Almost in the blink of an eye, generative artificial intelligence models have captured society’s collective imagination. It’s been called a “game-changer” with huge implications for technologists, investors, policymakers and broader society. 

And there are questions about how platforms such as ChatGPT will – and won’t – change the face of business.

But large foundational models including ChatGPT and Stable Diffusion are pre-trained on huge volumes of data. That poses some significant challenges when it comes to bringing these innovations to production, and as the companies involved have stated, it’s all a work in progress.

“We know that many limitations [remain], and we plan to make regular model updates to improve in such areas,” wrote OpenAI researchers in a Nov. 30, 2022 blog post.

During the “Top Startups Building Generative AI on AWS” event during the latest installment of the AWS Startup Showcase series, theCUBE industry analysts John Furrier and Lisa Martin will connect with leading founders and visionaries of companies seeking to utilize this technology to build out foundational models to production, including Neuralmagic, OctoML, Astonomer, Roboflow, Anyscale and ArthurAI. (* Disclosure below.)

Considering ChatGPT

Based on the advanced GPT-3 neural network OpenAI first debuted in 2020, ChatGPT has quickly become the poster child for generative AI. The technology uses a method known as reinforcement learning, where researchers give AI systems tasks and then provide them with feedback on how to complete those tasks, which leads to the AI finding ways to improve its accuracy.

With ChatGPT, the feedback used was provided by a second neural network. OpenAI trained the second neural network on a dataset of pre-prepared questions and answers. 

“Many lessons from deployment of earlier models like GPT-3 and Codex have informed the safety mitigations in place for this release, including substantial reductions in harmful and untruthful outputs achieved by the use of reinforcement learning from human feedback (RLHF),” wrote OpenAI researchers.

But what challenges remain beyond the model architecture and the data science, including large-scale training, fine-tuning and inference optimization?

Building generative AI

To solve these challenges, many companies seek to build on top of Amazon Web Services Inc. as they share their best practices for bringing foundational models to production.

One such company is Neuralmagic Inc., which seeks to enable capabilities to ease developers’ jobs as AI and machine learning play a more involved role in the next generation of enterprise computing.

“It’s not good enough just to work on models; you’ve got to put them into production,” said Brian Stevens, chief executive officer of Neuralmagic Inc., during a recent interview with theCUBE’s John Furrier. 

“So, what we do is we make it easier to optimize the models that have been developed and trained and then try to make it super simple when it comes time to deploying and managing them in production.”

In Stevens’ view, the outlook for growth in AI and ML couldn’t be brighter, as the industry seeks to tackle countless potential use cases in the coming months.

A look ahead

Another company seeking to innovate in this field is OctoML Inc., which targets machine learning workload deployment.

Luis Ceze, co-founder and chief executive officer of OctoML, said the company was spun out of the University of Washington by the creators of Apache TVM, an open-source stack for ML portability and performance.

“We saw a major pain point and built OctoML, which is about three and a half years old now,” Ceze told Furrier. “And the company’s mission is to enable customers to deploy models very efficiently in the cloud and enable them to do it quickly, run fast, and run at a low cost, which is something that’s especially timely right now.”

And OctoML and Neuralmagic are just two companies seeking to innovate in a field that is growing more varied and challenging by the day. Learn about more startups building generative AI on AWS during the AWS Startup Showcase event on March 9.

TheCUBE event livestream

Don’t miss theCUBE’s coverage of the AWS Startup Showcase: “Top Startups Building Generative AI on AWS” event on March 9. Plus, you can watch theCUBE’s event coverage on-demand after the live event.

How to watch theCUBE interviews

We offer you various ways to watch theCUBE’s coverage of the “Top Startups Building Generative AI on AWS” event, including theCUBE’s dedicated website and YouTube channel. You can also get all the coverage from this year’s events on SiliconANGLE.

TheCUBE Insights podcast

SiliconANGLE also has podcasts available of archived interview sessions, available on iTunes, Stitcher and Spotify, which you can enjoy while on the go.

SiliconANGLE also has analyst deep dives in our Breaking Analysis podcast, available on iTunes, Stitcher and Spotify.

Guests

During the “Top Startups Building Generative AI on AWS” event during the AWS Startup Showcase event, theCUBE analysts will talk with leading founders and visionaries of companies seeking to utilize generative AI on AWS, including Neuralmagic, OctoML, Astonomer, Roboflow, Anyscale and ArthurAI.

(* Disclosure: TheCUBE is a paid media partner for the AWS Startup Showcase: “Top Startups Building Generative AI on AWS” event. Neither Amazon Web Services Inc., the sponsor for theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Image: SiliconANGLE