Cloud

AWS finally gets serious about genAI


Love him or hate him, Duckbill Group chief economist and AWS snarkologist Corey Quinn is often right in his assessments of AWS product strategy—but not always.

Sometimes Quinn is completely wrong, like last week when he claimed “AWS is just a fundamentally unserious company when it comes to genAI.” This after the successful launches and customer adoption of Amazon CodeWhisperer (coding assistant), Amazon Bedrock (makes it easier to use a variety of large language models), and more, including the re:Invent mainstage release of Amazon Q (a genAI-powered chatbot).

Quinn’s comment might have passed muster a year ago when AWS was jockeying to be taken seriously alongside Microsoft, which had a deep partnership with and investment in OpenAI, and Google, which for years had been open sourcing impressive data science/AI-related projects. But in the wake of re:Invent, it’s clear that AWS is a fundamentally serious company when it comes to generative AI and AI, generally.

Putting genAI to work

To be fair to Quinn, he was making a gross generalization based on one erroneous data point: the pricing for Claude (Amazon Bedrock Edition). But pricing is set by Anthropic, an AWS partner (one that just received up to $4 billion from AWS). And it’s not even necessarily the big price change Quinn thought he’d spotted. Still, although Quinn was wrong in this particular instance, he’s often right to call “BS” on AWS product releases. Just not this time. AWS is clearly “fundamentally serious [about] genAI.”

At least, I think that’s what 10,000 Bedrock customers would say. The product just hit general availability in September and already Adobe, BMW, United Airlines, Merck, and at least 9,996 other customers are using it, according to AWS CEO Adam Selipsky’s keynote. Ditto CodeWhisperer customers like Buildstr that have seen security vulnerabilities decrease from using the tool. Following that topic of security, another commentator praised the “integration of generative AI into Amazon Detective and Amazon Inspector” as “a significant leap in cloud computing security and code management.”

If AWS were playing catch-up in AI (something I’m sure they’d dispute), at re:Invent they went past their competition in some key ways. For example, in Selipsky’s keynote, he announced that CodeWhisperer customizations (the ability to add internal libraries, APIs, etc., to improve output) were generally available to everyone, versus GitHub preannouncing theirs (but not making them available) and Google Cloud offering a gated preview of sorts. AWS also is unique in offering code transformation/language upgrade capabilities. In general, AWS made a point of announcing a variety of services and features that are available now, rather than in the future, a nice change from recent re:Invent launches.

This brings us to Q, a new service that offers AWS a big step forward in genAI and beyond.

A dramatically different Q

I’m sure the folks inside AWS understood the unfortunate associations “Q” would have with extreme conspiracy types, but they also must have banked on the James Bond association with “Q,” the R&D department of MI5 in 007’s world. They’re going for “cool” instead of “crazy.” Long term, they’re probably right (though this is an area where Quinn’s commentary isn’t necessarily incorrect).

Short term, Q is a definite line-in-the-sand moment for AWS, establishing their “fundamentally serious” investment in genAI.

Q, argues RedMonk analyst and cofounder James Governor, “is going to be a big deal,” in part because it offers a compelling vision of “the future of docs,” alongside Microsoft’s release of Copilots across its product lines. Additionally, Governor says, Q provides “an abstraction that could help” unify AWS services that have “become too complicated and unwieldy.”

Unlike Microsoft, which promises a multitude of Copilot assistants, AWS’s approach is dramatically different. As Governor explains, AWS offers just “one Q that is cross-service … across all products for ops, troubleshooting, development.” With Q you should be able to pick up a conversation in one console and continue it on another console (and, eventually, have a conversation start in your IDE, if you’re a developer, and have it persist to your console). AWS already offers Q across its documentation, website, Slack/Teams, mobile app, etc. It’s almost bizarre that Microsoft, the company that has historically offered customers a very unified experience, has shipped its org chart with many Copilots, as if that’s a virtue, whereas AWS, which has historically offered all sorts of competing, “purpose-built” services as a virtue, is the one giving a unified, cross-service experience. This unified approach will bring serious advantages out of the box.

Also, although AWS didn’t really play this up at re:Invent, Q just turned AWS into the productivity application vendor (think Salesforce or Microsoft) that it largely has never been. “This is the next squillions of dollars for the company,” Governor observed. “Hey look, we dropped Q into Amazon Connect for call centers.” AWS is now an application vendor, with the connective tissue (Q) to help it turn data into answers for enterprises. It’s a big deal.

A marketplace strategy

This is not to say that somehow AWS won the genAI sweepstakes with a few announcements, however hefty in importance. First, as with Microsoft’s and Google’s announcements, there’s still plenty of assembly required. Everyone is selling a genAI vision more than a ready-made genAI reality. Even so, until now AWS wasn’t really in the conversation. Now they are, and in some cases, they’re leading that conversation.

Also, somewhat refreshing for any tech vendor but particularly for AWS (with 1.2 trillion cloud services, give or take), AWS’s genAI message isn’t that enterprises must go all in with AWS. AWS was quick to suggest more of a marketplace strategy where customers can pick and choose the right model for the job. For example, if you ask Q which models it uses, it’ll tell you it’s built on Bedrock and uses a variety of models depending on the task. A more humble AWS? Definitely maybe.

Also, a fundamentally serious company about genAI.

Copyright © 2023 IDG Communications, Inc.





READ SOURCE

This website uses cookies. By continuing to use this site, you accept our use of cookies.