AI can now read your mind

AI’s Biggest Lawsuit Yet

Last Wednesday, the New York Times (NYT) sued OpenAI and Microsoft, alleging copyright infringement. Many are calling this AI’s Napster moment—poised to test the legal boundaries of generative AI.

The allegations:

• The NYT claims unauthorized use of its articles by OpenAI for training chatbots like ChatGPT and Copilot.

• According to the NYT, these chatbots generate output that replicates articles word-for-word—which competes with its paywalled content.

• The overall argument? NYT says this development threatens high-quality journalism, damages the relationship between news outlets and readers, and costs the NYT subscription and licensing revenue.

The demands: The NYT is asking the court to…

• Prevent OpenAI and Microsoft from training their LLMs on its content. 

• Remove NYT articles from the companies’ datasets.

• Demand “billions of dollars in statutory and actual damages.”

Some context: The suit comes just after OpenAI inked deals with Axel Springer (the German media titan behind Politico and Business Insider) and the Associated Press for permission to train AI models on their news content. But the NYT isn’t alone in its opposition to OpenAI’s practices—it’s one of many publications that have recently blocked OpenAI’s web crawler.

My take? Of all of the lawsuits against OpenAI, this one holds the most weight: The fact that ChatGPT rewrites paywalled articles verbatim is a strong argument. But I still think OpenAI will come out on top. Here’s why:

1. The NYT’s prompt in the lawsuit’s example seems…a little fishy. It looks like the company tailored a prompt to get the chatbot to output articles word-for-word. Most users don’t formulate prompts that way.

2. I think the NYT’s goal isn’t about upheaving ChatGPT’s training method; it’s about receiving financial compensation for the use of its content. 

The NYT has already been in talks with OpenAI about the usage of its articles, so it’s possible that this will settle out of court.  But if the case does move to court, its outcome could shape the future of AI training methods—potentially forcing LLM developers to strike expensive content deals with major publications.

Instant Answers to Any Question Across Notion

Today’s AI chatbots search the whole internet to find answers instantly. But what happens when you have a question about your company’s private information?

Notion’s new AI feature, Q&A, is an expert on everything you’ve ever written in Notion. Q&A functions as your assistant, sweeping through your workspace in seconds to answer any question you may have.

Q&A can be used for static information retrieval—like quickly identifying the sponsor of an upcoming video—or dynamic information retrieval—like instantly checking on the status of tasks. Plus, Notion AI protects data privacy and only references pages that users have access to.

If you're already a Notion AI user, you can get started today with Q&A already in your Notion workspace. If you’re interested in trying it out, join the waitlist to get early access.

AI That Reads Minds

Mind-reading sounds like the stuff of sci-fi, but AI is making it a reality. Australian researchers developed a new AI system called DeWave that can translate silent thoughts from brain waves into text. 🤯

TL:DR:

• DeWave utilizes an EEG cap—small electrodes that are placed on a user’s scalp—that records neural activity and eye movements.

• The tool then matches EEG patterns with eye fixations that indicate the recognition of specific words—and uses this information to output full thoughts.

A caveat: The results are still far from perfect—DeWave achieved over 40% accuracy in early experiments. However, the researchers are confident that the system can reach 90% accuracy with further work, which would put DeWave’s capabilities on par with speech recognition technology.

What’s next: Sure, mind-reading computers can seem a bit creepy. But DeWave could revolutionize communication for people who are unable to speak or type—presenting a whole new frontier of AI in assistive technology. This is just the start of the neuroscience x AI revolution, and I can’t wait to see what’s next.

Apple’s Secret LLM

Last week, Apple took us all by surprise when news broke that the company had silently launched its own multimodal LLM, Ferret, on GitHub—way back in October.

What makes Ferret special:

• Ferret goes beyond object recognition. By combining natural language understanding (NLP) with image analysis, it can understand image elements at various levels of detail.

• For example, when asked to identify an object within a picture, Ferret can analyze surrounding elements for a deeper contextual understanding.

• The LLM can simultaneously detect and understand both text and visual elements, responding to prompts with a deeper level of insight.

Open-source strategy:  Breaking from Apple’s (in)famous love of secrecy, Ferret was released as an open-source model. The move invites developers to explore and enhance Ferret’s capabilities. Ferret is currently limited to research purposes only.

Looking ahead: Ferret’s release hints at an AI-powered future for Apple’s hardware products. If integrated into Siri, Ferret could enable advanced visual search functionalities within iPhones (like requesting specific actions based on an image). The model could also improve accessibility for visually impaired users.

• Pika Labs’ new AI video generator is now available to all users. 

• AI-powered search engine Perplexity announces a $73.6M funding round—and investors include Nvidia and Jeff Besos.

• AI discovered a new class of antibiotics to fight drug-resistant bacteria. 

• AI startup Mistral plans to release a GPT-4 level open-source model this year. 

• OpenAI is raising fresh capital at a stunning valuation north of $100 billion. 

• Google breaks ground in AI video generation with the launch of VideoPOET.

More important AI news: Dive deeper into this week’s hottest AI news stories (because yes, there are even more) in my latest YouTube video:

DIY: I just found an exciting tool to run your very own chatbot locally on your computer. I show you how in this new video:

Setting the frame: This video gives a primer on where AI could be headed this year by diving into the current state of AI.