The AI Rundown by Lightscape Partners – 04/15/24

Good morning, happy Monday, and welcome back to this week’s AI Rundown.

This week in AI:

• Taiwan Semiconductor Manufacturing (TSMC) is the latest semiconductor company to receive US money from the CHIPS Act with an $11B grant to build 3 new factories in Arizona.

• Two researchers at OpenAI have been let go following accusations of leaking company information.

• The newest generative AI startup combines AI-generated music and social media, backed by former Black Eyed Peas member will.i.am and a16z.

Following another trend of the weekend, the word of the week is GPU:

Check out the AI conferences happening this month here.

If you’d like a quick refresher on industry terms and the current landscape of AI, skip to the bottom.

Tops AI Stories of the Week

US and Taiwanese chipmaker TSMC reach deal to build 3 new semiconductor factories in Arizona. Link.

• The 3 new factories will create over 25,000 jobs in the United States.

• TSMC will invest $65B in building the three factories and receive $11B in grants from the US.

• The deal highlights the United State’s push in the ongoing chip war, aiming to produce 20% of the world’s semiconductors by 2030.

• This is the fifth CHIPS Act deal.

OpenAI fired two employees for releasing company information. Link.

• One of the researchers fired was a close ally of chief scientist Ilya Sutskever, who led the movement forcing Sam Altman out as CEO last year.

• One of the researchers fired may have been Leopold Aschenbrenner, a researcher on the team dedicated to keeping AI safe for society.

• It is unclear what information the two were caught leaking at the moment.

Udio, the newest AI-generated music app launches with prominent backings. Link.

• The music and social media company received $10M from a16z, will.i.am, and others.

• The company uses AI to create music that can be shared publicly within the platform.

Foundational Models

Mistral AI debuts a new 176B parameter LLM: Mixtral 8×22B. Link.

• The new model is expected to outperform all of Mistral’s previous models.

• The new model has a 65,000-token (~48,000 words) context window, the amount of text a model can process and reference at one time.

• Mistral’s open-source model play has directly contrasted OpenAI’s closed model as the two foundational models continue to one-up each other.

Hardware + Infrastructure

Rumors about Apple’s newest M4 chip heavily revolve around AI capabilities. Link.

• Despite the recent launch of the M3 chip and the upcoming debut of the M3 ultra chip, Apple is rumored to be releasing an AI-focused M4 chip as soon as the end of the year.

• The M4 will have more AI computing cores and would likely be able to run AI inference on the edge.

Intel announces newest AI chip Gaudi 3. Link.

• This is the newest version of Intel’s artificial intelligence chip, challenging Nvidia’s lead in the semiconductor industry.

• The new processor will be available for mass distribution in Q3 2024.

• The chip will boost performance around training AI systems and running inference, the two main components in the AI lifetime stack.

Meta announced a powerful new custom AI chip for training and inference. Link.

• Meta announced their family of custom-made chips for Meta’s AI workloads called Meta Training and Inference Accelerator (MTIA).

• The latest version will help power Meta’s ranking and recommendation ads models.

• MTIA illustrates Meta’s growing focus on AI infrastructure and the growing focus on building custom silicon.

Upcoming AI Conferences

AI Accelerator Institute

• San Jose, CA

• April 16 – 17, 2024

Get up to speed on the current landscape

• AI Term Glossary

• Current AI Landscape from Sequoia Capital

Thank you for reading the AI Rundown by Lightscape Partners. Please send any questions, comments, or suggestions to [email protected].