Google Gemini partially launched this week with claims that a forthcoming high-end version will outperform OpenAI's GPT-4. But for software developers, choosing an AI coding assistant isn't just a matter of which foundational model has most recently leapfrogged another.
On the heels of Microsoft adding support for GPT-4 Turbo to Windows 11 and search engine Bing this week, Google launched a comparable Pro version of Gemini 1.0 for its Bard search engine and the lower-end Gemini Nano for Pixel 8 Pro phones.
Starting Dec. 13, developers will have access to Gemini Pro in Google AI Studio and Google Cloud Vertex AI. The high-end Gemini Ultra, for which Google published early testing benchmarks that exceed GPT-4, will be released early next year.
Google's AI updates this week also encompassed deeper layers of the generative AI IT stack. It released a new version of its tensor processing unit (TPU), which was used to train Gemini, as well as a Google Cloud AI Hypercomputer service -- a combination of hardware, software and resource management utilities similar to those used to train Gemini. Hypercomputer utilities include a new Dynamic Workloads Scheduler to address demand for GPU and TPU capacity in cloud computing.
Google's Palm 2 large language model (LLM) was seen as less mature than OpenAI's GPT, especially once GPT-4 became available in March. AWS made strides in its own AI-specific chips with the AWS Trainium and Inferentia product lines, and it formed a partnership with Nvidia.
With Gemini, Google has arguably surged ahead on all those fronts, at least in terms of theoretical technical specs, said Andy Thurai, an analyst at Constellation Research.
"Google is taking [the lead] on competitors in three major categories -- on OpenAI/Microsoft ChatGPT with their Gemini announcement, on AWS/Nvidia with their infrastructure [and] TPU chips ... [and] on IBM/HP/Oracle with their Hypercomputer," Thurai said.
At least on the surface, Gemini also has some key potential differentiators from OpenAI and Microsoft's GPT-4 based tools, he said.
"First, it is multimodal from the ground up. Technically, this means this LLM could cross the boundary limitations of modalities," such as text, code and image data, Thurai said. "Second, they also released three model sizes rather than one-size-fits-all. Third, [it will have] a lot of safety guardrails to avoid any toxic content."
This last point, however, has already been a sticking point for Google Gemini, which was reportedly delayed from an original planned launch this month in part because of issues with its chat responses in languages other than English. Google spokespeople including CEO Sundar Pichai were careful to emphasize trust and safety in a blog post and video associated with this week's launch.
"We're approaching this work boldly and responsibly," Pichai wrote. "That means being ambitious in our research and pursuing the capabilities that will bring enormous benefits to people and society, while building in safeguards and working collaboratively with governments and experts to address risks as AI becomes more capable."
Developers already entrenched on GitHub Copilot
The overall generative AI arms race remains at its early stages, and it is anyone's to win: Most enterprises do not yet have generative AI tools broadly deployed in production, according to data from TechTarget's Enterprise Strategy Group (ESG).
In an August 2023 survey of 670 IT professionals, ESG researchers found that just 4% had deployed generative AI widely in production, while 14% were at the early stages of production use, 24% were at a pilot phase, and 27% were considering but had no specific plans to deploy generative AI yet.
"That leaves plenty of room for market education and competition," said Paul Nashawaty, an ESG analyst, in an interview this week.
David StraussFounder and CTO, Pantheon
All of the big three cloud providers are competing in several areas, from chips to models to coding assistants. Each brings historical advantages to the competition, from the popularity among consumers of Google Workspace products, which will soon integrate Gemini within the Duet AI product line, to AWS' considerable clout in cloud infrastructure, which holds 40% market share in IaaS, according to Gartner.
In the software development field, with AI code generation and remediation, Microsoft retains a firm first-mover advantage, said David Strauss, founder and CTO at Pantheon. The WebOps vendor is a customer of Google Cloud Platform (GCP) and Workspace, but its software engineers use Microsoft GitHub Copilot.
"Google has a big problem in the market for code assistants, which is that they don't own a GitHub," or an IDE as widely adopted as Microsoft's Visual Studio Code, Strauss said. "Google could have the most incredible code assistant system ever ... and it will get some traction, but Microsoft's story here is really comprehensive all the way from cloud-hosted development environments to the Azure cloud platform."
So far, Google's performance claims for Gemini only just edge out GPT-4 -- it doesn't blow GPT-4 out of the water, Strauss said. Thus, it's unlikely to compel developers to switch tools.
One exception to this is when data subject to data sovereignty regulations is concerned, which is already stored in Google Cloud regions for Pantheon.
"If you look at how GDPR works with sub-processor announcements and data hopping over various boundaries, it's a pain to add vendors for this stuff," Strauss said. "I told our [developer] team working on one of our products, 'I want you to use Google's Vertex AI on GCP and not OpenAI' because we have a huge partnership with Google, we're already in those regions, [and] our customers care a lot about where their data goes."
Overall, however, analysts said this data gravity factor is ultimately more likely to favor AWS, with its commanding lead in cloud infrastructure.
"Microsoft has the advantage of having the highest market share among source code management tools. With AWS having the highest cloud market share and many enterprise developers using the platform, the connection of CodeWhisperer to AWS services will be a differentiator," said Larry Carvalho, an independent analyst at RobustCloud. "Google Duet AI is a distant third in demand."
AWS generative AI updates emphasize model choice
AWS made generative AI waves of its own last week at its annual re:Invent conference.
The biggest generative AI news out of the conference was Amazon Q, a broad-based virtual assistant launched in preview Nov. 28 and built on the Amazon Bedrock service. Amazon Bedrock is a managed service that exposes multiple foundation models through an API, including AI21 Labs' Jurassic-2, Anthropic's Claude, Cohere's Command and Embed, Stability AI's Stable Diffusion, Meta's Llama 2 and Amazon's own Titan models.
Amazon Bedrock has amassed more than 10,000 users since its launch in September, according to a keynote presentation by AWS CEO Adam Selipsky on Nov. 28. The flexibility and choice of models behind Bedrock are key advantages for the products AWS builds on top of it, including Amazon Q, Selipsky said.
"There's not going to be one model to rule them all, and there's certainly not going to be one company providing the models that everybody uses," he said.
At least one industry observer said he agreed that Amazon Bedrock's flexibility would prove a key differentiator for AWS generative AI tools -- along with transparency from AWS as it develops tools such as Amazon Q for Java code transformation, which the company disclosed is based in part on the OpenRewrite project.
"Amazon's going out there and integrating these different open source tools that exist, and being public about it, which then gives developers more confidence in it, even if they can't modify it," said Donnie Berkholz, founder and chief analyst at Platify Insights, a tech industry analysis firm. "Sharing that view inside the black box really is different from [Microsoft] Copilot, which has really been, 'Hey, you just have to trust us. We're going to recommend things [and] you can take it or leave it.'"
Amazon Q Code Transformation is available in preview for customers on the Amazon CodeWhisperer Professional Tier in the AWS Toolkit for IntelliJ IDEA and the AWS Toolkit for Visual Studio Code.
Other IT-focused Amazon Q integrations include a conversational AI assistant embedded in the AWS Management Console -- similar to Microsoft's Copilot for Azure, GitHub's Copilot Chat and Google's Duet AI for GCP. Amazon Q can be used through the AWS Management Console to optimize Elastic Compute Cloud instance selection and troubleshoot errors; another integration with the Amazon Virtual Private Cloud Reachability Analyzer can be used for network troubleshooting. With Amazon's Kendra data retrievers and its pledges about its data privacy and access controls, Amazon Q could also be used for retrieval-augmented generation systems similar to Glean AI's.
The preview version of Amazon Q is integrated alongside AWS Application Composer and the Amazon CodeWhisperer code generation tool into the Visual Studio Code IDE, as well as the Amazon CodeCatalyst unified app development service.
"You could argue that Amazon will use this to steer Microsoft C# developers away from Azure and toward AWS, but with GitHub Copilot support already available for Visual Studio, Amazon needed to release something or cede the space," said Andrew Cornwall, an analyst at Forrester Research. "It'll be helpful to the C# developers who use the AWS ecosystem."
Beth Pariseau, senior news writer at TechTarget Editorial, is an award-winning veteran of IT journalism. She can be reached at [email protected] or on Twitter @PariseauTT.