AI & Startups April 8- April 14

苹果 Macs getting AI overhaul

Apple is reportedly gearing up to overhaul its entire lineup of Macs with an upcoming M4 chip family — which will place a heavy emphasis on AI capabilities.

The details:

• The M4 chips are nearing production and will come in at least three main varieties: entry-level Donan, high-performance Brava, and top-end Hidra.
• Apple plans to update every Mac model with M4 chips, with releases starting in late 2024 and extending into early 2025.
• Initial M4 refreshes will include new iMacs, MacBook Pros, and Mac Minis, followed by MacBook Airs, Mac Studio, and Mac Pro.
• More details and features are expected at the company’s developer conference on June 10.

Udio levels up music creation

A new AI-powered music creation app from former Google DeepMind researchers called Udio just launched with the backing of prominent tech and music industry figures.

The details:

• Udio allows users to generate full audio tracks in under 40 seconds with simple text prompts from descriptions like genre, lyrics, or inspirational artists.
• The app offers powerful editing and capabilities, as well as a variety of vocal languages, music types, and remixing options.
• Udio has secured $10M in seed funding from investors including a16z, Instagram co-founder Mike Kreiger, and musicians will.i.am and Common.
• The company is also working with artists directly to explore new ways of creating and monetizing music using AI.

Meta unveils powerful new custom AI chip

Meta just released the next generation of its custom Meta Training and Inference Accelerator (MTIA) AI chip family, delivering significant improvements in compute performance and efficiency.

The details:

• The next-gen MTIA chips perform 3x better than last year’s MTIA v1 across four key model evaluations.
• The new chip is currently serving Meta’s recommendation and ranking models, with plans to expand support to GenAI models like Lllama.
• The company isn’t alone in the custom silicon world —?with rivals like Google, Amazon, and Microsoft all surging ahead with their own chips.

Mistral AI releases massive new open-source LLM

French AI startup Mistral just released Mixtral 8×22B, a powerful new frontier LLM, dropped quietly via a 281GB file on X available for download.

The details:

• Mixtral 8×22B features a 65,000-token context window and 176B parameters and is expected to surpass the previous Mixtral in capabilities.
• The model uses a sparse mixture of experts (SMoE) approach, optimizing performance and cost by combining specialized models for tasks.
• Mistral joins a wave of companies releasing or teasing model updates this week, including OpenAI, Google, and Meta.

Meta confirms Llama 3 coming within the month

At an event in London, Meta confirmed that it plans an initial release of Llama 3, the next generation of its large language model, within the next month.

The details:

• Meta will release several different versions of Llama 3 with different capabilities over the year — with the initial release within the next month.
• The company did not disclose the size of the parameters used in Llama 3, but it's expected to have about 140 billion parameters and rival OpenAI’s GPT-4.
• Meta previously released Llama 2 in three sizes, with the largest at 70B parameters.
• Meta has stockpiled 350,000 coveted H100 GPUs over the last year to ramp up AI infrastructure, a supply that dwarfs other competitors.

AI chip wars heat up with 英特尔 Gaudi 3

Intel just unveiled its Gaudi 3 AI chip at the company’s Vision event, taking direct aim at Nvidia's dominance and offering enterprises an open, flexible alternative for deploying generative AI at scale.

The details:

• Gaudi 3 promises 50% faster training and inference performance than Nvidia's H100 chip while delivering 40% more power efficiency at a lower cost.
• The chips will be available in Q2 of 2024 and are expected to be on par with Nvidia’s H200 line.
• Intel also revealed an expanded AI roadmap that includes an open enterprise AI platform, new Xeon processors, and more.

谷歌 , 英伟达 team up to accelerate AI development for startups of all sizes

As global startups continue to navigate the challenges of building AI products, Google is pushing the bar to give them everything they need in one place. Case in point: the company’s latest collaboration with Nvidia aimed at bringing the companies’ startup accelerators together.

The details:

• Under the partnership, announced at the ongoing Google Cloud Next conference in Las Vegas, Google said it is combining its Google for Startups Cloud Program with Nvidia’s Inception initiative. The move will give startups access to everything they need – from cloud credits to technical resources and go-to-market support – to build AI applications and services and deliver value to their customers faster.
• For AI-first startups in particular, the company has been providing up to $350,000 in cloud cost coverage, along with dedicated AI training and access to its AI ecosystem, including infrastructure, AI products like Vertex and foundation models.

Gen AI VC

IAG Capital Partners Leads $24M in Generative AI Startup for Drug Discovery, DeepCure

AI SaaS cloud startup Neysa bags $20 million from Matrix, Nexus, others

Sapien raises $5M to ease companies' AI journeys