I would use this as a human. That npm output is crazy. Maybe a better variable would be "CONCISE=1". For LLMs, there are a few easier solutions, like outputing in a file (and then tail)., or running a subagent
> France ("consistently invested in nuclear for the past half-century" ).
Not really and this is currenctly causing a big problem. France stopped building new reactors after 2002. They only built 1 new generation EPR, which was very late and 6x the cost.
Many of the reactors are very old and need to be replaced, but it's difficult to do because of the bad experiences of Flamanville's reactor.
I you hang them at 45 degree the depth will be reducted by sqrt(2). (about 0.7 x hanger length), and you will lose space on each side. And the more you increase the angle, the more you will lose space on the sides.
With this technique, you will reduce the depth to 0.5 x hanger length and not lose space on the side.
For example you could set cookies before visiting another website. This is currently impossible in an iframe but possible in a browser.
I've wanted to do this to automatically login users on some external websites.
With compositions A uses B but B can never use A.
With inheritance Child can use the Parent, but Parent will also call the Child
(virtual methods) which in turn can call the Parent again etc.., so the code can become difficult to follow. It can become very complicated with multiple inheritance and multiple levels.
Humans are pretty bad at these questions. Even with the simplest questions like "Sally (a girl) has 3 brothers. Each brother has 2 sisters. How many sisters does Sally have?" I think that a lot of people will give an incorrect answer. And for questions like "Argue for and against the use of kubernetes in the style of a haiku", 99.99% will not be able to do it.
> Whereas, in biological brains, the weights are updated continuously.
My personal impression is that many "weights" are updated during sleep time. For example when training juggling, I will make no progress at all for hours of training. But later, after a night of sleep, I will have a large and instant progress.
In humans, sleep is also required fo learning, so it's not fully continuous. An AI that occasionnaly retrains using the new knowledge would still be very interesting.
It could be double speak, but the obvious thing you understand from this formulation it that he won because he is a philosopher. Not that he won, and by the way he is a philosopher.
lambdas are very useful for example for sorting data using a key function.
Maybe the auther is saying that it should be called differently, or have a different syntax and the title is just clickbait.
Is outperforming GPT-3 still a good reference? It seems there are many models outperforming GPT-3 in the superglue benchmark: https://super.gluebenchmark.com/leaderboard/
GPT-3 is in position #21, with 71.8% score. The best model is at 91.2%.
Note the human baseline in #6 with 89.8%
And if every parameter is one byte, the minimum, it will take at least 70gb to save or share this model. So it's still way to big to package directly in a app.
I doubt that there are numbers to prove that. SMR might just be a buzzword. From the page below there don't seem too many operating or under construction (5 operating, and 4 under construction). https://www.world-nuclear.org/information-library/nuclear-fu...
Most of the time it's not compute or memory intensive so any computer will do.
The most important thing then is too have a good chair, keyboard and mouse!