230 | 🍓🤯 OpenAI & Google blow up the week!

In today’s Brainyacts:

1. Latest OpenAI model

2. Google just killed education, thought leadership, and podcasts

3. Meta doing shady things and other AI model news

4. Do GenAI summaries violate antitrust? and more news you can use

   👋 to all subscribers!

To read previous editions, click here.

Lead Memo

🍓🍓 OpenAI released new models

OpenAI Launches New Models: Thomson Reuters Had It First

OpenAI has unveiled its latest innovation, o1 (previously known as Strawberry), a revolutionary new model that takes a significant step forward in AI capabilities. This model doesn’t just respond instantly but pauses to think, carefully evaluating different ways of responding before delivering an answer. The technology behind o1 uses advanced reinforcement learning and chain-of-thought processing, enabling it to approach problems like humans do—by thinking before speaking.

What Makes o1 Different?

The key feature that sets o1 apart is its ability to “think” through problems, particularly those that require deep analysis and logical reasoning. Whether tackling intricate coding challenges or answering science-based questions at a PhD level, this model has proven to outperform both previous models and even human experts in specific domains. Here’s how o1 stacks up:

• Superior Performance: o1 surpasses expert humans on complex, PhD-level science questions and ranks in the top 11% for competitive programming.

• Mathematics Mastery: The model solved an impressive 83% of problems from the International Mathematics Olympiad qualifying exams, a stark contrast to GPT-4o’s 13% success rate.

• Tailored for Math, Science, and Coding: o1 is particularly adept at handling challenging queries in these areas, making it a game-changer for professionals who need precise answers.

Two versions are available to users: o1-preview and o1-mini, both of which are now available to ChatGPT Premium and Teams users, giving them direct access to the enhanced capabilities of this next-gen model.

Early Feedback: Thomson Reuters Sets the Stage

Thomson Reuters was one of the first organizations to get hands-on experience with OpenAI’s o1 model through their legal unit, CoCounsel. The early results have been promising, especially for tasks that require strict adherence to instructions and detailed analysis of specific documents.

In a statement to Axios, Jake Heller, the head of product at CoCounsel, remarked on the model’s precision:

“Its careful attention to detail and thorough thinking enables it to do a few tasks correctly where we have seen every other model so far fail.”

One of the notable trade-offs, Heller explained, is that responses may take longer as the model pauses to consider its options. However, this delay is seen as a positive in high-stakes environments like law and finance, where accuracy is paramount. As Heller put it:

“Professionals want the most thorough, detailed, and accurate answer—and they would much rather wait for it than get something wrong and quick.”

The Road Ahead for AI

OpenAI’s o1 model signals a new era for artificial intelligence, where the goal is not just speed but depth and precision. By mimicking human thought processes and focusing on delivering the most accurate response, o1 has the potential to reshape the way industries handle complex tasks, from legal analysis to scientific research.

Spotlight

🤯 👂 Google!!!

You know how we’re always juggling knowledge, documents and information? Trying to get the people to engage and get the most out of them? Well, Google’s Notebook LM just blew my mind. If you haven’t checked it out yet, seriously, stop what you’re doing and sign up for it on labs.google. It's like having Gemini 1.5 work directly on your own docs, turning them into a personal knowledge hub.

But here’s where it gets crazy—they just added an audio feature. At first, I thought it was a bit gimmicky, but it’s not. Yes, this version turns any content into a podcast with two AI voices. But stop to think about what is happening when you listen and watch - think of all the content we create to communicate with clients, colleagues, bosses, students, etc. The world is saturated with boring and ignored information.

For you law firm people - just think about all the info on your firm’s intranet. Usually these are glorious collections of super boring and ignored content. But I digress.

In the video below, I tested it on an ABA Ethics Opinion. Yes I did! And in less than a minute, I had a super engaging podcast version of a dry legal PDF.

Here’s why I think you’ll find this fascinating:

• Instantly converts content into a podcast—no setup needed.

• Engaging dialogue between AI voices makes even dense topics easy to digest.

• Think about what this could do for sharing information with clients, colleagues, or students.

• It’s not just a fun feature—it has serious potential for how we communicate and share knowledge.

• It process up to 50 sources, each up to 500,000 words, allowing for a total of 25 million words to be considered when generating the audio.

And here is the straight audio of the whole podcast.

AI Model Notables

► Facebook and Instagram are making AI labels less prominent on edited content.

► Meta/Facebook admits it scraped all Australian Facebook posts since 2007 to train its AI.

► Microsoft Copilot to be integrated into Singapore's legal technology platform.

► Google introduced DataGemma, to connect LLMs with real-world data from Google's Data Commons, aiming to reduce AI hallucinations by grounding responses in factual stats.

► CLARIFICATION: In Tuesday’s edition is was too hasty in writing my headlines at the beginning of the newsletter. I shared data that ChatGPT usage was dropping off significantly. This was based on a report that hits to ChatGPT’s website had been declining. Going to the website and using the tool are tow different things.

THANK YOU to the many many readers that emailed me or hit me up on socials to let me know about this. Love when readers engage!

So, to clear things up: OpenAI's ChatGPT reportedly surpassed 11 million paying subscribers, including 1 million on higher-priced business plans, potentially generating over $2.7 billion in annual revenue, according to COO Brad Lightcap.

News You Can Use:

➭ A group of Democratic leaders—led by Senator Amy Klobuchar—has written a letter to the US Federal Trade Commission (FTC) and Department of Justice (DOJ), urging them to investigate whether AI tools that summarize and regurgitate online content (like news stories) are violating antitrust laws.

➭ Tech, US officials discuss AI development, climate and power needs at White House.

➭ An underwater data center in San Francisco Bay? Regulators say not so fast.

➭ In Nevada Google’s AI will help decide whether unemployed workers get benefits.

➭ New AI service to help lawyers go through 15,000 judgments in Singapore’s legal history.

➭ A recent study finds that AI has the potential to change individuals' beliefs in conspiracy theories, challenging the conventional wisdom that evidence and arguments rarely sway those entrenched in such ideas.

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.