Continue Reading

Whisper Vs Chirp: The Hidden GPU Cost Of "Free" AI Models & Why Commercial Hosted Models Can Be Far Cheaper

The rapid proliferation of impressively capable "free" open source AI models with one-click installations and simplified workflows means companies increasingly…

Continue Reading

Experiments With Speech Transcription: Comparing The Cost Of Running OpenAI's Whisper Vs Google's Chirp In Production

Over the past few months we've transcribed more than 65 million minutes of global television news coverage spanning more than…

Continue Reading

Experiments With Speech Transcription & Translation: Meta's SeamlessM4T Vs OpenAI's Whisper Vs GCP's STT+GT Vs GCP's USM/Chirp+GT

As the final piece in our our series of evaluating Meta's new SeamlessM4T multimodal translation model, let's evaluate its speech transcription and…

Continue Reading

What OpenAI's Whisper Teaches Us About The Future Of Large Language Models Like ChatGPT

While specialty AI models like OpenAI's Whisper ASR speech recognition and translation target a more technical and niche audience, they…

Continue Reading

Jason Scott: The Whispering Episode

The Internet Archive's Jason Scott mentions on his latest podcast some of GDELT's work on exploring OpenAI's Whisper ASR on…

Continue Reading

Testing The New OpenAI Whisper ASR Large-V2 Model On A Russian TV News Broadcast

This past October we explored applying OpenAI's Whisper ASR to television news broadcast selections from more than 100 channels spanning…

Continue Reading

Experiments With Whisper ASR: No Speedup On V100 GPU From #370

As we continue our ASR experiments with OpenAI's Whisper ASR, we've examined a recent patch that restores proper attention caching…

Continue Reading

Applying OpenAI's Whisper ASR To 101 Sample TV News Broadcasts Spanning 50 Countries

OpenAI's open source Whisper ASR has received immense interest for its multilingual transcription and translation capabilities spanning nearly 100 languages….

Continue Reading

Experiments With Whisper ASR: Model Parameters & Non-Determinism: temperature_increment_on_fallback

Across our experiments with OpenAI's Whisper ASR this week, its unprecedented fluency has been challenged by its high non-determinism, with…

Continue Reading

A Deep Dive Exploration Applying OpenAI's Whisper ASR To A French-Language Télé Congo TV News Broadcast

Thus far this week we have conducted deep dives applying OpenAI's Whisper ASR to Russian and English-language television news broadcasts….

Continue Reading

OpenAI's Whisper ASR: How The "NATO Threat To Putin" Becomes "Putin's Threats [To NATO]": The Challenges Of Machine Translation

Earlier this week we explored how OpenAI's Whisper ASR translated a Russian-language television news broadcast into English. Towards the middle…

Continue Reading

A Deep Dive Exploration Applying OpenAI's Whisper ASR To A PBS NewsHour Broadcast

Yesterday we explored applying OpenAI's open source Whisper ASR to transcribe and translate a Russian television news broadcast, finding that…

Continue Reading

A Deep Dive Exploration Applying OpenAI's Whisper ASR To A Russian Television News Broadcast

Last month OpenAI released an open source ASR system called Whisper, trained on 680,000 hours of multilingual data. Earlier this week…

Continue Reading

OpenAI Whisper ASR + Russian TV News Deep Dive Coming Tomorrow

Tomorrow we will be unveiling a massive deep dive exploring OpenAI's new Whisper ASR system on a complete Russian television…

Continue Reading

Experiments Applying OpenAI's Whisper ASR To Russian Television News

Last month OpenAI released an open source ASR system called Whisper, trained on 680,000 hours of multilingual data. How does…

Continue Reading

WashPost: The Trump Media Era Ends Not With A Wow But A Whisper

The Washington Post's Philip Bump uses GDELT to examine television news mentions of Donald Trump. Read The Full Article.

Continue Reading

How We Transcribed 2.5 Million Hours Of TV News In Just 7 Days & How We Could Have Finished In A Single Afternoon

On Wednesday, we announced in collaboration with the Internet Archive TV News Archive the machine transcription of its complete 2.5…

Continue Reading

Generative AI & A Reminder That "Accuracy" Is Not The Same Thing As "Output"

One of the more remarkable reminders of the Wild West state of Generative AI came in a recent meeting I…

Continue Reading

Visual Explorer: Another 30 Million Minutes Transcribed Through Google's Chirp Transcription Model

Our massive collaborative initiative to transcribe the entire Internet Archive Television News Archive added another 30 million minutes of transcribed broadcasts…

Continue Reading

Visual Explorer: 65M Minutes Transcribed Through Google's Chirp Transcription Model

Our massive collaborative initiative to transcribe the entire Internet Archive Television News Archive has reached 65 million minutes of transcribed broadcasts…

Continue Reading

Generative Image AI & The Potential Of At-Scale Citizen-Led National Promotional Campaigns During Times Of Crisis: A Ukraine Experiment

As we continue to explore the potential of generative AI for autonomous production and ideation of national and organizational branding…

Continue Reading

Multimodal Generative AI Experiments: Gemini Pro + Bard & Describing YouTube Videos

Last month Google announced YouTube integration for Bard in which it can now retrieve transcripts and metadata about any video…

Continue Reading

Presenting Prompt Evolver: Fully Autonomous LLM-Driven Prompt Evolution/Optimization: Crafting The Ultimate Summarization Prompt & The Limits Of LLM Summarizers

Today, the majority of LLM prompts are written and laboriously optimized by human beings through manual prompt engineering. What if…

Continue Reading

Generative AI Experiments: Creativity As Prompt Vs Temperature In Google's Bison, Unicorn & Gemini Pro Models

Historically, the concept of "creativity" was largely embedded in LLMs through a combination of their prompt-level specification and API-level temperature…