The Promise & Perils of AI: What We’ve Learned from Some of the Largest Real World Deployments of Generative & Classical AI to Understand Global News

Speaker: Dr. Kalev Hannes Leetaru (GDELT Project)

Date: 11 July 2024

Speaker Session Summary

SMA hosted a speaker session with Dr. Kalev Hannes Leetaru (GDELT Project) as part of its SMA General Speaker Series. 

One of the GDELT Project’s goals is to use artificial intelligence (AI) to better understand our world. To accomplish this monumental task, the GDELT Project maps over 400 languages and cultures in real-time, looking at local events through local lenses to discover the conversations taking place across the globe. The ability to look at events and interpret them in their local languages, rather than translating them into English or Western languages first, is particularly vital, as important context and nuances are often lost during translation. The GDELT Project has thus far used generative AI tools and large language, speech, and multimodal models (LLMs, LSMs, and LMMs) to ingest vast amounts of text, video, and audio files and interpret world events, including the COVID-19 pandemic, Russia’s invasion of Ukraine, and the war in Gaza.

There are numerous benefits and pitfalls involving the use of generative AI and LLMs. The ability for an individual to collect millions of terabytes of visual and textual media and make sense of it is one of AI’s key values to society, as is AI’s ability to understand, interpret, and seamlessly transition between local languages. AI’s facial recognition and facial embedding capabilities are also tremendously valuable, as they enable individuals to interpret key trends occurring in a society, including why certain individuals regularly appear on certain news networks. One can also use AI to determine how stories evolve over time in terms of tone, the places and people focused on, the way events are described, and the velocity and frequency of the stories. One of the primary downsides of AI, however, is its inherent biases, which makes trusting the tool’s description and interpretation of events, in particular, very difficult. The fact that these models have been trained on largely Western data contributes to these biases. AI also has difficulty with summarizing and determining the importance of certain events amidst a plethora of less important events. Moreover, AI hallucinates, plagiarizes, misunderstands, becomes distracted, and regularly excels in its ability to lie, even to the point where a human editor may not be able to identify all of the tool’s lies and inaccuracies. Overall, it may look on the surface like AI is rapidly advancing, but in reality, individuals still need to overcome monumental hurdles in order to generate the perfect insights that we see online.

It is important to remember that generative AI and LLMs can only summarize what has been done in the past. It does not possess the ability to reason or think. What we will most likely see in the future is a blend of statistical, classical, and generative techniques, rather than just generative techniques due to their inherent flaws.

Speaker Session Recording

Briefing Materials

Dr. Kalev Hannes Leetaru, one of Foreign Policy Magazine’s Top 100 Global Thinkers of 2013, is a global advisor to governments, NGOs and the world’s largest corporations to help them solve tomorrow’s greatest challenges in an ever more uncertain world. His GDELT Project fundamentally transformed modern global risk forecasting, becoming one of the most iconic and largest real􀆟me open graphs over Planet Earth. For more than a quarter-century his landmark studies have been at the forefront of reimagining how we understand our world through some of largest datasets and computing platforms on the planet.