- The Daily Update
- Posts
- DeepRacer drives generative AI progress
DeepRacer drives generative AI progress
Spotlight: Georgia State uses AI to support disadvantaged students
Jack Shields
June 27, 2024
Welcome back!
Most of today’s news is coming straight from Amazon Web Services’ ongoing generative AI Summit in D.C. — (almost) no hyperlinks needed! Let’s get right to it.
In today’s Daily Update:
🗞️ NBC to use commentator’s voice clone at Summer Olympics
🤖 AWS DeepRacer tactics drive generative AI progress
📸 Georgia State leverages AI to support disadvantaged students
🚨 AI Roundup: Four quick hits
Read time: 3 minutes
TOP NEWS
🗞️ NBC to use commentator’s voice clone at Summer Olympics
Source: NBC News
NBC is planning to use an AI-generated version of legendary sportscaster Al Michaels’ voice during the Summer Olympics in Paris this summer.
The rundown:
Michaels joined NBC sports in 2006, but stepped away from regular Olympic coverage after the 2016 games in Rio.
NBC will use an artificial clone of the 79-year-old sportscaster’s voice to narrate daily streaming recaps of this year’s Olympics.
Michaels says he was “very skeptical” of NBC’s initial proposal, but was ultimately stunned by how accurately his voice clone mimicked his commentary style.
This is the first time that a major network is tapping into AI-generated commentary to cover a high-profile sporting event.
Why it matters: AI will probably never replace on-air talent entirely. Regardless, voice cloning technology carries significant implications for several industries (music, film, etc.) and the spread of deepfake content. This move by NBC highlights how advanced the tech already is.
AI INSIGHT
🤖 AWS DeepRacer tactics drive generative AI progress
Source: AWS
AWS DeepRacer helps developers of all skill levels learn machine learning (ML) using a fully autonomous 3D racing simulator. To demonstrate how ML approaches can drive advancements in generative AI, we’ll follow Commercient’s Max Allgaier through the process of building a DeepRacer model.
The process:
DeepRacer utilizes a specific type of ML called reinforcement learning, which deploys a “reward function” to incentivize models to complete desired actions on the track.
Allgaier’s first step was defining the actions that his virtual car could take in its environment. In this context, actions mainly refer to the car’s steering angle and speed.
He then coded a reward function that awarded his car points for navigating turns in advance and maintaining optimal positions on the track.
Allgaier's model trained itself on a virtual track through trial and error, learning which actions earned more points based on the reward function.
The resulting model is able to drive a real model car on a physical track.
Why it matters: DeepRacer is certainly a fun introduction to reinforcement learning, but this process also demonstrates how models trained in similar environments can be transferred to real-world applications. The same concept can be applied to generative AI models that are trained to handle complex language and image generation tasks.
INDUSTRY SPOTLIGHT
📸 Georgia State leverages AI to support disadvantaged students
DALL-E 3
Georgia State University plans to pilot an AI-powered personalized financial aid system this fall in an effort to boost graduation rates among students from disadvantaged backgrounds.
What you should know:
In 2003, Georgia State realized that students from disadvantaged backgrounds were graduating at significantly lower rates than white students from high-income households.
Since then, the Atlanta-based university has equipped its faculty with data technology that has helped raise African American and Hispanic graduation rates by roughly 200%.
Georgia State is now turning to a personalized financial aid system built on Anthropic’s Claude chatbot through Amazon Bedrock.
The system matches student data with highly relevant scholarship and financial aid applications, and curates action steps students can take to receive the aid.
Why it matters: Georgia State’s Senior Director of Student Success Analytics Ben Brandon says institutions can inadvertently hinder students’ success through their policies and practices. Universities can leverage data to identify and understand institutionally created barriers. Deploying AI can help institutions create more effective systems that improve everything from enrollment rates to graduation outcomes.
MORE TRENDING NEWS
🚨 AI Roundup: Four quick hits
Source: Figma
THAT’S ALL FOR TODAY