On data, causality, and AI

I started coding in C and used C# before switching to Java (those were the dark days). Today, R and Python make up my daily stack (and I dabble in Julia for fun). Now, in retrospect, one habit I wish the data science community had developed better is exception handling. More often than not, I come across code that is doomed to crash (hey, it’s called an exception for a reason).

The next time you code to analyze data, remember this video of division by zero. This is the pain and suffering your computer goes through when your denominator hits zero. This is also a reminder to appreciate (not complain about) the errors and warnings in R and Python (my former and future students, hear me?).

Received the video in personal communication and happy to add a source.

…because experts are more likely to know what they don’t know. When users don’t know what they don’t know, so-called “hallucinations” are less likely to be detected and this seems to be a growing problem, likely exacerbated by the Dunning-Kruger effect. Well, this is my take on them.

In the study cited in the article, several LLM models are asked to summarize news articles to measure how often they “hallucinated” or made up facts.

The LLM models showed different rates of “hallucination”, with OpenAI having the lowest (about 3%), followed by Meta (about 5%), Anthropic’s Claude 2 system (over 8%), and Google’s Palm chat with the highest (27%).

Source

𝘈𝘱𝘱𝘳𝘰𝘢𝘤𝘩 𝘺𝘰𝘶𝘳 𝘥𝘢𝘵𝘢 𝘸𝘪𝘵𝘩 𝘢 𝘮𝘪𝘹 𝘰𝘧 𝘴𝘬𝘦𝘱𝘵𝘪𝘤𝘪𝘴𝘮 𝘢𝘯𝘥 𝘤𝘶𝘳𝘪𝘰𝘴𝘪𝘵𝘺. 𝘛𝘳𝘦𝘢𝘵 𝘥𝘢𝘵𝘢 𝘢𝘴 𝘢 𝘴𝘵𝘰𝘳𝘺𝘵𝘦𝘭𝘭𝘦𝘳 𝘵𝘩𝘢𝘵 𝘥𝘰𝘦𝘴𝘯’𝘵 𝘥𝘪𝘳𝘦𝘤𝘵𝘭𝘺 𝘵𝘦𝘭𝘭 𝘵𝘩𝘦 𝘵𝘳𝘶𝘵𝘩, 𝘣𝘶𝘵 𝘰𝘧𝘧𝘦𝘳𝘴 𝘤𝘭𝘶𝘦𝘴 𝘵𝘩𝘢𝘵, 𝘸𝘪𝘵𝘩 𝘳𝘪𝘨𝘰𝘳𝘰𝘶𝘴 𝘢𝘯𝘢𝘭𝘺𝘴𝘪𝘴 𝘢𝘯𝘥 𝘤𝘳𝘪𝘵𝘪𝘤𝘢𝘭 𝘵𝘩𝘪𝘯𝘬𝘪𝘯𝘨, 𝘳𝘦𝘷𝘦𝘢𝘭 𝘵𝘩𝘦 𝘥𝘦𝘦𝘱𝘦𝘳 𝘯𝘢𝘳𝘳𝘢𝘵𝘪𝘷𝘦. 𝘊𝘶𝘭𝘵𝘪𝘷𝘢𝘵𝘦 𝘵𝘩𝘦 𝘢𝘳𝘵 𝘰𝘧 𝘢𝘴𝘬𝘪𝘯𝘨 𝘵𝘩𝘦 𝘳𝘪𝘨𝘩𝘵 𝘲𝘶𝘦𝘴𝘵𝘪𝘰𝘯𝘴 -𝘯𝘰𝘵 𝘰𝘯𝘭𝘺 𝘰𝘧 𝘵𝘩𝘦 𝘥𝘢𝘵𝘢, 𝘣𝘶𝘵 𝘢𝘭𝘴𝘰 𝘰𝘧 𝘵𝘩𝘦 𝘴𝘵𝘢𝘬𝘦𝘩𝘰𝘭𝘥𝘦𝘳𝘴. 𝘜𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥 𝘵𝘩𝘦 𝘤𝘰𝘯𝘵𝘦𝘹𝘵 𝘢𝘯𝘥 𝘶𝘯𝘥𝘦𝘳𝘭𝘺𝘪𝘯𝘨 𝘱𝘳𝘰𝘤𝘦𝘴𝘴𝘦𝘴 𝘵𝘩𝘢𝘵 𝘨𝘦𝘯𝘦𝘳𝘢𝘵𝘦 𝘵𝘩𝘦 𝘥𝘢𝘵𝘢 𝘵𝘰 𝘢𝘷𝘰𝘪𝘥 𝘮𝘢𝘬𝘪𝘯𝘨 𝘪𝘯𝘧𝘦𝘳𝘦𝘯𝘤𝘦 𝘰𝘯 𝘴𝘦𝘦𝘮𝘪𝘯𝘨 𝘴𝘪𝘨𝘯𝘢𝘭𝘴 𝘵𝘩𝘢𝘵 𝘮𝘢𝘺 𝘢𝘤𝘵𝘶𝘢𝘭𝘭𝘺 𝘣𝘦 𝘯𝘰𝘪𝘴𝘦. 𝘈𝘯𝘥 𝘯𝘦𝘷𝘦𝘳 𝘧𝘰𝘳𝘨𝘦𝘵 𝘵𝘩𝘦 𝘷𝘢𝘭𝘶𝘦 𝘰𝘧 𝘤𝘰𝘮𝘮𝘶𝘯𝘪𝘤𝘢𝘵𝘪𝘰𝘯 𝘢𝘯𝘥 𝘴𝘵𝘰𝘳𝘺𝘵𝘦𝘭𝘭𝘪𝘯𝘨; 𝘺𝘰𝘶𝘳 𝘪𝘯𝘴𝘪𝘨𝘩𝘵𝘴 𝘢𝘳𝘦 𝘰𝘯𝘭𝘺 𝘢𝘴 𝘷𝘢𝘭𝘶𝘢𝘣𝘭𝘦 𝘢𝘴 𝘺𝘰𝘶𝘳 𝘢𝘣𝘪𝘭𝘪𝘵𝘺 𝘵𝘰 𝘦𝘧𝘧𝘦𝘤𝘵𝘪𝘷𝘦𝘭𝘺 𝘤𝘰𝘮𝘮𝘶𝘯𝘪𝘤𝘢𝘵𝘦 𝘵𝘩𝘦𝘮.

The Executive Order defines “AI” as:
“a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments.”

This means that the scope is not limited to generative AI, which is good. Using “AI” as an umbrella term may still not be a good idea for the reasons my assistant lists below but I hope this is a first step in the right direction.

“𝘈𝘐” 𝘢𝘴 𝘢 𝘣𝘭𝘢𝘯𝘬𝘦𝘵 𝘵𝘦𝘳𝘮

𝘖𝘯 𝘵𝘩𝘦 𝘱𝘰𝘴𝘪𝘵𝘪𝘷𝘦 𝘴𝘪𝘥𝘦, 𝘪𝘵 𝘴𝘪𝘮𝘱𝘭𝘪𝘧𝘪𝘦𝘴 𝘤𝘰𝘮𝘮𝘶𝘯𝘪𝘤𝘢𝘵𝘪𝘰𝘯 𝘣𝘺 𝘨𝘳𝘰𝘶𝘱𝘪𝘯𝘨 𝘵𝘰𝘨𝘦𝘵𝘩𝘦𝘳 𝘢 𝘸𝘪𝘥𝘦 𝘳𝘢𝘯𝘨𝘦 𝘰𝘧 𝘮𝘰𝘥𝘦𝘭𝘴 𝘵𝘩𝘢𝘵 𝘦𝘮𝘶𝘭𝘢𝘵𝘦 𝘩𝘶𝘮𝘢𝘯 𝘤𝘰𝘨𝘯𝘪𝘵𝘪𝘷𝘦 𝘧𝘶𝘯𝘤𝘵𝘪𝘰𝘯𝘴 𝘴𝘶𝘤𝘩 𝘢𝘴 𝘭𝘦𝘢𝘳𝘯𝘪𝘯𝘨, 𝘱𝘳𝘰𝘣𝘭𝘦𝘮 𝘴𝘰𝘭𝘷𝘪𝘯𝘨, 𝘢𝘯𝘥 𝘱𝘢𝘵𝘵𝘦𝘳𝘯 𝘳𝘦𝘤𝘰𝘨𝘯𝘪𝘵𝘪𝘰𝘯. 𝘛𝘩𝘪𝘴 𝘴𝘪𝘮𝘱𝘭𝘪𝘧𝘪𝘤𝘢𝘵𝘪𝘰𝘯 𝘤𝘢𝘯 𝘣𝘦 𝘣𝘦𝘯𝘦𝘧𝘪𝘤𝘪𝘢𝘭 𝘧𝘰𝘳 𝘦𝘥𝘶𝘤𝘢𝘵𝘪𝘰𝘯𝘢𝘭 𝘱𝘶𝘳𝘱𝘰𝘴𝘦𝘴, 𝘱𝘰𝘭𝘪𝘤𝘺𝘮𝘢𝘬𝘪𝘯𝘨, 𝘢𝘯𝘥 𝘱𝘳𝘰𝘮𝘰𝘵𝘪𝘯𝘨 𝘱𝘶𝘣𝘭𝘪𝘤 𝘶𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥𝘪𝘯𝘨. 𝘐𝘵 𝘱𝘳𝘰𝘷𝘪𝘥𝘦𝘴 𝘢 𝘤𝘰𝘯𝘷𝘦𝘯𝘪𝘦𝘯𝘵 𝘴𝘩𝘰𝘳𝘵𝘩𝘢𝘯𝘥 𝘧𝘰𝘳 𝘥𝘪𝘴𝘤𝘶𝘴𝘴𝘪𝘯𝘨 𝘪𝘯𝘯𝘰𝘷𝘢𝘵𝘪𝘰𝘯𝘴 𝘳𝘢𝘯𝘨𝘪𝘯𝘨 𝘧𝘳𝘰𝘮 𝘴𝘪𝘮𝘱𝘭𝘦 𝘢𝘭𝘨𝘰𝘳𝘪𝘵𝘩𝘮𝘴 𝘵𝘰 𝘤𝘰𝘮𝘱𝘭𝘦𝘹 𝘯𝘦𝘶𝘳𝘢𝘭 𝘯𝘦𝘵𝘸𝘰𝘳𝘬𝘴 𝘸𝘪𝘵𝘩𝘰𝘶𝘵 𝘨𝘦𝘵𝘵𝘪𝘯𝘨 𝘣𝘰𝘨𝘨𝘦𝘥 𝘥𝘰𝘸𝘯 𝘪𝘯 𝘵𝘦𝘤𝘩𝘯𝘪𝘤𝘢𝘭 𝘥𝘦𝘵𝘢𝘪𝘭𝘴.

𝘖𝘯 𝘵𝘩𝘦 𝘥𝘰𝘸𝘯𝘴𝘪𝘥𝘦, 𝘩𝘰𝘸𝘦𝘷𝘦𝘳, 𝘵𝘩𝘦 𝘵𝘦𝘳𝘮 𝘤𝘢𝘯 𝘣𝘦 𝘮𝘪𝘴𝘭𝘦𝘢𝘥𝘪𝘯𝘨 𝘣𝘦𝘤𝘢𝘶𝘴𝘦 𝘰𝘧 𝘪𝘵𝘴 𝘣𝘳𝘰𝘢𝘥 𝘴𝘤𝘰𝘱𝘦 𝘢𝘯𝘥 𝘵𝘩𝘦 𝘱𝘶𝘣𝘭𝘪𝘤’𝘴 𝘷𝘢𝘳𝘺𝘪𝘯𝘨 𝘪𝘯𝘵𝘦𝘳𝘱𝘳𝘦𝘵𝘢𝘵𝘪𝘰𝘯𝘴 𝘰𝘧 𝘸𝘩𝘢𝘵 𝘈𝘐 𝘦𝘯𝘤𝘰𝘮𝘱𝘢𝘴𝘴𝘦𝘴. 𝘐𝘵 𝘤𝘢𝘯 𝘤𝘰𝘯𝘧𝘭𝘢𝘵𝘦 𝘳𝘶𝘥𝘪𝘮𝘦𝘯𝘵𝘢𝘳𝘺 𝘴𝘰𝘧𝘵𝘸𝘢𝘳𝘦 𝘸𝘪𝘵𝘩 𝘢𝘥𝘷𝘢𝘯𝘤𝘦𝘥 𝘮𝘢𝘤𝘩𝘪𝘯𝘦 𝘭𝘦𝘢𝘳𝘯𝘪𝘯𝘨 𝘮𝘰𝘥𝘦𝘭𝘴, 𝘭𝘦𝘢𝘥𝘪𝘯𝘨 𝘵𝘰 𝘪𝘯𝘧𝘭𝘢𝘵𝘦𝘥 𝘦𝘹𝘱𝘦𝘤𝘵𝘢𝘵𝘪𝘰𝘯𝘴 𝘰𝘳 𝘶𝘯𝘥𝘶𝘦 𝘧𝘦𝘢𝘳. 𝘐𝘯 𝘢𝘥𝘥𝘪𝘵𝘪𝘰𝘯, 𝘵𝘩𝘦 𝘣𝘳𝘰𝘢𝘥 𝘶𝘴𝘦 𝘰𝘧 𝘵𝘩𝘦 𝘵𝘦𝘳𝘮 𝘤𝘢𝘯 𝘰𝘣𝘴𝘤𝘶𝘳𝘦 𝘵𝘩𝘦 𝘯𝘶𝘢𝘯𝘤𝘦𝘥 𝘦𝘵𝘩𝘪𝘤𝘢𝘭, 𝘭𝘦𝘨𝘢𝘭, 𝘢𝘯𝘥 𝘴𝘰𝘤𝘪𝘰𝘦𝘤𝘰𝘯𝘰𝘮𝘪𝘤 𝘪𝘮𝘱𝘭𝘪𝘤𝘢𝘵𝘪𝘰𝘯𝘴 𝘴𝘱𝘦𝘤𝘪𝘧𝘪𝘤 𝘵𝘰 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘵 𝘈𝘐 𝘢𝘱𝘱𝘭𝘪𝘤𝘢𝘵𝘪𝘰𝘯𝘴, 𝘵𝘩𝘦𝘳𝘦𝘣𝘺 𝘩𝘪𝘯𝘥𝘦𝘳𝘪𝘯𝘨 𝘧𝘰𝘤𝘶𝘴𝘦𝘥 𝘥𝘦𝘣𝘢𝘵𝘦 𝘢𝘯𝘥 𝘵𝘩𝘰𝘶𝘨𝘩𝘵𝘧𝘶𝘭 𝘳𝘦𝘨𝘶𝘭𝘢𝘵𝘪𝘰𝘯. 𝘛𝘩𝘦 𝘣𝘭𝘢𝘯𝘬𝘦𝘵 𝘵𝘦𝘳𝘮 𝘤𝘢𝘯 𝘢𝘭𝘴𝘰 𝘰𝘣𝘴𝘤𝘶𝘳𝘦 𝘵𝘩𝘦 𝘴𝘪𝘨𝘯𝘪𝘧𝘪𝘤𝘢𝘯𝘵 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘤𝘦𝘴 𝘪𝘯 𝘵𝘩𝘦 𝘤𝘢𝘱𝘢𝘣𝘪𝘭𝘪𝘵𝘪𝘦𝘴 𝘢𝘯𝘥 𝘳𝘪𝘴𝘬𝘴 𝘰𝘧 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘵 𝘈𝘐 𝘮𝘰𝘥𝘦𝘭𝘴, 𝘱𝘰𝘵𝘦𝘯𝘵𝘪𝘢𝘭𝘭𝘺 𝘭𝘦𝘢𝘥𝘪𝘯𝘨 𝘵𝘰 𝘢 𝘰𝘯𝘦-𝘴𝘪𝘻𝘦-𝘧𝘪𝘵𝘴-𝘢𝘭𝘭 𝘢𝘱𝘱𝘳𝘰𝘢𝘤𝘩 𝘵𝘰 𝘱𝘰𝘭𝘪𝘤𝘺 𝘢𝘯𝘥 𝘨𝘰𝘷𝘦𝘳𝘯𝘢𝘯𝘤𝘦.

Source

Every time I fly, I am struck by how archaic airport baggage handling still is. Sure, the airline industry is infamous for maintaining its legacy Fortran and Cobol software, but that’s because aviation is a pioneer in using computers to run its operations. Meanwhile, baggage handling remains a bottleneck in air travel because the process is highly manual and inefficient (except for using the same conveyor system that seems to have been in use since 1971).

When robots take over (or assist with) baggage handling, overall passenger satisfaction is likely to improve. Increased use of robots to solve such low-stakes bottleneck problems may also help the public perception of robots.

evoBot looks like one of the robots that can solve this problem. The robot achieves excellent balance using an inverted pendulum design, and can reach speeds of 37 mph and carry over 220 pounds. Pretty impressive.

Not shown in the video, but it can also lift luggage off the ground and deliver it to its destination (airplane or the 1971 conveyor belt). Munich Airport seems to have tested it already. I hope to see it in action soon.

If you ask it, its answer is “Yes.” If you ask it if it “understands” Chinese, its answer is again “Yes” without hesitation. Searle’s 1980 Chinese Room argument is more relevant than ever in the age of LLMs:

𝘚𝘶𝘱𝘱𝘰𝘴𝘦 𝘢 𝘮𝘰𝘥𝘦𝘭 (𝘣𝘰𝘹 𝘪𝘯 𝘵𝘩𝘦 𝘱𝘪𝘤𝘵𝘶𝘳𝘦) 𝘵𝘩𝘢𝘵 𝘣𝘦𝘩𝘢𝘷𝘦𝘴 𝘢𝘴 𝘪𝘧 𝘪𝘵 𝘶𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥𝘴 𝘊𝘩𝘪𝘯𝘦𝘴𝘦. 𝘐𝘵 𝘵𝘢𝘬𝘦𝘴 𝘊𝘩𝘪𝘯𝘦𝘴𝘦 𝘤𝘩𝘢𝘳𝘢𝘤𝘵𝘦𝘳𝘴 𝘢𝘴 𝘪𝘯𝘱𝘶𝘵 𝘢𝘯𝘥 𝘱𝘳𝘰𝘥𝘶𝘤𝘦𝘴 𝘰𝘵𝘩𝘦𝘳 𝘊𝘩𝘪𝘯𝘦𝘴𝘦 𝘤𝘩𝘢𝘳𝘢𝘤𝘵𝘦𝘳𝘴 𝘢𝘴 𝘰𝘶𝘵𝘱𝘶𝘵. 𝘛𝘩𝘪𝘴 𝘮𝘰𝘥𝘦𝘭 𝘱𝘦𝘳𝘧𝘰𝘳𝘮𝘴 𝘪𝘵𝘴 𝘵𝘢𝘴𝘬 𝘴𝘰 𝘤𝘰𝘯𝘷𝘪𝘯𝘤𝘪𝘯𝘨𝘭𝘺 𝘵𝘩𝘢𝘵 𝘪𝘵 𝘤𝘰𝘮𝘧𝘰𝘳𝘵𝘢𝘣𝘭𝘺 𝘱𝘢𝘴𝘴𝘦𝘴 𝘵𝘩𝘦 𝘛𝘶𝘳𝘪𝘯𝘨 𝘵𝘦𝘴𝘵: 𝘪𝘵 𝘤𝘰𝘯𝘷𝘪𝘯𝘤𝘦𝘴 𝘢 𝘩𝘶𝘮𝘢𝘯 𝘊𝘩𝘪𝘯𝘦𝘴𝘦 𝘴𝘱𝘦𝘢𝘬𝘦𝘳 𝘵𝘩𝘢𝘵 𝘵𝘩𝘦 𝘮𝘰𝘥𝘦𝘭 𝘪𝘴 𝘪𝘵𝘴𝘦𝘭𝘧 𝘢 𝘭𝘪𝘷𝘦 𝘊𝘩𝘪𝘯𝘦𝘴𝘦 𝘴𝘱𝘦𝘢𝘬𝘦𝘳. 𝘛𝘰 𝘢𝘭𝘭 𝘰𝘧 𝘵𝘩𝘦 𝘲𝘶𝘦𝘴𝘵𝘪𝘰𝘯𝘴 𝘵𝘩𝘢𝘵 𝘵𝘩𝘦 𝘱𝘦𝘳𝘴𝘰𝘯 𝘢𝘴𝘬𝘴, 𝘪𝘵 𝘮𝘢𝘬𝘦𝘴 𝘢𝘱𝘱𝘳𝘰𝘱𝘳𝘪𝘢𝘵𝘦 𝘳𝘦𝘴𝘱𝘰𝘯𝘴𝘦𝘴, 𝘴𝘶𝘤𝘩 𝘵𝘩𝘢𝘵 𝘢𝘯𝘺 𝘊𝘩𝘪𝘯𝘦𝘴𝘦 𝘴𝘱𝘦𝘢𝘬𝘦𝘳 𝘸𝘰𝘶𝘭𝘥 𝘣𝘦 𝘤𝘰𝘯𝘷𝘪𝘯𝘤𝘦𝘥 𝘵𝘩𝘢𝘵 𝘵𝘩𝘦𝘺 𝘢𝘳𝘦 𝘵𝘢𝘭𝘬𝘪𝘯𝘨 𝘵𝘰 𝘢𝘯𝘰𝘵𝘩𝘦𝘳 𝘊𝘩𝘪𝘯𝘦𝘴𝘦-𝘴𝘱𝘦𝘢𝘬𝘪𝘯𝘨 𝘩𝘶𝘮𝘢𝘯 𝘣𝘦𝘪𝘯𝘨. 𝘐𝘯 𝘵𝘩𝘪𝘴 𝘤𝘢𝘴𝘦, 𝘥𝘰𝘦𝘴 𝘵𝘩𝘦 𝘮𝘢𝘤𝘩𝘪𝘯𝘦 𝘭𝘪𝘵𝘦𝘳𝘢𝘭𝘭𝘺 𝘶𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥 𝘊𝘩𝘪𝘯𝘦𝘴𝘦? 𝘖𝘳 𝘪𝘴 𝘪𝘵 𝘮𝘦𝘳𝘦𝘭𝘺 𝘴𝘪𝘮𝘶𝘭𝘢𝘵𝘪𝘯𝘨 𝘵𝘩𝘦 𝘢𝘣𝘪𝘭𝘪𝘵𝘺 𝘵𝘰 𝘶𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥 𝘊𝘩𝘪𝘯𝘦𝘴𝘦?

More recently in his book, Searle linked his original argument to consciousness, but that’s probably a higher bar than needed to reason that ChatGPT is a box that has no idea what it’s talking about.

I took several rides in Google’s Waymo robotaxi. This is a short video of the experience, which is great, almost flawless. One problem is that it gets boring really fast.

About half the time during my trip, I used Uber or Lyft instead of a Waymo, and I met a professional dancer, a retired chef, a compliance officer, a criminal justice expert, an Amazon truck driver, and a painter.

I had really fun conversations that touched on “AI” and dance music, the best old school restaurants in town, the private equity fundraising process, cybersecurity, privacy, more “AI” and so on. As a bonus, almost all of the conversations included some useful, local information about the city.

None of the robotaxi rides had any of this, serendipity was nonexistent. The robot feels friendly, sure, but that’s about it. The longer the ride, the more boring it gets.

Replacing influencers with generative AI looks like a great use case. The real question is whether influencers who promote “AI” will also be replaced by AI.

Some details:
With just a few minutes of sample video from the person to be cloned and a payment of $1,000, brands can clone a human streamer to work 24/7.

The AI videobots may already be having some economic impact: the average salary for livestream hosts in China is down 20% from 2022 (just another YoY figure, not a causal effect).

Source

This is based on a recent Nature study, and it’s useful with a caveat that may make the findings and visuals less striking than they look:

“𝑁𝑎𝑡𝑢𝑟𝑒 𝑠𝑒𝑎𝑟𝑐ℎ𝑒𝑑 𝑓𝑜𝑟 𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠, 𝑟𝑒𝑣𝑖𝑒𝑤𝑠 𝑎𝑛𝑑 𝑐𝑜𝑛𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑝𝑎𝑝𝑒𝑟𝑠 𝑖𝑛 𝑆𝑐𝑜𝑝𝑢𝑠, 𝑤𝑖𝑡ℎ 𝑡𝑖𝑡𝑙𝑒𝑠, 𝑎𝑏𝑠𝑡𝑟𝑎𝑐𝑡𝑠, 𝑜𝑟 𝑘𝑒𝑦𝑤𝑜𝑟𝑑𝑠 𝑐𝑜𝑛𝑡𝑎𝑖𝑛𝑖𝑛𝑔 𝑡ℎ𝑒 𝑡𝑒𝑟𝑚𝑠 ‘𝑚𝑎𝑐ℎ𝑖𝑛𝑒 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔’; ‘𝑛𝑒𝑢𝑟𝑎𝑙 𝑛𝑒𝑡*’, ‘𝑑𝑒𝑒𝑝 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔’, ‘𝑟𝑎𝑛𝑑𝑜𝑚 𝑓𝑜𝑟𝑒𝑠𝑡’, ‘𝑑𝑒𝑒𝑝 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔’, ‘𝑠𝑢𝑝𝑝𝑜𝑟𝑡 𝑣𝑒𝑐𝑡𝑜𝑟 𝑚𝑎𝑐ℎ𝑖𝑛𝑒’, ‘𝑎𝑟𝑡𝑖𝑓𝑖𝑐𝑖𝑎𝑙 𝑖𝑛𝑡𝑒𝑙𝑙𝑖𝑔𝑒𝑛𝑐𝑒’, ‘𝑑𝑖𝑚𝑒𝑛𝑠𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦 𝑟𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛’, ‘𝑔𝑎𝑢𝑠𝑠𝑖𝑎𝑛 𝑝𝑟𝑜𝑐𝑒𝑠𝑠𝑒𝑠’, ‘𝑛𝑎𝑖̈𝑣𝑒 𝑏𝑎𝑦𝑒𝑠’, ‘𝑙𝑎𝑟𝑔𝑒 𝑙𝑎𝑛𝑔𝑢𝑎𝑔𝑒 𝑚𝑜𝑑𝑒𝑙𝑠’, ‘𝑙𝑙𝑚*’, ‘𝑐ℎ𝑎𝑡𝑔𝑝𝑡’, ‘𝑔𝑎𝑢𝑠𝑠𝑖𝑎𝑛 𝑚𝑖𝑥𝑡𝑢𝑟𝑒 𝑚𝑜𝑑𝑒𝑙𝑠’, ‘𝑒𝑛𝑠𝑒𝑚𝑏𝑙𝑒 𝑚𝑒𝑡ℎ𝑜𝑑𝑠’.”

So, SVM, Naive Bayes, Random forest, and Ensemble methods are all called AI (not untrue, but…). Gaussian processes? Well, papers with a GP regression count then. Dimensionality reduction? So, papers using PCA or LDA count.

This feeds the trend of using AI as an umbrella term unfortunately.

Source

The business problem is to put a lifeguard station on a beach to save some lives (i.e., find the best location for the lifeguard station). This is not really a predictive modeling problem. But that’s the hammer our data scientists have and they have access to fancy libraries. There is also some historical data: swimmers rescued and drowned at other beaches. It all checks out. Resistance to 𝘱𝘪𝘱 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘱𝘳𝘰𝘱𝘩𝘦𝘵 is futile.

Transforming the problem into an objective function could have signaled that this is an optimization problem (a prescriptive modeling problem), but that step was skipped. In the picture shown, we may need a solution:

– minimizes distance => 𝗦𝗼𝗹𝘃𝗲𝗱 𝘂𝘀𝗶𝗻𝗴 𝗽𝗶𝗽 𝗶𝗻𝘀𝘁𝗮𝗹𝗹 𝗳𝗮𝗻𝗰𝘆_𝗹𝗶𝗯𝗿𝗮𝗿𝘆
while also…
– minimizing time => 𝗧𝗵𝗲 𝗱𝗼𝗺𝗮𝗶𝗻 𝗲𝘅𝗽𝗲𝗿𝘁 𝗲𝗻𝘁𝗲𝗿𝘀 𝘁𝗵𝗲 𝗿𝗼𝗼𝗺
– minimizing swimming => 𝗧𝗵𝗲 𝗹𝗮𝗯𝗼𝗿 𝘂𝗻𝗶𝗼𝗻 𝗶𝗻𝘁𝗲𝗿𝘃𝗲𝗻𝗲𝘀
– minimizing time to ice cream => 𝗧𝗵𝗲 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝘃𝗲 𝗹𝗲𝗮𝗱𝗲𝗿𝘀𝗵𝗶𝗽 𝘀𝘁𝗲𝗽𝘀 𝗶𝗻
– [not shown] minimizing walking on sand => 𝗧𝗵𝗲 𝗗𝗲𝗽𝗮𝗿𝘁𝗺𝗲𝗻𝘁 𝗼𝗳 𝗟𝗮𝗯𝗼𝗿 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗺𝗲𝗻𝘁
and hopefully not…
– maximizing time => 𝗔 𝗷𝘂𝗻𝗶𝗼𝗿 𝗱𝗮𝘁𝗮 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁 𝘀𝗼𝗹𝘃𝗲𝘀 𝘁𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺

So, the ideal solution requires more thinking about the problem. For example, maximizing the number of lives saved may actually require constraints on how to minimize time so that lifeguards don’t risk their lives during the rescue.

The law of the instrument works a little too well in predictive modeling (and more generally in machine learning). Objective functions are often lost in translation when they should be an explicit step in the modeling process. Best practice tends to favor performance metrics, even though achieving the highest performance on the wrong function is clearly useless (and sometimes detrimental).

More focus on objective functions and less obsession with “better performance” may be what we need. This would underline the importance of problem formulation and domain knowledge, and undermine the 𝘱𝘪𝘱 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘱𝘳𝘰𝘱𝘩𝘦𝘵 solution.

A combination of Warren Powell‘s writing and the accompanying xkcd comic inspired this post (courtesy of xkcd.com).

I have had a copy of Scott Cunningham‘s “Mixtape” since it came out. I’ve skimmed through it before, but last night, while putting together a few slides, I read an entire chapter and loved it. It has just enough detail to keep the reader from going to the cited work. It is also candid and fun. The print version is nicely sized and designed so you can blend in and look cool while others around you are reading the latest fiction. The book has already made its impact and this is probably a late call but I had to share.

Source

We seem to have a growing barrier to discussing AI: the use of AI as an umbrella term. My former students will say “Here we go again,” but if something means everything, it means nothing. If we take the time to define what we mean when we refer to AI, it will probably help the conversation.

Attached is a figure I’ve been using in my classes since 2017 to make this point (sorry, not the cat picture but the following figure of a timeline from AI to ML to Deep Learning). We might be better off referring to specific models and algorithms (or at least a group of models, such as LLMs, instead of AI).

Over the weekend, I attended a series of discussions on “AI” at the Academy of Management‘s annual conference. I had the opportunity to hear the perspectives of great scholars from a variety of backgrounds. Once again, I was puzzled as to what was meant by “AI” in most of the discussions.

Source

In Data Duets, Duygu Dagli and I offered our take on Ashesh Rambachan and Jonathan Roth‘s recently published but long overdue paper now titled “A more credible approach to parallel trends.”

Problem:
We want to test the causal effect of a promotion on sales, let’s say a coupon. The coupon was sent to newer customers. Did the coupon increase the sales? Or would the new customers have bought more anyway?

Solution:
We’ll never know the answer to the last question but we can answer the first question after making some assumptions. More on this in the post.

This is less elaborate than our earlier posts on synthetic controls and Lord’s paradox. We will probably keep it this way so that we can post more often.

Source

The fact that 73% of consumers trust content created by generative AI models is intriguing. And it’s not just people playing around with a chatbot for trivial conversations:*

– 67% believe they could benefit from receiving medical advice from a generative AI model
– 66% would seek advice from a generative AI model on relationships (work, friendships, romantic relationships) or life/career plans
– 64% are open to buying new products or services recommended by a generative AI model
– 53% trust generative AI-assisted financial planning

To put this into perspective, only 62% of people trust their doctor the most for medical advice.**

* 2023 survey by Capgemini of 10,000 consumers
** 2023 survey by OnePoll for Bayer of over 2000 adults

Source

H/T to Travis Gerke, I’ve just discovered the wonderful work of Benjamin Elbers. tidylog provides feedback for dplyr and tidyr operations in R. This is another simple and powerful idea that basically uses wrapper functions for the dplyr and tidyr functions with feedback added. This will help greatly with both teaching and “doing.”

Source

Pandas AI is an interesting and somewhat natural direction for embedding large language models into data science/analytics. This is less of a black box than automated exploratory data analysis tools, but still makes things easier.

We will likely see more ideas like Gabriele Venturi‘s here. For any serious project, though, we’ll still need skilled humans who can understand how the algorithm is responding to queries, and can check and confirm that it’s responding as expected.

Source

In this new post, Duygu Dagli and I took a quick look at the analysis of experimental data. I really enjoyed writing this piece because Lord’s revelation is one of my favorites (pardon the pun).

Lord’s paradox is related to the better known Simpson’s paradox and it highlights the importance of constructing the right conceptual model before moving on to modeling the data. In the post, I speculated about one potential conceptual model and discussed its implications for modeling the data at hand.

Frankly, the example in the post had much to unpack. I just picked up on the part that relates to causal models and Lord’s paradox. I also ended up touching on an interesting discussion around the use of diff-in-diff vs. lagged regression models.

After running an experiment, how do you estimate the average treatment effect (ATE)? Which model do you choose to use? In this post, we use five different models with different assumptions to answer the same question. We find five different ATEs… Which one is the correct average treatment effect in this experiment? How do we decide?

In this post, Gorkem Turgut Ozer and I explore these questions (and more) by discussing the differences across models and potential implications. We ended up covering an interesting paradox I enjoyed learning about!

To me, a main takeaway is, business value from data is maximized when the right conceptual model meets the right method. For this to happen more often, data science and pricing leaders need the technical skills to be able to ask the right questions. They also need to build a trusting relationship with their teams to delegate and learn from them.

Source

Across the board, ChatGPT is passing exams by answering a mix of short-answer, essay, and multiple choice questions:

– U.S. medical licensing exam (says the attached study)
– Wharton School MBA exam on Operations Management
– University of Minnesota Law School exams in Constitutional Law, Employee Benefits, Taxation, and Torts

If ChatGPT is able to pass these exams, it is not because ChatGPT is revolutionary (while it is surely impressive) but because they are just bad exams. These exams must be lacking enough components that require some form of creative thinking and use of imagination.

Source

What is demonstrated here is a successful translation from human language to code. OpenAI has another project for this purpose: Codex. Microsoft’s GitHub Copilot serves as a specialized version (both are descendents of GPT-3). DeepMind’s AlphaCode and the open source PolyCoder also target English to code translation.

What is missing (and provided by Marco) is the articulation of a solution that stems from a conceptual model, which in turn, is informed by causal links. For example: diversification reduces asset-specific risk.

Unless ChatGPT reasonably limits the weights of each individual stock based only on the objective at the beginning (minimize SD) without being explicitly instructed, we’d better curb our enthusiasm here.

Source

Just tried out ChatGPT, the new large language model trained by OpenAI, and I was blown away by its capabilities! It can generate human-like text responses to any prompt, making it a powerful tool for conversation simulation, language translation, and more.

I also had a chance to play around with the code, and it’s surprisingly simple to use. Here’s a quick example of how to generate a response from ChatGPT using the Python API:
…

—

Not bad but a bit too excited (blown away, really?). Also the shameless self-promotion using my voice without any disclosure. We’re not off to a good, trusting relationship.