- Mindstream
- Posts
- Apple’s chip genius joins Rain AI
Apple’s chip genius joins Rain AI
+Meta brings 200 rare languages to life
PRESENTED BY |
NASA is preparing studies on passenger comfort for future air taxis, which could revolutionise urban transportation.
Love or hate this intro? Hit reply and tell us why!
What’s in store:
Rain AI welcomes top Apple chip exec.
Meta comes up with a translation tool for lesser-spoken languages.
Around the web: Massachusetts rideshare drivers gain benefits and $32.50 minimum pay under new Uber and Lyft settlement.
Grow on LinkedIn with Taplio!
Read Time: 6 minutes
AI STARTUP
Rain AI has welcomed Jean-Didier Allegrucci, a former Apple chip executive, to lead its hardware engineering.
This is the second major hire this month for the startup which aims to revolutionise semiconductor design for AI.
Allegrucci, known for his work on Apple processors, that replaced intel in iPhones and Macs, will focus on developing more energy-efficient chips at Rain AI.
He’ll collaborate with Amin Firoozshahian, a former Meta architect.
Here’s what’s happening:
Jean-Didier Allegrucci joins Rain AI to lead hardware engineering.
Focus on developing energy-efficient chips using in-memory compute.
Backed by industry leaders and exploring brain-inspired technology.
Allegrucci’s new challenge
Backed by OpenAI co-founder Sam Altman and Y Combinator, Rain AI is exploring in-memory compute technology.
This method, inspired by the human brain, processes data where it’s sorted, promising significant energy savings.
Despite the challenges around its economic and environmental feasibility, in-memory computing is pursued by industry giants like Intel, TSMC, and Samsung.
Efficient chips are key to broadening AI’s applications across devices.
According to Rain AI CEO William Passo, Rain AI's compute-in-memory technology is set to unlock the potential of modern AI models, making them faster, cheaper, and more advanced.
Turning AI dreams into silicon reality, one chip at a time.
Do you think Rain AI's new hire will revolutionise AI hardware? |
Vote for live results and see results + opinions from yesterday at the bottom of the email.
AI TOOLS
Get Taplio for $1 - Exclusive to Mindstream
Taplio has transformed our LinkedIn game here at Mindstream.
Their AI-powered LinkedIn software is used by over 6,000 of the platform's top creators, including our CEO Adam, who has nearly 90,000 followers in one year on the platform.
Taplio helps us create top posts and carousels quickly with advanced scheduling beyond LinkedIn's features.
Mindstream has teamed up with Taplio and our readers can get their first month of Taplio for just $1.
Here’s how:
Start the usual sign-up process by clicking on "Start for free" on the Taplio home page.
When you reach the payment page during user onboarding, enter the code MINDSTREAM.
This is valid for the first 15 readers that sign up!
Don’t miss out on this exclusive Mindstream offer!
Use code: MINDSTREAM
On This Day
July 1st 1908 - SOS is adopted as the international distress signal.
Trending Tools
ImageFX: Developed by Google, this tool uses AI to generate high-quality images from text prompts.
Yarn: Yarn is a similar tool that finds clips from TV shows, movies, or songs based on user-inputted words or phrases.
HeyGen: A cloud-based platform that transforms text into professional-quality videos.
Saner AI: Designed to enhance productivity by acting as your AI-powered 'second brain,' helping organise notes and ideas efficiently.
RunwayML: A powerful tool for artists and designers, RunwayML uses machine learning to generate original and realistic videos from text prompts.
META
Meta is enhancing its translation capabilities on Facebook and Instagram to include 200 lesser-spoken languages worldwide.
The No Language Left Behind (NLLB) project aims to make these languages more accessible, but experts suggest consulting native speakers to refine the tool.
Meta’s NLLB project recently announced they’ve scaled their technology to support these languages.
The project includes European languages like Scottish Gaelic, Galician, Irish, Ligurian, Bosnian, Icelandic, and Welsh.
Meta classifies these as "low resource" languages, with less than one million sentences available for AI training.
Experts recommend Meta involve native speakers and language specialists to improve the service.
The current tool still needs work to achieve better accuracy.
How it works:
Meta trains its AI using data from the Opus repository, an open-source platform with a collection of authentic text.
Contributors are experts in natural language processing (NLP), enabling computers to translate and understand human language.
Meta also uses mined data from sources like Wikipedia to create a multilingual language model. This AI can translate between languages without relying on English.
200 languages, one goal
William Lamb, a professor of Gaelic ethnology and linguistics at the University of Edinburgh, points out that Meta’s Scottish Gaelic translations are still not very good.
He suggests that Meta should consult native speakers to improve the tool.
However, most native speakers are older and do not use computers, making this challenging.
Lamb recommends Meta collaborate with the BBC, which produces high-quality online content in Gaelic.
Meta's translations have improved by 44% since the first model was released in 2020.
When fully implemented, there will be more than 25 billion translations daily on Facebook and Instagram.
The NLLB team uses a benchmark of human-translated sentences to evaluate translation quality, filtering out toxic words and phrases.
Alberto Bugarín-Diz, a professor of AI at the University of Santiago de Compostela in Spain, believes linguists and tech experts should work together to refine datasets.
He notes that using Wikipedia data can be advantageous, as it covers almost every aspect of human life, but stresses the need for legal compliance in data usage.
Lamb remains sceptical about recommending Meta's tool due to current inaccuracies.
Bugarín-Diz, however, argues that user engagement is essential for Meta to invest in improvements, recognising that understanding the tool's weaknesses is key before widespread use.
Watch me make this my whole personality in 2024.
Mindstream Picks
China faces growing pressure to reconsider its strict crypto mining ban, balancing its carbon neutrality goals with economic impact and technological innovation.
Wall Street expects robust earnings growth from major tech names like Microsoft, Alphabet, and Amazon.
Eumetsat's decision to shift the MTG-S1 weather satellite launch from Ariane 6 to Falcon 9 in 2025 surprised and frustrated European space officials.
Gaimin Gladiators and Team Spirit secured spots in the grand finals with 2-0 playoff victories, as the 1win Series Dota 2 Summer crowned its champion last Sunday, with the broadcast on Kick and YouTube.
Don’t Miss - Massachusetts rideshare drivers to get benefits and $32.50 minimum hourly pay from August 15th following a settlement with Uber and Lyft. The companies will also pay $175 million to underpaid drivers and provide health stipends and sick leave.
AI Art
Our Image of the Day
Artwork submitted by Mindstream reader Maddox: “An abstract painting of the moon”
Daily Image Prompt
A futuristic looking Gandalf the Grey
Yesterday’s Poll
“Do you think Adept's collaboration with Amazon will finally bring their AI agents to market success?”
Yes, Amazon will make it happen! - 65% ✅
No, it’s just a temporary lifeline. - 35%
Your Views:
“Amazon knows how to deliver outcomes and innovation is in its DNA. This will succeed.” -davidsbrent
“Would love an Agent to do manage AWS admin for me.” - jim
Submit your opinions in our daily poll to be featured!
Starting fresh this week with more news and the latest in the industry. ❤️ We need your feedback to make our newsletter better. 📣 Refer our newsletter to your friends and make us feel all warm and fuzzy. 🚀 Advertise in our newsletter to reach 150,000+ founders, engineers, and content creators. |
You There! What Did You Think? |