Collaborative Intelligence Week in Review - 21Jul2023

Jul 21, 2023

The great acceleration: CIO perspectives on generative AI
This is a survey of CIOs conducted by MIT Technology Review magazine in partnership with Databricks. The article lists several insights on how generative AI could create business value. The full report is available from Databricks, but requires sharing your contact information.

Gradient strikes back: How filtering out high frequencies improves explanations
One role that humans play in CI is explaining to other humans why intelligent machines made the decisions they did (such as accepting or denying a loan application). As such, there is a lot of ongoing research into “explainability.” This article proposes a more computationally-efficient approach for explaining the decisions that come from deep machine learning models.
Sig-Splines: Universal approximation and convex calibration of time series generative models
I do a lot of work with multi-variate time series (both analysis/forecasting of market data for clients and for quantitative investment management). Ofttimes, there isn’t enough data to train ML models. Current synthetic data generation algorithms leave much to be desired. This article proposes a new approach to generation of multi-variate time-series data.

10 Breakthrough Technologies 2023
Each year, MIT Technology Review magazine lists “the 10 technologies that matter the most right now.” They have just published their list for 2023. While this isn’t about CI specifically, I am always interested in this list.

Playbook.ai
This may not appeal to you, but it did to me. It’s a storage place for AI-generated art and images, but it also saves the prompts used to create them. If you’ve spent any time with Dall-E or Stable Diffusion, then you have probably run into the problem of wanting to tweak a previously-generated image, but you can’t remember what the prompt was. This app solves that problem.

High-Impact AI: Strategies & Tactics for Leaders