- Starter AI
- Posts
- Stability’s Stable Cascade, Amazon’s largest model, Nvidia worth more than Amazon
Stability’s Stable Cascade, Amazon’s largest model, Nvidia worth more than Amazon
Guess who’s one of the most valuable in the world?
Sthefania
February 15, 2024
Hello, Starters!
As we dive deep into our AI journey, you'll start to recognize which major players are always on the verge of sharing new advancements. And if you haven't yet, let us tell you all about them.
Here’s what you’ll find today:
Stability AI introduces Stable Cascade
Largest-ever text-to-speech model emerges at Amazon
Nvidia surpasses Amazon and Alphabet in value
GPT-5 is “better at everything” according to Sam Altman
ElevenLabs expands its AI voice changer
And more.
🖼️ Stability AI introduces Stable Cascade (1 min)
Stability AI has introduced what seems to be a great successor to its remarkable Stable Diffusion. Stable Cascade showcases upgraded performance in prompt alignment and aesthetic quality, offering users an array of options such as improving resolution, inpainting, outpainting, and creating a new image just from the edges of an existing one. All of this is achieved through simple text prompts.
An interesting thing to note is that Stability has employed three LLMs based on the Würstchen architecture, well-known for its efficiency and time-saving capabilities. This allows Stable Cascade to require less memory, fewer training hours, and provide results faster than ever before.
A team of researchers at Amazon has recently developed the largest text-to-speech model to date, dubbed BASE TTS (Big Adaptive Streamable TTS with Emergent abilities). What sets this model apart, besides its groundbreaking 980 million parameters, is the fact that it showcases "emergent" abilities, allowing it to speak complex sentences naturally.
BASE TTS was trained using 100,000 hours of public domain speech, mostly in English, but also including a bit of German, Dutch, and Spanish. As models grow beyond a certain size, they seem to become more robust and perform more versatile tasks, which is one of the aims of Amazon's AGI team with this breakthrough.
Nvidia's impact on the AI landscape is undeniable, as the firm is the main provider of the chips used to train most of the LLMs currently existing. It's no surprise, then, that it has claimed its place as the world's fourth most valuable company, surpassing Amazon and Google's owner, Alphabet.
In addition to advancing AI technologies, many big tech companies are aiming to challenge Nvidia's dominance by developing their own GPU chips. This includes companies like Intel and AMD, as well as OpenAI, which, interestingly, is also one of Nvidia's main clients.
🤖Sam Altman shared insights into GPT-5 directly from the World Government Summit. In the words of the CEO, the model is "magical" and "a little bit better at everything," hinting at its performance and multimodal capabilities. Although there's no release date yet, many predict its arrival in the spring of 2024.
🗣️ElevenLabs has expanded its powerful "Speech to Speech" AI Voice Changer to support 29 languages. This enables users worldwide to leverage AI to enhance their videos, games, and projects while maintaining full control of the overall delivery of their character's voice.
⚡️Quick links
What did you think of today's newsletter? |