Collaborative Intelligence Week in Review - 31May2024

May 31, 2024

The One Generative AI Blindspot Leaders Have Now
This video is Part One of a two-part interview of AI experts and business/technical leaders that took place at the recent MIT Sloan CIO Symposium.

Alchemist: Parametric Control of Material Properties with Diffusion Models
A team of researchers from Google Research and MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) describe a method for controlling (changing) specific parameters in photos. The link above demonstrates the effects. The technical paper can be accessed by clicking this link.
Computer-aided diagnosis system for Alzheimer’s Disease
This article discusses how researchers used neural networks and support vector machines as classifiers and principal component analysis as a feature extraction technique to detect and diagnose Alzheimer’s Disease from MRI and PET images. According to the authors, “the combined scheme achieves good accuracy and offers a significant advantage over other approaches.” While specific to diagnostic medicine, this approach can be generalized for use in other application domains.

Seeking Out the Innermost Secrets of Generative AI
This is a lengthy but non-technical article about the ongoing work to figure out how generative AI “works its magic.” If you are interested in learning more about the intricacies of the technology, you should read this.
The Shift from Analog to Digital to Cognitive
This Psychology Today piece traces humans' use of technology from the Analog Age to the Digital Age to Today, which the author terms the “Cognitive Age.” In this emerging age, humans will be interacting with intelligent machines that demonstrate cognitive abilities similar to their own. The author discusses some of the benefits and challenges of this new era.
NIST releases updated AI Risk Management Framework
This was released in late April, but I just saw it this week. The document identifies 10 risks that are “unique to or exacerbated by the use of AI.” Some are well-known, and some are new to me. A PDF of the framework can be accessed by clicking this link.
OpenAI announces it has begun training its newest “frontier model”
There is no link because this announcement was included in OpenAI’s announcement of a new Safety and Security Committee. The new model, which could be dubbed “ChatGPT-5,” will not be released for at least 90 days, but it should be released before the end of the year.

ICON’s New 3D Printer for Construction Creates 27-Foot Tall Building
A giant 3D printer extrudes concrete to accelerate construction and enable new architectural forms. See for yourself:

Monocle
This company caught my eye because of its value proposition. It claims that its use of machine intelligence enables online retailers (and other verticals) to create profitable sales promotions for their customers. The founders make a point in a blog entry that price promotions are much less optimized that ads, a point with which I agree.

High-Impact AI: Strategies & Tactics for Leaders