cases to focus on next.
5. IBM Watsonx focuses on end-to-end AI workflows with data and governance
IBM recently announced the Watsonx platform, designed to help businesses build and deploy AI applications using foundation models. This includes a studio for training and deploying models, a data store for storing and managing data, and a governance toolkit for managing ethical and responsible AI.
IBM also announced partnerships with Hugging Face to make Watsonx more accessible to the open source community and with Adobe to deliver a content supply chain tool using generative AI. The company also plans to continue exploring industry-specific and domain-specific applications. And IBM Consulting has established a Center of Excellence for generative AI, with more than 1,000 consultants with specialized generative AI expertise.
I view this announcement as IBM reinventing the Watson brand. As this gets released in stages, starting in July and continuing through October, I believe that organizations will appreciate the holistic guidance from IBM across the data, analytics and AI lifecycle.
One item to keep an eye on here is that Watsonx for governance is currently scheduled to launch in October. I'm a firm believer that AI should start with governance and trust, so I'm hoping for availability sooner. The good news is that IBM has been a strong leader in the AI governance and responsible AI space for some time, so I don't think it's an afterthought. Additionally, IBM continues to expand its AI partner ecosystem to include companies such as Adobe, AWS, Microsoft, Salesforce and SAP.
6. Microsoft accelerating enterprise-ready generative AI and empowering developers
Microsoft's initial announcements stemmed from its deep partnership with OpenAI. The Azure OpenAI Service brings the power of OpenAI's GPT and Dall-E models to enterprises everywhere, backed by Microsoft's infrastructure.
Microsoft has also released Copilot across several products, including Microsoft 365, Bing, Edge, Dynamics 365 and Power Platform. This enables organizations to use LLMs with customer data or adopt AI-powered, no-code tools for software development. Additionally, a slew of partnership announcements with the likes of Nvidia and, most recently, Snowflake are enabling Microsoft to build new integrations with Azure ML, Azure OpenAI and Microsoft Cognitive Services.
Microsoft has been tightly tied to LLMs and generative AI since its investment in OpenAI. And although the company started with search integration, we've seen rapid expansion across its portfolio. Given its status as the leader in productivity and collaboration, it's no surprise Microsoft started with several Copilots. And knowing that Microsoft's developer community is the largest in the world, I expect announcements from Microsoft Build to be rapidly adopted as developers hope to use generative AI capabilities within their apps.
With that said, even after Microsoft Build, I continue to hear questions about enterprise traction. I recognize there are many use cases for developers and end users, but I'm looking forward to seeing more from the Azure ML and Power BI teams on use cases within the data stakeholder ecosystem. And with Microsoft Ignite coming up in the fall, be prepared for Microsoft to highlight the pervasiveness of generative AI throughout its portfolio, including customer traction.
7. Nvidia hardware, software and partnerships accelerating generative AI adoption
At the GTC Conference this spring, Nvidia announced a generative AI suite of cloud services called Nvidia AI Foundations. This suite contains pre-trained foundation models, frameworks, optimized inference engines and APIs to simplify building and consumption of customized generative AI on customers' terms.
The cloud services cover several use cases, including text with Nvidia NeMo, visual content with Nvidia Picasso and biology with Nvidia BioNeMo, all of which are powered by the Nvidia DGX Cloud. NeMo is a development framework that enables the building, customization and deployment of generative AI models with billions of parameters. Within the NeMo framework, NeMo Guardrails is a toolkit to enable customers to build generative AI applications that are trustworthy, safe and secure.
And on the partner front, Nvidia is partnered not only with everyone mentioned in this article but many more companies that are looking to enable customers to gain access to high-powered compute or tools to confidently build generative AI applications, including platforms such as Hugging Face, Cohere, Adept, AI21 Labs, Anthropic and Jasper.
If you're a key player in the generative AI market of any kind and don't have a partnership with Nvidia, you're behind. Nvidia can bring hardware and software angles to the table that help virtually every vendor in this space better complete its story. Whether customers need access to their GPUs in the public cloud, private deployments on premises or in a colocation facility for training LLMs, or anywhere in between, Nvidia can easily add value and likely should be part of the story.
And on the software side, Nvidia offers domain-specific platforms and expertise, clearly outlined use cases, a generative AI developer framework that includes safety and security, a growing open source ecosystem presence through strategic partnerships with all the big names and more. I recognize that GPUs and accelerated compute are driving headlines based on resource shortages, but I truly believe that Nvidia is underselling its software capabilities. I expect to hear more on that front going forward, especially as customers explore private LLMs within their own environments.
8. Oracle cloud infrastructure with native generative AI
Oracle announced it will provide native generative AI services built on Oracle Cloud Infrastructure (OCI), using Oracle's unique SuperCluster capabilities, in partnership with Nvidia. The goal is to accelerate training of LLMs while simultaneously reducing cost.
Oracle also recently announced a partnership with Cohere to provide the foundational models for its generative AI services. The services will be available in the Oracle Cloud Marketplace in the coming months. Customers can expect generative AI to be pervasive across Oracle, and its generative AI services will become generally available later this year.
Oracle's analytics, data science and AI services are a hidden gem. Oracle has big customers doing some amazing things across virtually all industries. The company also has in-house talent, a reputation for serving the enterprise -- and therefore a great understanding of enterprise requirements -- and a good track record when it comes to security, privacy and governance. I expect Oracle to lean in on the latter as the company continues to shape its generative AI story.
I also like the concept of embeddable generative AI services across Oracle's portfolio and cloud applications, including ERP, human capital management, supply chain management and customer experience. I think Oracle has the potential to deliver something differentiable with Cloud@Customer to enable customers to take advantage of generative AI in private cloud environments.
I do think Oracle is a bit exposed competitively with later general availability for its generative AI services. But at least in the interim, customers can experiment with Oracle's internal playground to try out Cohere models on OCI. Now it comes down to execution between now and Oracle CloudWorld in the fall to show all of this in action on the main stage.
9. Snowflake aims to help customers develop generative AI apps in the Snowflake Data Cloud
In May, Snowflake acquired Neeva, a search engine vendor whose platform is fueled by generative AI. Then, at Snowflake Summit 2023, Snowflake announced the private preview of Snowpark Container Services, a framework that developers can use to run containerized AI and machine learning workloads, as well as the launch of a new partnership with Nvidia that gives users access to Nvidia's GPUs in the cloud.
Snowflake also unveiled the private preview of Document AI, Snowflake's own LLM aimed at enabling users to derive deeper insights from documents. The company additionally showcased a new application framework in public preview on AWS, in which developers can access applications built by others within the Snowflake ecosystem and share and monetize their own custom-built applications.
As I mentioned earlier, Snowflake will also expand its strategic partnership with Microsoft in the realm of AI tools for data scientists and developers. This includes building new integrations between Snowflake Data Cloud and Azure ML, as well as making use of integrations with Azure OpenAI and Microsoft Cognitive Services.
Snowflake had been pretty quiet on the generative AI front building up to Snowflake Summit 2023; outside of the Neeva acquisition, we hadn't heard much. But Snowflake has since made it clear that generative AI is now a priority.
And although Snowflake is hoping to be the platform of choice for building generative AI experiences, assistants, copilots and user applications, it really comes down to customer data. All of the company's announcements are focused on enabling customers to use their data in Snowflake to embrace generative AI without having to move or copy it. The big question for me is: What if your data is in another environment?
The announced partnerships are with the biggest names you would expect, leading with Nvidia and Microsoft. So, while the announcements are exciting, Snowflake is still behind the competitive pack when it comes to generative AI. The good news is that the company has such a loyal customer base that I expect the announcements to be embraced, and I look forward to seeing the creative ways customers build generative AI applications to help them transform. And with this becoming a top priority, I expect a lot more from Snowflake throughout the rest of the year.