Antony Cook

Last year, when developing ContractMatrix, we did so in partnership with one of the leading technology companies in the world: Microsoft. Partnering with Microsoft has enabled us to scale ContractMatrix in Azure, which is a robust and secure environment, and make it available to customers all over the world, via Microsoft AppSource and Azure Marketplace. As I said in a recent interview with Microsoft (https://lnkd.in/eFZJwd9Z), part of the reason we’ve been able to collaborate seamlessly is because both of our organisations share the same vision: to ensure that AI is adopted responsibly in a way that augments humans. ContractMatrix is designed to streamline a specific legal function: contract management. As such, we were able to reduce the risk of hallucinations, by grounding the AI output in established precedents. Risk management was baked into the tool’s design from the beginning; an expert is always in the loop, reviewing any amendments via a tracked changes function. Check out the article linked above or get in touch to find out more.

225

2 Comments

#Google faces new #antitrust trial after ruling declaring search engine a monopoly One month after a judge declared Google's search engine an illegal monopoly, the tech giant faces another antitrust lawsuit that threatens to break up the company, this time over its advertising technology. The #Justice #Department and a coalition of states contend that Google built and maintains a monopoly over the technology that matches online publishers to advertisers. Dominance over the software on both the buy side and the sell side of the transaction enables Google to keep as much as 36 cents on the dollar when it brokers sales between publishers and advertisers, the government contends in court papers. Google says the government's case is based on the internet of yesteryear, when desktop computers ruled and internet users carefully typed precise World Wide Web addresses into URL fields. Advertisers now are more likely to turn to social media companies like TikTok or streaming TV services like Peacock to reach audiences.

8

Lots of interesting observations about how hallucinations/confabulations can be reduced by certain architectures in this article, but here's what I'm wondering about this morning: Suppose the models do end up solving these problems through what is, in effect, a division of labor: Models are specialized to answer queries within a defined subdomain. Do we then develop new models that are good at integrating answers across subdomains into a full solution that represents the best answer across domains? Even if we do, it seems like it will have to be humans that are playing that role in the near term. 1 for those who are good at context recognition and integrating perspectives across different domains — specialists in generalization. https://lnkd.in/g5RMaEx6

14

5 Comments

I’m interested in how AI is impacting law practice. The U.S. Patent and Trademark Office has issued guidance on how to determine inventorship when AI was involved in creating the invention. In 2022, in Thaler v. Vidal, the Federal Circuit held that “only a natural person can be an inventor, so AI cannot be.”  So, how can something developed by AI be patented? ·      The guidance reaffirms the Federal Circuit decision: Named inventors must be “natural persons.” A non-human, such an AI system, cannot be listed as inventor, even if it was “instrumental” in creating the invention. ·      A person named as inventor must make a “significant contribution” to any AI-assisted invention. ·      This is determined by existing inventorship law: The inventor must contribute in some significant way to the invention’s conception or reduction to practice. ·      USPTO gives an example: A person who merely 1) recognizes the need for a new design to solve a specific problem, 2) prompts an AI system to create a design, and 3) reviews the design outputted by the AI system, has not made a significant contribution. ·      On the other hand, a significant contribution “could be shown by the way the person constructs the prompt in view of a specific problem to elicit a particular solution from the AI system.” ·      These distinctions could be difficult to make in practice. It will be interesting to see how this plays out in real world examples as AI develops and its use in creating inventions increases. ·      The USPTO guidance tells us that inventors using AI should be prepared to demonstrate, with supporting documentation, the specific things they did to contribute to the invention and how AI was used to assist. https://lnkd.in/g6gx_eTV #artificialintelligence #technology #AI

6

Any questions about AI for IP? Go on and read this outstanding and comprehensive blog post from my Tradespace colleague Justin Rerko to demystify your fears and misconceptions about this game-changing technology. We are in the innovation business after all - do not be afraid to innovate your practices with AI and ask us how we can help you!

4

Join Director Lestin Kenton at The AI Strategy Summit: IP, Data and Compliance, hosted by CenterForce USA on May 22, 2024, in Palo Alto, CA. Lestin is a speaker on the "Navigating Intellectual Property Challenges in AI-Driven Innovations" panel. Register here: https://lnkd.in/e-tXxeT8 The panel will discuss: 1️⃣ AI as Inventor: Legal Implications and Future Directions 2️⃣ Ownership and Licensing Strategies for AI-Generated IP 3️⃣ Harmonizing Global IP Laws for AI Innovations Don't miss out on this insightful discussion! #AI #IntellectualProperty

8