The AI Daily Brief
Posts
AI Audio Generation from Text

AI Audio Generation from Text

Plus Alibaba releases new models to compete with Meta

Nathaniel Whittemore
August 03, 2023

The AI Breakdown First Five - Thursday August 3, 2023

Today on the First Five:

5. Alibaba Releases New Models To Compete w/ Llama 2
4. An Nvidia AI Chip Challenger Lands $100M
3. Goldman Says $200B to Be Invested in AI
2. Google Search Generative Experience Gets Updates
1. Meta’s Text-to-Audio Generator is Live

5. Alibaba Releases New Models To Compete w/ Llama 2

Recently, Alibaba Cloud surprised many when it announced it would support Meta’s Llama 2 model for customers. This week, it announced it’s own “open source” models Qwen-7b and Qwen-7b-Chat which have some commercial restrictions similar to Llama 2.

Huge...
“Today, Alibaba Cloud is the first to launch a training and deployment solution for the entire series of Llama 2 in China. All developers are welcome to create customized large models on Alibaba Cloud."
— Paul Triolo (@pstAsiatech)
12:13 PM • Aug 2, 2023

4. An Nvidia AI Chip Challenger Lands $100M

Another day, another contender trying to disrupt Nvidia’s lead in the AI chip space (or at least represent some amount of market competition). Tenstorrent is a Canadian startup that has added a fresh $100m from Hyundai, Samsung and others to its war chest.

The life of a startup is filled with highs and lows.
This is one of the highs.
@tenstorrent has closed a note on a higher valuation & brought in 2 strategic investors.
Thank you @Hyu@Hyundai_Globalia@Kia_Worldwideam@Samsung all of our investors for your vote of confidence. http
— David Bennett (@DavidBennett__)
12:16 AM • Aug 3, 2023

3. Goldman Says $200B to Be Invested in AI

By 2025, Goldman Sachs thinks that AI investment could represent up to 4% of US GDP. GS says this would make it bigger than electricity or personal computing in terms of economic impact. Hyperbole or representative of the true opportunity?

.@GoldmanSachs predicts investment in AI could soar to nearly $200 billion by 2025 and could eventually have a bigger impact on U.S. GDP than electricity and PCs.
— Cointelegraph (@Cointelegraph)
2:10 AM • Aug 3, 2023

2. Google Search Generative Experience Gets Updates

If you were wondering about the future of Google Search, it’s clearly what they currently call their Search Generative Experience (which is right up there with some of OpenAI’s finer duds in the pantheon of bad product names). Announced this week, SGE will now increasingly return visual results — particularly video - to user queries.

Google Search Generative Experience adds videos to AI-generated answers searchengineland.com/google-search-…
— Barry Schwartz (@rustybrick)
4:01 PM • Aug 2, 2023

1. Meta’s Text-to-Audio Generator is Live

Our suspicion is that, culturally-speaking, music and sound are going to be one of the key AI battlegrounds. Meta has released AudioCraft which combines three different audio generation models. This week they released the pretrained models as well as the weights and code.

Today we're sharing details on AudioCraft, a new family of generative AI models built for generating high-quality, realistic audio & music from text. AudioCraft is a single code base that works for music, sound, compression & generation — all in the same place.
More details ⬇️
— Meta AI (@MetaAI)
4:31 PM • Aug 2, 2023

Meta open sourcing all their AI projects is the greatest whitepill of 2023
ai.meta.com/blog/audiocraf…
— Garry Tan 陈嘉兴 🛡️— e/acc (@garrytan)
4:24 PM • Aug 2, 2023