Tesla

Recently, Elon Musk took to Twitter to announce that Tesla AI Day #2 is scheduled for August 19. Similar to last year's event, Musk stated, "The aim of AI Day is to attract top talent in AI, software, and chip development to Tesla." Personally, I’m more intrigued by Tesla's advancements in AI rather than its vehicles. I’m eager to learn more about developments regarding the Models, Dojo chips, and the anticipated Tesla Bot at this upcoming AI Day #2. What are you looking forward to?

Apple silicon & PyTorch

On May 18, the popular machine learning framework PyTorch, primarily created by Facebook's AI research lab, announced it will support GPU-accelerated training on Macs starting with PyTorch v1.12. This means users can now train larger networks or batch sizes directly on their M1 Macs, reducing expenses tied to cloud development or the need for additional local GPUs. While there’s a significant boost in speed for GPU training and evaluation compared to CPU performance, a notable gap remains when compared to local systems utilizing Nvidia GPUs. Refer to the test results shared by a Twitter user.

In my opinion, the Apple M1 GPU currently lacks the API capabilities of Nvidia’s CUDA. It will require further support from machine learning frameworks and optimization efforts to fully harness its potential. For now, the M1 Mac is best suited for tasks like prototyping and fine-tuning locally.

Google I/O 2022

While I often find Google’s keynotes less compelling than those of Apple, the 2022 Google Keynote still featured several noteworthy points.

1. Machine Learning in Everything

   Google's products—Maps, Gmail, Translate, Photos, YouTube—are all deeply integrated with machine learning. When I think of machine learning and AI innovation, Google immediately comes to mind. Their initiatives such as Google Brain, Waymo, DeepMind, and Transformers are led by top scientists producing cutting-edge research.

2. Google Tensor Chip

   The Google Tensor chip is a premium system on a chip (SoC), akin to Apple’s M1 series. For those who have utilized TPUs for accelerating neural network training on Google Cloud, you understand the significance of this technology. TPUs, or Tensor Processing Units, are specialized AI accelerators developed specifically for neural network machine learning, particularly in conjunction with TensorFlow on Google Cloud.

Nowadays, Google employs a combination of TPU, CPU, and GPU technologies to create SoCs for its Pixel phones. This trend toward advanced packaging is becoming the norm in chip manufacturing.

3. Google AR Glass

   Toward the end of the I/O keynote, Google's CEO unveiled an early version of Google AR Glass, which offers real-time translations across multiple languages, including sign language.

This development raises the question: If Google can launch AR glasses, how long until Apple follows suit?

VS Code Update (April 2022)

One of the key features in the latest VS Code update (version 1.67) allows users to drag and drop files into the editor to create Markdown links. By holding the Shift key while dragging a file or image over the Markdown editor, you can easily drop it in and establish a link.

MarkDoc

In conclusion, this week’s tech updates highlight significant advancements and ongoing developments across various platforms. Stay tuned for more news, and thank you for reading! See you next week!

Inside Wales Tech Week: Diversity Panel Experience | Designer Diaries - This video features discussions on diversity in tech, showcasing various perspectives and experiences.

Week In My Life | Staying Inside, Preparing For Tech Interviews, and Deep Cleaning - A glimpse into the life of someone preparing for tech interviews while managing daily responsibilities.