Collaborative Intelligence Week in Review - 16Aug2024
Best business article(s) I read this week…
Where Data-Driven Decision-Making Can Go Wrong
The rapid adoption of AI and machine learning in companies has also increased organizational focus on data-driven decision-making. This article lists five ways in which data-driven decision-making can lead to suboptimal or poor decisions. They are:
Conflating correlation and causation
Misjudging the potential magnitude of effects
Creating a disconnect between what is measured and what matters
Misjudging generalizability
Overweighting a specific result
Best technical article(s) I read this week…
Localized Neural Network Modeling of Time Series: A Case Study on US Monetary Policy
I think I have previously written that I have a side business and a strong interest in applying AI/ML to quantitative finance. This may not appeal to many, but I liked it.
Other items I found valuable/interesting…
This 300-Year-Old Winemaker Is Embracing AI
This story melds two of my favorite topics: wine and AI. It describes how Château Ducru-Beaucaillou is using AI to improve quality and increase production. For example, the winery uses AI-powered sensors to monitor vine health and sap flow and to predict disease and frost risks.
Researchers Ranked AI Models Based on Risk—And Found a Wild Range
Researchers developed a benchmark that uses thousands of prompts to test chat model performance against government regulations and usage guidelines published in the US, China, and the EU. Each model did well on some aspects of risks but not others. None did well on all.
In a development related to the previous article, researchers at MIT released a publicly available database of 700 risks related to AI, categorized by risk domain. The database is available as a Google Sheet, and there is an accompanying explainer video.
Coolest thing I saw…
The video below describes how Siemens, in collaboration with a startup (Sewbo) and the Advanced Robotics for Manufacturing Institute, has introduced a robot that can complete sewing tasks.
A company that caught my eye…
The company has released a platform that enables software developers to create realistic digital twins for use on conversational tasks. The video at this link demonstrates the platform’s capabilities, and the company’s website link in the title provides an opportunity for a live, interactive demo. When I tried it, it wasn’t as responsive as it is in the video, but I guess that’s to be expected.