For more crisp and insightful business and economic news, subscribe to The Daily Upside newsletter. It's completely free and we guarantee you'll learn something new every day.
Amazon wants to help AI models learn even quicker.
The company is seeking to patent a method for "fast annotation of samples" for developing machine learning models. Amazon's system would essentially be The Sims for AI developers: It allows them to run experiments, discover and take note of problems, and make slight changes until they get their preferred outcome.
Here's how this works: Amazon's system offers a "model development environment" that allows a development team to perform "iterative model experiments" to develop machine learning models, specifically for media-based models (aka, those that control things like facial recognition, text and speech processing, autonomous vehicle control and more).
This development environment is equipped with a "media data management interface" that allows its users to annotate and make changes to the model's training data as they run experiments. Users can run both training runs and test runs of the models in this environment. This system also comes with a "diagnosis interface," which breaks down how a model performs to help users determine what needs to be fixed.
Amazon noted that machine learning model training and development is typically time-consuming and error-prone, as it involves a ton of labor-intensive, tedious human-led tasks and close supervision by data scientists. Plus, once that training is done, novice users often can't diagnose problems with the model or determine how exactly to fix it.
Like most tech companies, Amazon has been recently loud about its commitment to AI. As it races to play catch-up, the company last week vowed to invest $100 million in what it calls the AWS Generative AI Innovation Center, which will help its cloud tenants use generative AI in healthcare, financial services and manufacturing. In the consumer space, the company is building a conversational AI chatbot for "reimagining Amazon Search," according to a job description reported by Bloomberg in May.
But Amazon has some fierce competition. Both Google and Microsoft are making headlines for infusing AI into just about every move they make. However, Amazon may have a leg up as the AI race continues: its dominance in cloud technology, said Irusha Peiris, retail analyst and portfolio manager for William O'Neil. Amazon isn't too far behind competitors, he said, but boosting its AI offerings to serve up to its already massive AWS customer base will likely help keep them from moving over to Azure or Google Cloud.
"Whenever they come out with more AI offerings, it makes it that much easier to upsell to those customers to help them take more advantage of their data," Peiris told me. "Amazon really has to -- everyone has to, on the cloud side -- have some of these offerings. If you don't, all the sudden, you're severely lacking in something that's in tremendous demand."
Amazon is facing more than just competition. The company, along with the industry as a whole, faces a wave of rhetoric from lawmakers, critics and technologists about the future that AI could bring. In March, a group called The Future of Life Institute started a petition to pause "giant AI experiments," which caught the attention of the likes of Elon Musk, Steve Wozniak and Andrew Yang.
Since then, more insiders have stepped forward warning about AI's danger. In early May, AI trailblazer Dr. Geoffrey Hinton stepped down from his position at Google to "speak freely" about the risks of the tech. Later that month, the Center for AI Safety released an open letter claiming that AI poses a "risk of extinction" that needs to be mitigated, signed by Hinton as well as OpenAI's Sam Altman, Google DeepMind's Demis Hassabis, Bill Gates and dozens of other tech leaders.
But as it stands, big tech firms show no signs of stopping their breakneck AI race. If this patent filing reveals anything, they're only interested in speeding things up.
Have any comments, tips or suggestions? Drop us a line! Email at [email protected] or shoot us a DM on Twitter @patentdrop. If you want to get Patent Drop in your inbox, click here to subscribe.
