- Starter AI
- Posts
- Nvidia’s new AI chip rivals, Sora’s first impressions, MIT presents DMD
Nvidia’s new AI chip rivals, Sora’s first impressions, MIT presents DMD
We're at the front of the AI chip battle.
Sthefania
March 26, 2024
Hello, Starters!
If you've been following our coverage, you're probably aware that Nvidia is the leader when it comes to chips. AI companies rely on them. But what would happen if they were to challenge that dominance?
Here’s what you’ll find today:
Introducing: The Unified Acceleration Foundation (UXL)
OpenAI releases Sora’s first impressions
MIT researchers present “Distribution Matching Distillation”
Voice Engine is OpenAI’s next product
Neuralink’s first patient plays Mario Kart
And more.
Nvidia's time as a key player in the AI chip market might be coming to an end, as a group formed by Intel, Google, Arm, Qualcomm, and Samsung is planning to create an open-source software suite to democratise access to AI for developers. This approach will allow them to run code on any machine with any chip, diverging from Nvidia's CUDA architecture.
The Unified Acceleration Foundation is still an early project; however, they expect to solidify their efforts by the second half of 2024, showcasing how several companies continue to develop alternatives to challenge Nvidia's dominance and address the current scarcity of AI chips.
🎬 OpenAI releases Sora’s first impressions (1 min)
After reaching out to Hollywood for feedback, OpenAI has presented Sora's first impressions granted by a diverse group of creatives, including visual artists, designers, creative directors, and filmmakers. The company hopes that this feedback can help further improve the model.
These initial responses shed a positive light on Sora's capabilities, as many of them agree that its integration into their workflow allows them to bring impossible ideas to life, eliminating budget and technical constraints.
MIT researchers have discovered a groundbreaking method to simplify the multi-step process of traditional diffusion models. Distribution Matching Distillation (DMD) opens a new pathway for image generation, maintaining quality while accelerating the process and significantly reducing computational time.
This one-step generation process aims to transform diffusion models, aiding in the development of new design tools, improving 3D modelling, and advancing drug discovery, a field where promptness is essential.
😯A patent filing has revealed what appears to be OpenAI's upcoming product. Introduced one day after Sam Altman's interview hinting at new releases, “Voice Engine” might be related to digital voice assistants like Siri and Alexa.
🏎️After the footage of the first Neuralink patient playing chess, another video has been shared demonstrating his progress, showing him playing Mario Kart with his father and controlling the game with his mind.
💼Jobs we liked
Google, Senior Software Engineer, AI/Machine Learning: Warsaw, Mazowieckie, Poland | $56,142 - $75,000
AMD, C++ Machine Learning Engineer, AI Software Solutions: Austin, TX - US | $125,000 - $192,000
Facebook, Software Engineer, Machine Learning: Seattle, WA - US | $131,000 - $285,000
Reddit, Senior Machine Learning Engineer, Ads Prediction: Fully remote | $216,700 - $303,400
⚡️Quick links
What did you think of today's newsletter? |