Unlock the full potential of AI with Building LLMs for Production—our 470+ page guide to mastering LLMs with practical projects and expert insights!


AI Trends on TED
Artificial Intelligence   Data Analysis   Data Science   Latest   Machine Learning

AI Trends on TED

Last Updated on July 9, 2024 by Editorial Team

Author(s): GaryGeo

Originally published on Towards AI.

AI Trends on TED
AI TED Talk Trends — Source: Author

Introduction: The AI Zeitgeist Through the TED Lens

If you are like me, you turn to TED videos to satisfy your curiosity or to be educated on innovative ideas. In recent years, I have regularly watched their Artificial Intelligence videos to learn more about AI’s capabilities, potential, and risks. Almost every week, I notice a new TED AI video on my YouTube homepage, which inspired me to do some digging. There are over 550 AI-themed TED videos dating back to 2007!

The dataset, interactive app and YouTube playlist at the end of the article.

As I started to explore TED’s rich library of content, it dawned on me that these conversations are a window into how AI technology and adoption is evolving. With this in mind I started my most extensive data analysis project to date, to give some structure and track the TED trends. I used the YouTube videos as my source, a little LLM, and a lot of Python to build a knowledge graph and started analyzing. This article is about the analysis, not the graph, but follow me for a future article about the build.

Ray Kurzweil explaining the potential of AI back in 2007 at a TED Event — Source: Screenshot from YouTube by Author

The Evolution of AI on TED

My first step in the analysis was to better understand the video publishing trends over time. Using this dataset as a foundation I started to investigate what was the “story” behind these trends.

Video Count by Year, by Channel — Source: Author

Early Days: Visionaries and Pioneers (2007–2015)

In the mid-2000s, when AI was still a niche topic for most, TED was already featuring talks by visionaries like Ray Kurzweil, known for his work on “The Singularity,” and Jeff Hawkins of Palm Computing fame. IBM’s Jeopardy playing Watson, launching in 2011, was the biggest AI story of the time. During this period, AI discussions were sporadic, appearing occasionally but not consistently every year.

It is also notable that TEDx events were already in the thousands by 2012 (source: Forbes), so either the content was not focused on AI, or these videos were not published on YouTube (or are now archived)

The Tipping Point (2016–2017)

Based on the dataset, a shift began in 2016–2017, marked by an increase in AI coverage. DeepMind’s AlphaGo was mastering Go, not by memorization, but by creating its own new strategies that beat the Go masters, such as its victory over world champion Lee Sedol in 2016.

At the same time TEDx events were spreading (with over 100,000 talks by 2017 — source: TED Blog) and the topic of AI intrigued many of the new community-based presenters.

The Deep Learning Boom (2017–2019)

The increase in AI-related TED talks during 2017–2019 resulted from several factors converging at once. This period saw advances in deep learning and neural networks research and at the same time, companies/venture capitalists increased their investments in AI startups. Data Science became a popular career choice, and Big Data was a hot topic. AI technologies also reached the top of Gartner’s Hype Cycle for Emerging Technologies, reflecting high public interest and high expectations.

These factors — tech progress, more funding, growing expertise, and public excitement — led to more AI discussions at TED talks. People were seeing how AI would impact different aspects of society and industry. TED became a forum for exploring this AI shift as it happened.

The Pandemic Interlude (2020–2021)

During 2020–2021, much of the focus on TED’s main channel shifted to healthcare, remote work, and the social impacts of the COVID-19 pandemic. AI was not the main topic but was an undercurrent in the discussions about technological solutions to pandemic-related challenges.

The ChatGPT Era (Late 2022-Present)

ChatGPT-3’s release in late 2022 sparked renewed interest in AI, especially in Large Language Models (LLMs). Throughout 2023 and 2024, AI and LLMs have taken center stage at TED. Presenters have covered a wide range of topics, from the technology’s capabilities and opportunities to its societal impacts and potential risks.

A timeline with events and trends in AI that are reflected in the TED Talks — Source: Author

And to no one’s surprise, TED is not alone. A snapshot from Google Trends shows the impact of AI on search is even more dramatic. Interest in AI experienced a parabolic shift and is only now stabilizing at levels 10x what they were before ChatGPT.

Google Trends for “Artificial Intelligence” searches — Source: Author

The volume and publishing cadence of the videos tell part of the story, now let’s see what we can extrapolate from the videos themselves.

What Can We Learn from the Video Data

Next we will dig into the content of this collection of YouTube transcripts and metadata. My analysis involved extracting key concepts (topics, people, organizations, etc.) as well as categorizing the videos to build a knowledge graph. With this we can learn about the categories, people and organizations that dominate the TED AI video and also provide insights into the general zeitgeist of Artificial Intelligence.

A word-cloud created from the 50 most frequently extracted concepts in the TedTalk YouTube dataset — Source: Author

Key Categories

AI is a general-purpose technology, much like electricity or the Internet, with the potential to achieve significant results across a wide range of applications. This diversity is reflected in the categories and topics in the video dataset. AI has the potential to impact various areas of life, including business, society, healthcare, education, work, art, entertainment, and more. Alongside these emerging applications, we also see videos addressing a broad set of concerns, including ethics, governance, safety, and security.

In terms of distribution, the TED catalog is actually very balanced across these two extremes. Applying AI to business and industry is a major focus of the TED catalog, with 126 videos dedicated to this category. However, this focus is balanced by a significant number of videos addressing societal impacts (113) and AI ethics and governance (99). The pattern continues with substantial categories focused on healthcare (63) and education (55), balanced by concerns about the future of work (36). As we move into the smaller categories, this pattern of balance persists. Overall, about 55% of the videos primarily focus on opportunity topics, and 45% focus on more risk-related topics.

A pie chart representing the percentage of videos by category (one category per video) — Source: Author

The fact that opportunities and risks weigh evenly in TED presentations mirrors the dilemma we face as a society — what will it cost us to embrace the potential of AI?

Influential People

Now let’s move on to what can be learned about AI by examining the individuals mentioned in these TED videos. Key individuals frequently mentioned in the videos fall into three categories:

  1. Technical Thought Leaders: Known for their pioneering contributions and thought leadership in AI (e.g., Alan Turing, Stephen Hawking, Ray Kurzweil, Marvin Minsky).
  2. Business Leaders: Visionaries in the business world who have significantly influenced the adoption and application of AI/Technology (e.g., Elon Musk, Bill Gates, Mark Zuckerberg, Steve Jobs).
  3. Expert Reference Points: Masters in their fields who have been profoundly impacted by AI advancements (e.g., Garry Kasparov in chess, Lee Sedol in Go, Michelangelo in art).
A horizontal bar chart with 18 of the most mentioned individuals by video count — Source: Author

While many of these names are well-known, there were a few that I had to research, with the larger list feeling almost like a “who’s who” in AI quiz.

More so than the abstract trends and concepts, understanding the individuals in AI helps to give a broader context to what we see in the video library. This AI moment is historical, and these individuals will be an important part of that history.

Leading Organizations

Organizations also play an important role, and while I don’t think the list of most referenced organizations will surprise anyone, it does highlight key shifts over the 17 years of TED videos.

  • Google is mentioned almost twice as often as the next organization, even considering their DeepMind acquisition as a separate entity.
  • OpenAI has rapidly gained prominence, despite being a relative newcomer.
  • MIT and Stanford are the leading academic institutions for AI research and development.
  • IBM, Amazon, and Meta have been minimally referenced in this latest LLM wave, and over 80% of their mentions happened before 2022.
A cumulative stacked area chart reflect the video count that organizations accumulated over the years — Source: Author

Organizations have much more inertia than individuals, and I think we will continue to see Google, Microsoft, MIT, Amazon, etc., for many more years. That is not to say there will not be upstarts like OpenAI , but it is far more likely their star will fade or they get consumed (e.g. DeepMind’s acquisition by Google). For this trend, our 17 year window might not be enough.


These TED Talks serve as a window into the AI revolution, reflecting its journey from a niche subject to a transformative force in our society. This analysis leverages video content to provide insight into AI technology trends, societal opportunities and risks, and the individuals and companies driving its emergence.

As AI continues to evolve, TED videos will remain valuable resources for understanding its potential, challenges, and the critical conversations surrounding its development and implementation. While these individual presenters and videos are incredibly powerful on their own, analyzing them in aggregate reveals trends that enhance our broader understanding.

The story of AI is still in its early chapters, and it will be fascinating to see how these trends evolve and what new topics emerge in this dynamic field.

Resources to Explore Further

Screenshot of my TED AI Video Exploration App — Source: Author

Data Call-Outs

The goal of the dataset is to be directional and insightful. I believe it achieves this, but alas it is not perfect.

  1. The video playlist contains videos published through May 2024. As a result many of the charts have full year data for other years and partial for 2024.
  2. This playlist was manually generated, by myself. I may have made errors or applied judgment on what to include inconsistently… I did my best.
  3. There are other TED Videos published that are not on the YouTube channel, so this playlist and dataset is incomplete.
  4. The playlist includes all of the TED channels in this analysis. By including all of these channels, we can get a broader cross-section of what people are interested in sharing and discussing. The main TED channel features videos from the official TED conference and TEDx videos that have been promoted to the main channel. TEDx actually has many more videos, as it comes from numerous community-organized events. There is also a TED-Ed channel, which focuses on educational content. Lastly a seemingly inactive TED Institute channel that was more corporate-focused.
  5. The extractions and category assignments were done with OpenAI ChatGPT-4o. There can be inconsistencies and errors.
  6. While not a focus of this analysis, the YouTube stats (Views, Likes, Comment, etc) were updated at the beginning of July 2024. There is an inconsistency with YouTube metrics in that a video published months or years before another video has had more time to accumulate, the Views, Likes, and Comments. Since the last video was added at the end of May, there was at least a one-month period for the video statistics to accumulate


Below is the general methodology I used in conducting this analysis. I plan to do a separate article on the process and techniques in the future (follow me to learn more).

  1. Identified YouTube and established a playlist of relevant videos thru ~June 1, 2024
  2. Used APIs and Python to gather both YouTube metadata and transcripts.
  3. Processed the data in a Python notebook, including transcript summarization, concept extraction, and categorization. This was done with the OpenAI API (i.e. LLMs).
  4. The results were stored in a knowledge graph comprising over 3,500 nodes and 11,000 relationships.
  5. Manually reviewed the captured nodes and relationships to remove issues/errors, and merge similar concepts (Stanford vs Stanford University, etc).
  6. Created datasets useful for analysis (e.g., video count by year/channel, video count by person, etc) then created visualizations.
  7. As a side effort I loaded this knowledge graph data into a JSON file for the web app.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI

Feedback ↓