• The AI Pulse
  • Posts
  • šŸ¤– Groq Unveils Lighting-Fast LLM Engine

šŸ¤– Groq Unveils Lighting-Fast LLM Engine

PLUS: OpenAIā€™s Startup Fund Backs AI Healthcare Venture, Google DeepMindā€™s New JEST Method

Welcome back AI enthusiasts!

In todayā€™s AI Report:

  • šŸ’„Groq Unveils Lighting-Fast LLM Engine

  • šŸ’µOpenAIā€™s Startup Fund Backs AI Healthcare Venture

  • šŸ’ØGoogle DeepMindā€™s New JEST Method

  • šŸ› 5 Trending Tools

  • šŸ’°Venture Capital Updates

  • šŸ’¼Whoā€™s Hiring?

Read Time: 3 minutes

šŸ—žRECENT NEWS

GROQ

šŸ’„Groq Unveils Lighting-Fast LLM Engine

Image Source: Canva AI Image Generator

Groq unveiled a lightning-fast Large Language Model (LLM) inference engine that responds to user queries instantly.

Key Details:
  • Groq, a hardware and software company that delivers exceptional compute speed, added a specialized inference engine to the companyā€™s website that enables LLMs to supply outputs instantly.

  • Groq identified that current processors, such as Graphics Processing Units (GPUs), canā€™t handle the speed and demand required to scale conversational chatbots like OpenAIā€™s ChatGPT.

  • The company developed a Language Processing Unit (LPU) thatā€™s specifically tailored for the efficient processing of language models.

  • Groqā€™s LPU achieves blazing speeds of 500 tokens per second compared to OpenAIā€™s GPT-3.5 Turbo of 30 tokens per second.

  • Tokens are the smallest units of data used by an LLM to process and generate text. Similarly, we break down sentences into words or characters.

Why Itā€™s Important:
  • While GPUs are powerful processors that excel at handling the complex calculations required for AI model training, Groqā€™s LPU is designed to help trained AI models use new data to make predictions or solve tasks.

  • Groqā€™s LPU is a potential competitor to Nvidiaā€™s GPUs. Its specialized architecture boasts superior processing speeds for tasks related to LLMs.

šŸšØTry Groqā€™s LPU for free here using Metaā€™s Llama 3.

šŸ©ŗ PULSE CHECK

Can Groqā€™s LPU dethrone Nvidiaā€™s GPUs for LLM tasks?

Vote Below to View Live Results

Login or Subscribe to participate in polls.

OPENAI

šŸ’µOpenAIā€™s Startup Fund Backs AI Healthcare Venture

Image Source: Mirko Ries/ā€œWorld Economic Forum (WEF)ā€/Flickr

OpenAI CEO Sam Altman and Huffington Post founder Arianna Huffington are funding a new venture called Thrive AI Health.

Key Details:
  • Thrive AI Health will create an app to focus on habit-forming and behavior change that promotes a healthier lifestyle.

  • The ā€œAI Health Coachā€ will offer precise recommendations tailored to each person through real-time, health-related ā€œnudgesā€ and suggestions.

  • For example, it recommends swapping your third soda with water and lemon or starting your wind-down routine at 10:00 PM for optimal sleep.

  • ā€œYes, behavior change is hard. But through hyperpersonalization, itā€™s also something that AI is uniquely positioned to solve,ā€ said Huffington.

  • DeCarlos Love, former Fitness and Health Experiences Lead at Google, has been appointed CEO of Thrive AI Health.

Why Itā€™s Important:
  • Over 129 million Americans suffer from at least one major chronic disease, and 90% of the U.S.ā€™s $4.1 trillion in annual healthcare spending goes toward treating these physical and mental health conditions.

  • Thrive AI Health can use AI and real-time data to suggest highly curated and proactive daily health recommendations to patients.

AI RESEARCH

šŸ’ØGoogle DeepMindā€™s New JEST Method

Google DeepMind researchers introduced the Joint Example Selection Technique (JEST), a method that accelerates AI model training with fewer computing requirements.

JEST leverages two AI models to streamline the data selection process for AI model training:

  1. Curator: a smaller AI model that acts as a ā€œcurator,ā€ evaluating the quality of different data batches.

  2. Trainer: a larger AI model that ā€œtrainsā€ the highest-quality data batches identified by the smaller AI model.

The method relies on ā€œdata quality bootstrapping,ā€ which refers to using small, curated datasets to guide learning on larger, unstructured datasets.

JEST makes AI model training up to 13X faster and 10X more efficient than current state-of-the-art techniques.

JEST achieved top-tier performance while only using 10% of the training data required by previous leading AI models.

Overall, JEST is a significant step forward in AI model training efficiency. Focusing on selecting the highest-quality data batches allows for faster AI model training with lower energy consumption.

šŸ› TRENDING TOOLS

šŸ•øHubflo builds client portals to streamline operations.

šŸ–DishGen generates recipe ideas with your ingredients.

šŸ’­Bubble is the full-stack, no-code app builder for everyone.

šŸ“‘CompResearch creates competitor research reports to grow your business.

šŸŒ‚Mindpal builds your AI workforce to automate market research and sales proposals.

šŸ”®Browse our always Up-To-Date AI Tools Database.

šŸ’°VENTURE CAPITAL UPDATES

  • Hebbia nets a $130M Series B to build the go-to AI platform for knowledge retrieval.

  • Rackhouse Ventures secures a $45M Second Fund to invest in early-stage AI in overlooked niches.

  • Leading AI primary care platform K Health raises a $50M Equity Fund to expand clinical AI within leading healthcare systems.

šŸ’¼WHOā€™S HIRING?

  • Qualcomm (San Diego, CA): ML/AI Engineering Intern, Fall 2024

  • AMD (Austin, TX): Graduate Data Science Co-Op/Intern, Fall 2024

  • DRW (Chicago, IL): Software Developer Intern, Summer 2025

  • BlackRock (Americas): 2025 Summer Internship Program, AMERS

  • Weights & Biases (San Francisco, CA): Senior Software Engineer, GenAI Team

šŸ¤–PROMPT OF THE DAY

MANAGING TEAMS

šŸ’ŖContinuous Improvement

Imagine being tasked with motivating your [Team] in [Business] with [Product/Service] in a [Niche Market]. Provide five practical steps to foster a culture of continuous improvement among your team.

Consider how you would encourage feedback, identify improvement opportunities, and sustain momentum over time.

Team = [Insert Here]

Business = [Insert Here]

Product/Service = [Insert Here]

Niche Market = [Insert Here]

šŸ“’FINAL NOTE

If you found this useful, follow us on Twitter or provide honest feedback below. It helps us improve our content.

How was todayā€™s newsletter?

ā¤ļøTAIP Review of the Day

ā€œTodayā€™s newsletter covering liability issues with Waymoā€™s self-driving car service was informative.ā€

-Pete (ā­ļøā­ļøā­ļøā­ļøā­ļøNailed it!)
REFER & EARN

šŸŽ‰Your Friends Learn, You Earn!

You currently have 0 referrals, only 1 away from receiving āš™ļøUltimate Prompt Engineering Guide.

Refer 5 friends to enter šŸŽ°Julyā€™s $200 Gift Card Giveaway.

Reply

or to participate.