The Deep Media Digital Digest

VOLUME 7 / SEPTEMBER 13TH, 2024

Welcome to Volume 7 of the Deep Media Digital Digest! This week, the AI world is buzzing with groundbreaking releases that are set to redefine how we think about and interact with technology. OpenAI’s much-anticipated ‘Strawberry’ model has finally rolled out, bringing with it unparalleled reasoning capabilities and a host of new features. Meanwhile, Google and Apple are making strides in integrating AI into everyday tools, marking significant steps forward in how AI will be used by consumers. These developments are just the beginning of what promises to be an exciting period for AI.

What's On Our Radar

This week, the release of OpenAI’s ‘Strawberry’ model has taken the spotlight. This model is a significant leap forward in AI, boasting the ability to solve complex math problems it has never encountered before and autonomously perform deep research across the internet. It also has the power to code entire video games from a single prompt and can "think" before answering, improving its reasoning with more processing time. Available to all Plus and Team users on ChatGPT, this model is poised to change how we approach problem-solving and creative tasks.

On the productivity front, Google’s NotebookLM has introduced an “Audio Overview” feature that allows users to have their notes read out loud, making it easier to digest complex topics. Whether it’s summarizing legal briefs or course readings, NotebookLM’s new feature caters to those who prefer auditory learning, marking another step in AI's integration into our daily lives.

Apple is also making waves with its upcoming AI features for the iPhone 16. Set to be available in a beta test next month, these features include writing tools like text rewriting, summarizing, and proofreading, as well as a revamped Siri, smart replies, and transcription services. Though they won’t be available at the launch of iOS 18, these capabilities will soon be enhancing user experience on iPhones, Macs, and iPads with M1 or higher chips.

What's Happening Now at Deep Media

At Deep Media, we're continuing our critical work in setting benchmarks for the rapidly expanding Deepfake Detection space. This week, we released a GenAI voice benchmark dataset with over 200,000 voice samples, encompassing 16 different generative capabilities and 8,000 unique speaker identities. This dataset is designed to aid researchers in improving the accuracy and reliability of deepfake detection.

We’re also gearing up for our upcoming webinar on September 24th, "Benchmarking the Battle Against Deepfakes: Strengthening Your Security Strategy," where our CEO, Rijul Gupta, will discuss the importance of security governance in the age of AI. Register here to secure your spot.

We've noticed that the rapid growth of Generative AI technologies is transforming software development but also brings significant risks. As organizations rush to adopt these tools, they often overlook the security implications. GenAI tools, if not properly governed, can introduce vulnerabilities into software products, potentially leading to data breaches or the spread of misinformation. Our latest blog post on security governance emphasizes the need for clear internal guidelines, robust media monitoring, and integrated multimodal deepfake detection to mitigate these risks.

Deepfakes Dominating the Headlines

The challenges surrounding AI-generated content continue to grow, with Grok-2 making headlines yet again. Concerns are mounting over its lack of guardrails, particularly in the context of the upcoming elections. Reports from NPR highlight the dangers posed by Grok-2’s ability to create fake political scenarios, which could influence public opinion and election outcomes.

This week, Taylor Swift spoke out about her fears regarding AI after a deepfake incident where Donald Trump posted fake images claiming her endorsement. In endorsing Kamala Harris for president, Swift expressed her concerns about the misuse of AI, emphasizing the impact these technologies can have on public figures and political discourse.

In a significant move, several major AI vendors, including Adobe, Cohere, Microsoft, Anthropic, and OpenAI, have committed to taking steps to combat nonconsensual deepfakes and child sexual abuse material. According to TechCrunch, these companies have agreed to responsibly source and safeguard the datasets they use to train AI, reflecting the increasing pressure on tech giants to address the ethical challenges posed by AI-generated content.

~ The Deep Media Team