Paper on how to think about how LLMs explain their answers as plausible (do they make sense) & faithful (do they accurately represent how the LLM "thought"). For some cases, like the LLM explaining how to calculate 5!, you want plausible but not faithful. https://
arxiv.org/pdf/2402.04614
.pdf
…
@emollick
-
LLM Explanations: Plausibility Versus Faithfulness Analysis
By
–
-
Industrial Infrastructure as Engineering Achievement and Human Innovation
By
–
This is why I like Infrastructure, a sort of birding guide for the built environment. Reading it, you realize that the industrial landscape is an achievement, the result of complex optimization and hard engineering and human factors. We do that, not pyramids, for better & worse.
-
AI-Generated Magic Cards Showcase Innovation Through Creative Remixing
By
–
I like this feed of AI-generated magic cards, and not just because I am a nerd. They illustrate how AI can be valuable in innovation, even when it is just remixing ideas. Recombinations result in unexpected & unexplored novelty. Humans can easily filter out bad ideas & keep good
-
Expert Identity Threats Drive Overconfident Predictions
By
–
Relevant paper: six studies show when an expert is wrong about a public prediction and is called out for a mistake by another expert, their identity becomes threatened. They tend to respond by doubling down with more overprecise predictions rather than acknowledging errors.
-
Creative AI as Companion to Human Creativity and Discovery
By
–
Creative AI is going to be an interesting companion to human creativity, helping find unexpected connections. For example, Metaphor uses predictive search to guess the links that might shed light on your question. The answers are uneven, but fascinating. https://
metaphor.systems -
Learning from Failure Stories More Effective Than Success
By
–
Interestingly, stories of failure may be more useful for learning. They can be more easily applied than stories of success, especially if the story is interesting and you believe that it is important to learn from mistakes. Just be sure the story is true! https://
ink.library.smu.edu.sg/cgi/viewconten
t.cgi?article=1021&=&context=lkcsb_research_all&=&sei-redir=1&referer=https3A2F2Fscholar.google.com2Fscholar3Fhl3Den26as_sdt3D0252C3926q3Dlearning2Bfrom2Bothers2Bfailure26btnG3D#search=learningfromothersfailure
… -
Organizational factors matter more than technical aspects for code quality
By
–
The best way to ensure good code is to have a good organization. This paper from Microsoft looking at 50M lines of code found that organizational factors were far more powerful in explaining the number of bugs than any technical aspect of the project. https://
microsoft.com/en-us/research
/wp-content/uploads/2016/02/tr-2008-11.pdf
… -
Twitter’s Economic Power: Impact on Stock and Bond Markets
By
–
How powerful is Twitter economically? Not only does Twitter sentiment affect stocks (as we recently saw), but this new paper finds it also “appears to convey economically important information to even the presumably sophisticated investors who dominate the corporate bond market”
-
Survivorship Bias: Why Success Stories Mislead Entrepreneurs
By
–
Why learning from successful companies & entrepreneurs is so hard. Selection bias means that you hear only about successes, many of those successes should have/will fail (but got lucky) & a desire to find secrets of success means we learn the wrong stuff: https://
oneusefulthing.substack.com/p/when-survivo
rship-bias-meets-superstitious?utm_medium=reader2
…
