Can’t Choose the Right AI Model? Amazon Wants to Help
BY PYMNTS | NOVEMBER 30, 2023
|
For nearly a year, there has been one clear generative artificial intelligence leader: OpenAI.
But now, with generative AI technology advancing rapidly — and hastily patched-up cracks showing at Sam Altman’s OpenAI — the marketplace is starting to get chippy.
This week, Amazon, which industry observers have tended to view as lagging behind its Big Tech competitors like Microsoft and Google in the AI race, has taken the gloves off against its peers.
The company unveiled a slew of AI-centric announcements at the Amazon Web Services (AWS) re:Invent event taking place this week in Las Vegas. The innovations show that the Seattle-based giant is taking its position in the AI field seriously — and that it wants to move to the front of the pack.
On Tuesday (Nov. 28), AWS announced Amazon Q, a new generative AI corporate chatbot that is “specifically for work and can be tailored to a customer’s business.” It also launched a new partnership with NVIDIA to deliver advanced AI infrastructure and added four new generative AI services to its cloud contact center, among a drumbeat of other announcements.
On Wednesday (Nov. 29), AWS launched new multimodal image generation capabilities for its Titan large language model (LLM) as well as a new AI model benchmarking service called Model Evaluation on Bedrock, available in preview.
The product announcements paint a compelling picture made up of data points that not only delineate how the AI commercial ecosystem is evolving but also underscore where Amazon wants to play and compete.
Read also: Amazon Enters Corporate Chatbot Race, Looks to Compete on Cost
Crafting a Compelling Enterprise Value Proposition
Amazon wants to take the lead in supporting enterprise companies building generative AI projects, and one way it is looking to differentiate itself is by emphasizing the choices that customers have on its platform.
After all, Microsoft — for all its success partnering with OpenAI — still has only products from one AI company to offer its own business customers, those of OpenAI.
Among its own models, AWS provides enterprise access to third-party foundation models including Anthropic’s Claude 2.1, Meta’s Llama 2, and Stability AI’s image generator.
Amazon invested over $1 billion into Anthropic and may end up taking a $4 billion stake down the line, and that sizeable check allowed it to become the first cloud provider to support the latest Claude model, which was released Nov. 21 at the height of the OpenAI drama.
But with so many neural networks available on AWS’s Bedrock application programming interface (API), how is an enterprise to know which model is the best fit for their needs?
After all, when a customer goes to Microsoft or Google, or another AI competitor, they don’t have to think about the options at hand. And without a way to transparently test models, organizations may end up getting locked into expensive subscriptions for AI programs that may be too large for their particular use case, or even not accurate enough for the domain-specific goals they may have.
“Knowing about AI will let people who use the tool understand how it works and do their job better,” Akli Adjaoute, founder and general partner at venture capital fund Exponion, told PYMNTS in an interview posted Thursday (Nov. 30). “… Just as with a car, if you show up to the shop without knowing anything, you might get taken advantage of by the mechanic.”
That’s where the new Model Evaluation on Bedrock service from AWS comes in.
See also: Tailoring AI Solutions by Industry Key to Scalability
Choice Is Good, but Knowledge Is Better
“I’m happy to share that you can now evaluate, compare and select the best foundation models (FMs) for your use case,” wrote Antje Barth, principal developer advocate for generative AI at AWS, in a blog post announcing the new solution.
“With automatic model evaluation, you can bring your own data or use built-in, curated datasets and pre-defined metrics for specific tasks such as content summarization, question and answering, text classification and text generation,” she added in the post.
The benchmarking tool offers both automated evaluation and human evaluation of models, and the goal is not to provide an industry standard but instead to help companies measure the impact of various models on their specific projects before committing to one solution or another.
Typically, developers looking for an AI solution would undergo bespoke assessments on their own time using their own resources, which requires a significant amount of custom code. Amazon is looking to streamline that process and take the manual work out of it.
For further reading on choosing the best AI solution, the PYMNTS Intelligence “Generative AI Tracker®,” a collaboration with AI-ID, sorts the myths from the realities of AI and explains how businesses can leverage AI technology wisely and effectively.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
