Welcome (back) to AI Minds, a newsletter about the brainy and sometimes zany world of AI, brought to you by the Deepgram editorial team.

In this edition:

• 📚 The entire, 502-page AI Index Report

• 🧪 Industry continues to dominate academia in AI research

• 💽 Will AI models run out of data?

• 💾 Why Training on Synthetic Data leads to a drop in output quality

• 💸 Private investment in generative AI skyrockets

• 🤖 China versus United States: The AI Robotics Battle

• 💻 NVIDIA Reaches $1 trillion valuation

• 🏇 How the AI Business really works: defensibility, stocks, startups

• 🌏 Global opinions on AI Products and People have changed

• 🐦 Social Media’s hot takes on the AI Index

• 📲 Three new trending AI Apps to check out!

• 🎤 AI Minds Podcast: Matt Hammel, COO & Co-founder of AirOps

• 📉 The hidden problem with today’s LLMs

• 🐶 Which LLM is the best at fetching information for you? (RAG & Information Retrieval)

• 🚑 The future of AI Healthcare

• 🤣 Can AI Make you Laugh? See how far LLMs have come in the past year

Thanks for letting us crash your inbox; let’s party. 🎉

🔍 The 2024 AI Index Report

Stanford just released its seventh annual AI Index Report, which showcases data related to AI in an effort to “provide unbiased, rigorously vetted, broadly sourced [information] for policymakers, researchers, executives, journalists, and the general public to develop a more thorough and nuanced understanding of the complex field of AI.”

Click here to see the report for yourself!

This edition of AI Minds will cover some of the 502 pages worth of AI content found in this report (among other pieces of content).

🧑‍🔬 Industry continues to dominate frontier AI Research

Epoch AI maintains a database of AI and machine learning models released since the 1950s, selecting 1.3 Frontier AI Research entries based on criteria such as state-of-the-art advancements, historical significance, or high citation rates. Analyzing these models provides a comprehensive overview of the machine learning landscape’s evolution, both in recent years and over the past few decades.

Highlight: Will Models Run Out of Data? (Chapter 1; Page 26)

As illustrated in the graphic above, when the number of synthetic generations increases, the tails of the distributions vanish, and the generation density shifts toward the mean. This pattern means that over time, the generations of models trained predominantly on synthetic data become less varied and are not as widely distributed.

Self-Consuming Generative Models Go MAD - A study from Rice University and Stanford University found that generative image models trained solely on synthetic data cycles—or with insufficient real human data—experience a significant drop in output quality.

🏇 Economic Impact: Generative AI Investment Skyrockets and China Dominates Industrial Robotics

AI Index Chapter 4, Pages 30-32: Despite a decline in overall AI private investment last year, funding for generative AI surged, nearly octupling from 2022 to reach $25.2 billion. Major players in the generative AI space, including OpenAI, Anthropic, Hugging Face, and Inflection, reported substantial fundraising rounds.

NVIDIA Reaches $1 trillion valuation: Nvidia’s market capitalization consistently exceeds $1 trillion USD, driven by rising demand for its AI-powering chips. Nvidia becomes the fifth company to reach a valuation of $1 trillion, joining the ranks of Apple, Alphabet, Microsoft, and Amazon.

China vs. United States (Chapter 4: Page 82-83) - The image above illustrates the number of industrial robots installed across various sectors in China over the past three years. In 2022, the leading sectors for industrial robot installations in China were electrical/electronics (100,000), automotive (73,000), and metal and machinery (31,000). To see how the United States compares, click the link!

🏦 How the AI Business Really Works

To see how the AI Business really works and what causes all of the numbers above to fluctuate in the way that they do, check out this directory! You’ll learn about everything from AI Defensibility, to the big tech vs. startup battle, and even how good AI is at predicting the stock market.

🌍 Global Opinions on AI Products and People

Ipsos ran a survey on global attitudes toward AI products and services. The survey consisted of interviews with 22,816 adults ages 16 to 74 in 31 countries. From feelings about AI to Trust in AI, this report reveals how the public’s feelings evolve as machine learning models improve over time.

🐝 Social media buzz

🧭 Trending AI Apps to Check Out!

SendFame allows you to create AI generated celebrity video messages, from happy birthday wishes to business presentations. Currently, you can create videos featuring Elon Musk, Donald Trump, Joe Biden, Christiano Ronaldo, Taylor Swift, Kanye West, Drake, and more.

Tavrn leverages state-of-the-art AI technology to automate and streamline tasks that traditionally require manual effort and extensive time. For ad agencies, it acts as a comprehensive project management bot, reviewing timelines, communicating with team members, and ensuring seamless progress towards task completion. 

CheggMate emerges as a leading brand in the realm of educational platforms, designed to cater to the diverse needs of students across various academic disciplines. By integrating advanced technologies like ChatGPT, GPT-3.5/4, and resources from Chegg, Course Hero, StuDocu, Quizlet, and Tutor.com, CheggMate provides an all-encompassing solution for academic assistance. 

🎙️ AI Minds Podcast! 

In this episode of AiMinds, Demetrios chats with Matt Hammel, COO, and Co-Founder of AirOps!

They discuss Matt's transition from consulting to startup life and the journey of AirOps in making AI workflow building accessible. Discussions include the importance of making complex AI concepts user-friendly, the potential and challenges of an AI-centric business model, and the concept of forward compatibility in technology. 

The episode wraps up with a look into the evolution of AirOps from solving data challenges to leveraging AI models for customized business applications, highlighting the power of artificial intelligence in the business sphere.

🤖 Additional Bits and Bytes

• 🎥The Hidden Problem with Today’s LLMs (Video) - In this lightning talk given at MLOps World 2023, Jose Nicholas Francisco discusses how the way we train LLMs and the data we choose to train them impacts their quality… causing an unforeseen problem.

• 🐶 Information Retrieval: Which LLM is best at looking things up? - With RAG becoming trendier and LLMs taking the spotlight in the AI community, the question arises: Which LLM is the best at retrieving information for you?

• 🚑️ The Future of Healthcare: Enhancing Patient Experience with Voice-Enabled AI Copilots - Many people feel much more comfortable speaking with a chatbot rather than typing at one. This article explores the future of voice-enabled healthcare AI. We'll walk through its advantages over text-only chatbots and show how it could vastly improve the efficiency of hospitals all over the world.

• 😂 Can AI make you laugh? (Video) - If you want to see how far AI art and LLMs’ writing has come over the past year, check out this video! The AI art of February 2023 is much, much different than what you see today. Warning: Strong language.