The AI & Legal Update #5 - 🔮 Big news from OpenAI

The AI & Legal Update #5 - 🔮 Big news from OpenAI

Cecilia Ziniti

May 14, 2024

Hi, friends. Holy moly, some say we've achieved AGI (artificial general intelligence, aka the singularity, aka the robots are here). Have we? Read on to find out. 
Read time: 6 minutes 11 seconds

OpenAI because ... they launched big stuff

OpenAI launched GPT-4o (see blog). Updates:

🎼 The o stands for "omni" because it's "omnichannel" because it can take in audio, video, or images directly and respond in kind, versus translating to text each way. 

 🇮🇹 GPT can now be much more Alexa-like and translate conversations live, per esempio in Italiano, doubly endearing your editor. 

🎁 They are giving limited access to GPT-4o for free and launched a desktop app. If you are currently using the free model (GPT-3.5) you will notice a very big difference in speed and power. Compared to GPT-4, it's faster and supposedly smarter... more on that below.

©️ With the new model, it seems OpenAI has regressed on content protections, as it happily gives copyrighted and trademarked art that it used to refuse. 

⚖️ So ....  the big question ... does GPT-4o still need a legal-focused AI product like GC AI to be amazing or is off-the-shelf ChatGPT good enough? Here's a side-by-side of GC AI and ChatGPT both with GPT-4o analyzing decacorn Notion's MSA for a few commercial basics. We are bullish on legal-tuned AI!  

Tap the images to see the original chats. PS - Notion legal team, good work on your readable MSA. 

How do we do it? The answer ... software and a whole lot of legal care. We have tuned GC AI to provide answers you can trust and build things under the hood with features like Exact Quote (™️ pending - do you think we can get it?).

Other Big Tech AI News 

💀 Apple and OpenAI are close to a deal, per Bloomberg, to integrate ChatGPT on the next version of iOS, coming this fall. Interesting but unsurprising Apple feels it can’t catch up here. Siri once reminded your editor to “die alone at 1:59 pm”, when she asked for reminder to “dial in” at that time. 

🛠️ Microsoft is building its own AI model, code name MAI-1. Leading the charge is Mustafa Suleyman, who lead DeepMind at Google. Wonder what his employment agreement said! 

📧 Interesting 2019 email to Bill Gates and CEO Satya Nadella warning Google was too far ahead on AI, so Microsoft needed to invest in OpenAI. Smart move.  

IP

🎬 Top Hollywood agency Creative Artists Agency (CAA) has started creating AI twins of actors that CAA will license on the actors’ behalf. Interestingly, studios typically require artists to create digital avatars when they start shooting a film - but licenses to those are limited for use for that film. PS - your editor’s amazing sister is a CAA agent.  

⌨️ An AI patent reviewing bot from a former Amazon patent counsel scanned 300,000 US patents and found 90% had errors like typos and missing words.  

🤑Want Reddit data for training? Sure. Just pay them. CEO Steve Huffman unveiled Reddit’s new content policy and says they’re “going to stay open, but to crawl Reddit or have access to reading content, you need to have some sort of agreement." 

🤝 AdAge got a leaked copy of OpenAI’s partnership pitch deck.  It details the company’s “Preferred Publishers Program,” under which OpenAI gives incentives in return for collaboration, aka permission to use their content. The Financial Times took the deal. Meanwhile, the copyright cases on AI pile up. See map below.

Regulatory and Policy 

🏄🏻‍♀️DHS announced 22 members of the Artificial Intelligence Safety and Security Board - members include Sam Altman, Sundar Pichai, Lisa Su (CEO of AMD), and the mayor of Seattle.  

🚓 A software called Cybercheck helps law enforcement “search[] all corners of the internet, whether surface or darkweb, regarding a suspect, victim or person of interest.”  Its creators claim it's over 90% accurate, but EFF is challenging it, and judges are throwing out Cybercheck evidence. See also Cybercheck's EULA

🏛️ The Office of Personal Management released guidance on federal employees' use of AI. ⚠️ Senator Amy Klobuchar warns about AI. She’s introducing bills that would ban deceptive deepfakes and increase transparency around AI-generated political content.

✝️ Pope Francis will participate in talks on AI at the G7 next month. Italian PM Giorgia Meloni announced it, emphasizing the need for an ethical framework to ensure AI remains "human-centered and human-controlled."  

🇪🇺 A new NBER study finds that following GDPR's implementation, EU firms significantly reduced their data storage by 26% and data processing by 15% compared to comparable US firms. In that time US GDP outpaced Europe’s, which even contracted (source). Win? 

🏔️Colorado's legislature passed AB 205 which if the governor signs would make Colorado the first state with a law specifically targeting high-risk AI uses. It focuses on preventing algorithmic discrimination and codifies the NIST AI Risk Management Framework and ISO/IEC 42001. 

Corporate & AI Adoption

⚜️ Thanks to AI, we’re in the golden age of freelancing. We agree. Fractional GCs have told us GC AI is their secret superpower!   

👨‍⚕️ Nurses gather at Kaiser San Francisco to protest AI in health care - they say "no computer, no AI can replace a human touch," which, true.  

📈 Linkedin, a subsidiary of Microsoft which also owns made a strategic investment in OpenAI, reports 75% of the knowledge workers they polled use AI at work, and most BYOAI. 

Other findings match what we see at GC AI - the key is trying AI, thinking to use it, and finding great use cases. Our favorite excerpt: 

“The path to becoming a power user starts with developing new habits. Power users are 68% more likely to frequently experiment with different ways of using AI — [ ] the #1 predictor of whether someone will be a power user … [power users are] more likely to frequently pause before a task and ask themselves if AI can help (+49%), to keep trying if they don’t get the perfect response the first time (+30%), and to research and try new prompts (+56%).” 

Are you trying AI? If knowledge is stopping you, take one of GC AI's Maven AI prompting for lawyers class. No judgement, just fun. Really! PS -  running a class now (Tuesday May 14 at 10:30 PT), and will let you all know when we run the next one.  

AI & Law 

🙂 Your editor's tweet about AI and law went viral. Tl;dr - in-house is where AI works best, while law firms are thinking about how to adopt AI. It's a great time to be a lawyer!  

🫠 Big firm Latham's head of litigation is optimistic, saying that "Artificial intelligence is great for litigators as a source of cases" and she "doesn’t think the technology has reached a level where she could rely on it." Maybe she could try GC AI. In seriousness, offer is out there if this email makes its way to her, Michelle, we'll enroll you in our Maven class for free.  

🤝 Docusign is buying CLM Lexion for $165m to beef up its AI capabilities. Congrats WGCN GC Jessica Nguyen.  

🎰 The CLOC legal operations conference was last week in Vegas (welcome back!). The Dean of Suffolk Law headlined, and explained that AI will change the practice of law.  

GC AI Product Updates & Quote of the Week

🔮 Faster, smarter AI

GPT-4o is live in GC AI. You can now use OpenAI's newest and most cutting-edge model which was just released yesterday. It's twice as fast and you'll also find it reasons better and can handle more complex questions and tasks. 

📄 Upload up to 300-page PDFs

We've doubled the page limit for PDFs so you can upload PDFs up to 300 pages. 

🫳 Easier File Upload

Drag and drop files (PDFs, *.docx files, images) from your computer to attach them to chat messages. See it in action below.

🆕 New Official Prompts 

MSA commercial basics - best for reviewing inbound paper - check for invoices (no credit cards), payment terms (defaults to net-45, must be nice lol), no autornew, term, effect of termination. Pro tip: add your own item to check! See analysis of Notion's MSA.  

DPA checklist prompt - run a DPA against a comprehensive checklist. See Okta's here

❤️ User Love & Use Case of the Week

🦸 "GC AI has been a fabulous tool. While I could sit and craft explanations for hours to various questions from business stakeholders, it's so much easier to provide context to GC AI and have the tool draft it for the appropriate audience -- and for Slack, email, text...it also prevents me from typing something up angrily as I explain negligence to the same person for the nth time in one week. :)" - GC Software with transactions specialty  

🪪 "I'm really enjoying figuring out what GC AI can do. Yesterday I asked it to scan all of my organization's public open-source projects on Github and produce a table for me with name of project and type of open source license associated with it (and to note if there was no license at all). It produced an accurate table in a matter of seconds. I love the prompt library -- it's useful on its own and also as a reference for seeing ways to adapt and create my own prompts. Keep 'em coming." - GC, internet infrastructure provider

Random

Reid Hoffman interviewed his AI twin.  Wild. 

🐞 Your editor and her daughter enjoy getting illustrated stories out of voice with ChatGPT. We did a series about a ladybug Amelia. The kiddo asked for progressive pictures and marveled at the voice responses which are more natural than Alexa which she's used to. "Now make dad" with some explanation by me yielded the below. Can you spot the funnies? How about the trademark violation? 

Fun times! Thank you all for your support. 
Best,

Cecilia and the GC AI Team

{{text2Gradient}}