178 | 🥷 📣 From DeepFakes to DeepTakes

In today’s Brainyacts:

1. From DeepFakes to DeepTakes

2. DeepFake panel announcement

3. GPT4.5 and GPT5 just got leaked and other AI model news

4. Judges defend lawyers using AI and other AI-related content

👋 to new subscribers!

To read previous editions, click here.

Lead Memo

🪙 📣 I am coining a new term the “DeepTake”

This is OpenAI’s CTO, Mira Murati. The Wall Street Journal posted a recent interview with Mira specifically about OpenAI’s video-generating model Sora. Sora can produce amazing, high-quality, and realistic videos (aka a really great deepfake generator). But there have been serious questions raised as to what videos or content it was trained on.

Here is her answer:

Now, I am a huge fan of Mira, Sora, and OpenAI but we are skirting way too many big and looming legal issues like IP protection and evidentiary implications. And some of these companies are way too cavalier about it. And it is happening constantly which is why we need a name for it.

What is DeepTake?

DeepTake refers to the practice of extracting vast amounts of data and content, often without proper authorization or attribution, to train artificial intelligence (AI) models. This concept is especially relevant in the context of deep learning, where AI systems require large datasets to learn and generate realistic outputs, such as in the creation of deepfakes. DeepTake highlights a growing concern in the digital age: the unauthorized use of data, ranging from images and videos to text and personal information, to fuel the advancement of AI technologies.

Why is it Called DeepTake?

1. Connection with Deep Learning: The term "Deep" in DeepTake is derived from deep learning, a key technology behind many AI advancements, including deepfakes. Deep learning algorithms depend on extensive data to function effectively, and DeepTake encapsulates the problematic side of this dependence—acquiring content in ways that may infringe on privacy and intellectual property rights.

2. Depth of Impact: "Deep" also signifies the profound effects that acquiring this content can have on individuals, businesses, and society. The implications of DeepTake are deep-rooted, affecting ethical, legal, and economic domains. It goes beyond simple theft, challenging our understanding of ownership, consent, and ethical AI use.

3. Association with Deepfakes: Naming it DeepTake creates an intuitive link with "deepfakes," emphasizing the intertwined nature of content acquisition and the creation of synthetic media. Both DeepFakes and DeepTakes raise significant concerns about authenticity, truth, and the potential for misuse in the digital realm.

4. Complexity and Nuance: The inclusion of "Deep" reflects the complex, layered challenges that DeepTakes present. It's not just about the act of taking content but also about navigating the intricate web of legal, ethical, and technological issues that arise from such practices.

Spotlight

Florida Bar Tackles Evidentiary Impact of AI Deepfakes

Happy to share that I have been invited to speak as part of the President’s Showcase portion of the Florida Bar’s Annual Convention.

In preparing for that, the organizers and I considered adding a specific focus on Deepfakes and evidence procedures/implications. It got approved and so in June, we will be discussing this live in front of 100+ Florida Bar members.

I am amped! As you know, I have been running down this path for many months now trying to flag all the warning signs of an evidentiary apocalypse. Deepfakes have been with us for a long time. Authenticating evidence can be challenging already. But in the AI Era that courts and all of us are now in, the old rules and ways of authenticating evidence are likely to be shattered. We must get ahead of this now as there is zero doubt a deepfake has already been accepted into evidence somewhere.

Have a take on this issue you want to share? Or, know of a resource or related effort underway?

Know of someone who might make an excellent and informed panelist?

** Please reach out to me by replying directly to the email that sent you this newsletter - it comes directly to me and me only.

AI Model Notables

► OpenAI's GPT-4.5 Turbo leaked on search engines and could launch in June

► Did Microsoft just tease GPT-5 Turbo in Copilot Pro subscription?

► Microsoft quietly upgraded Copilot's free version to GPT-4 Turbo. Here's why it matters

► Microsoft singles out Google's competitive edge in generative AI to EU antitrust regulators

► Anthropic releases Claude 3 Haiku, an AI model built for speed and affordability

► Meet Devin, the first AI software engineer who can build websites from a prompt

► Apple buys Canadian AI startup as it races to add features

News You Can Use:

➭ Judges defend AI use by attorneys for early brief drafts

➭ ‘AI-powered’ ad ignites creator controversy on Instagram

Here is the ad

➭ US spearheads first UN resolution on artificial intelligence — aimed at ensuring equal access

➭ Morgan Stanley names a head of artificial intelligence as Wall Street leans into AI

➭ Waymo pushes forward despite robotaxi setbacks

➭ Breakthrough AI can now predict Alzheimer's up to 7 years in advance

 ➭ AI assisted death therapy: Empathy closes $47M for AI to help with the practical and emotional bereavement process

Who is the author, Josh Kubicki?

Some of you know me. Others do not. Here is a short intro. I am a lawyer, entrepreneur, and teacher. I have transformed legal practices and built multi-million dollar businesses. Not a theorist, I am an applied researcher and former Chief Strategy Officer, recognized by Fast Company and Bloomberg Law for my unique work. Through this newsletter, I offer you pragmatic insights into leveraging AI to inform and improve your daily life in legal services.