Welcome to our This Week in AI roundup. Our goal with this roundup is to provide an overview of the week's most important news and industry developments.
This week we have stories about Tesla AI day, DeepMind's new algorithm PonderNet, and more.
Top Highlights from Tesla AI Day
Tesla hosted their annual AI day on Thursday at which Elon Musk said he wants the company to be "much more than an electric car company." He also said that their AI activity on the hardware, inference, and training level all have applications beyond self-driving cars, most notably the company is currently building a humanoid robot.
Below are the top 4 highlights from Tesla AI day reported by Techcrunch:
- Tesla Bot: The company presented an early prototype of their humanoid robot that they are hoping to release next year.
- D1 Chip: The company also presented Tesla's new D1 chip, which is designed in-house and meant to run on their supercomputer, Dojo.
- Full Self-Driving & More: Musk noted that Dojo is not just meant for Tesla's full self-driving system—they also hope to open it up to other car companies and other applications down the line. Dojo is expected to be fully operational by next year, at which point we'll know more about other potential applications.
- Computer Vision Problems: Tesla reiterated their computer vision-based approach to FSD, which they described as "building an animal from the ground up". The two main problems they're focusing on are temporary visual impediments and remembering information from earlier road signs (ex. that two lanes are merging in 500m).
As Musk stated, Tesla AI Day is about encouraging people to work on solving real-world AI problems:
We basically want to encourage anyone who is interested in solving real-world AI problems at either the hardware or the software level to join Tesla, or consider joining Tesla.
DeepMind Introduces PonderNet, which Allows the AI to "Think for A While"
DeepMind introduced a new algorithm called PonderNet, which is a way for a neural network to think about its answer for a while before answering. The goal with this algorithm is for the AI to generalize better after training, and answer more challenging questions.
PonderNet will essentially help the AI extrapolate beyond its training data and, ideally, achieve a higher degree of accuracy in complex problems, for example, for questions that require multi-step reasoning.
Top 15 Quantum Computing Stocks to Watch
Quantum computing is beginning to transition from its experimental research phase to commercially viable machines capable of solving the world's most complex problems. In this guide, we look at 15 quantum computing stocks that are entering the space and pioneering the industry.
Baidu Unveils a "Robocar" as it Prepares for Self-Driving Cars
Baidu unveiled Kunlun 2, a second-generation AI processor aimed at applications like as self-driving cars. The Chinese search engine company also unveiled a prototype "robocar", rebranding its robotaxi app. These moves are all part of a larger shift for Baidu away from advertising and toward artificial intelligence and self-driving cars.
That's it for this edition of This Week in AI, if you were forwarded this newsletter and would like to receive it you can sign up here.