So much this. It's important to keep in mind what will last and what is the buzz of the day. Especially since the lines between news and entertainment became so blurry.

So transformer models produce things that look plausible... and that's it. What would it look like if we started to make hybrid models in which a transformer model is also tied to proper computation tools with general knowledge? This piece is a good illustration of what it could provide.

Too often forgotten. Data is indeed a mean to an end. It's not outright knowledge and will require work to be useful. It better be aligned with your needs if you want to use it for decision making.

This is indeed very much true... there's a clear crisis in research. It turned into a hamster wheel of publishing articles at a constantly faster pace. The incentives are misguided which pushes that behavior to even have a career. Meanwhile, knowledge building suffers.

OK, this is an interesting practice... I do some of that in a less formal fashion, maybe it's worth exploring further.

Interesting article about how we badly design AI systems which make them very vulnerable to the quality of the data they receive. That's in part why I'd expect that somehow we'll see knowledge representation somehow come back in fashion because they have some potential to lead to better explicability in models.

It's a very nice paper on spreadsheets and how we use them. It got enough history in it to make me tick (goes back all the way to the 1300s!). Also it's well balanced, it's not just about blindly blaming tools but looks at their shortcomings but also how we often use the wrong tool for the task... and then end up managing data and knowledge really badly.