I use LLMs to answer certain questions, but those are often questions that I wouldn't have bothered using a search engine for in the past, rather I would have asked a colleague or just thought through the question on my own. And when I try to ask an LLM questions that search engines are good at, I'm most often disappointed.
In other words, it's not clear to me LLMs are going to eat into the market share of search engines, rather than just providing a tool with largely orthogonal use cases. But we'll see how the tech develops from here.
"When asked to draw circles representing themselves and friends or family, for example, people tend to self-inflate their own circle but they self-inflate more in individualist cultures."
This sort of methodology sounds sketchy to me - how much can we really learn from this? Does it reproduce? If it does, how do we know there isn't some other cause?
The average income of FT readers is £221k [0] and the average income of WSJ readers is $234k [1]. I'm guessing the average reader of most online only newspapers is closer to the US median of $74k [2]. These readers belong to different economic classes with different financial interests, so the agenda divide is not surprising!
The geometric mean (6.9) is all that really matters for investors, not the arithmetic mean (8.4) - the arithmetic mean under-weights the importance of negative years to long term performance.
For example, if the market is down 20% one year and up 20% the next year, the arithmetic mean will be 0%, but you'll be down 4% (0.8*1.2 = 0.96), which is reflected in the geometric mean of (about) -2%.
I want to believe this. Do you have any citations?
My experience would suggest shipping higher quality software is a lot cheaper in the long run, but has up-front costs that often times startups can't afford.
With mastery learning, it's the feedback loop that's making the difference. In some domains where the subject matter itself can be computationally modeled (such as math), it seems plausible that the feedback loop as well can be emulated by software. There have been many attempts and many failures to do so, but I'm optimistic and believe it's a matter of further investment and research.
The cost of nuclear energy has been flat for decades, but the cost of sources like solar has been plummeting. Nuclear doesn't have much of change without some technological break through.
This shows that wealthier people are moving to California, while middle and lower income people are leaving (People with an income lower than 110k a year).
The problem isn't that he was wrong, but that his method of reasoning about risk is dangerously flawed. Of course, you won't get much of that analysis in a buzzfeed article, but I recommend this article by Nassim Taleb: https://forecasters.org/blog/2020/06/14/on-single-point-fore...
This is a tragedy in its own right. Small businesses create richer communities and help stave off cultural homogeneity. They can be a source of pride for locals. They help de-centralize economic power. I hope we can find a way to rebuild what's being lost here.
Counterargument: If you believe that a system is immoral, its better to minimize your participation in that system, even if you're coerced to participate to a certain degree.
Interesting. Looking at Palmer's claim: "Try saying anything negative about the 五毛, or even mentioning them at all. Your comment will last about 30 seconds and get deleted without warning or notice, CCP-censor style." This seems to be evidence against the broader claim that anything gets deleted, rather than just negative comments.
Yes, that exact sequence of words isn't particularly common. And yet a child, even if they have hey have never been exposed to it, has no problem understanding what it means.
Whereas all these services seem to be processing the input in such a superficial way that they give the searcher results that aren't just inaccurate but are the opposite of what was asked for.
1. With complex social phenomena like this, we're not going to have rock-solid scientific demonstration. Doesn't mean it's not worth considering the possibility of a connection and what implications that would have.
2. Because it's a disturbing (to many) aspect of inequality in the U.S. today.
3. Yes. Times change. The point of articles like this is to make observations about those changes, their significance, and sometimes to make some value judgements about them. I'm not sure what your point is.
4. Is the author claiming that they alone can unwind structural inequalities? When participants in an institution can be critical of their and associated institutions' role in a social problem, and be honest about the limits of said institutions in addressing those problems, that seem useful.
Perhaps worth noting that Kuhn wasn't attempting to lay out a methodology - he wasn't saying "this is how you should do science". Rather, he was a historian making generalizations about how different sciences developed.
From a business perspective, the volatility of Silver's models seem like a feature - they enhance the drama / sensationalism of election coverage. One week, he's telling me candidate X will likely win, the next week it's candidate Y, and I'm on the edge of my seat.
If Silver really believed these probabilities were correct, he should be willing make bets with these odds, otherwise his incentives are distorted.
Okay, but the problem is, 45-60 minutes is usually all hiring managers have to assess a candidate. What can you do in that amount of time to assess someone‘s real world performance? Having a high level conversation about technology and problem solving approach helps but often doesn’t translate into writing code. Tech-stack specific challenges are no good because general programming ability is more important, you can learn a new language on the job. Algorithm puzzle problems aren’t great because then we’re selecting for people who spend a lot of time on hackerrank or whatever.
Maybe effective programmer hiring is just an unsolved problem.
In other words, it's not clear to me LLMs are going to eat into the market share of search engines, rather than just providing a tool with largely orthogonal use cases. But we'll see how the tech develops from here.