LeanderK

31 ms·

301.
by LeanderK 4y ago
I have seen a score of 83 on twitter
302.
by LeanderK 4y ago
It has long been experimentally shown that neural network do in fact generalise and do not just memorise the training samples. What we do not see here is the convergence of the empirical distribution to the ideal distribution, the data is t
303.
by LeanderK 4y ago
I think a missing piece is that ChatGPT doesn't know when it's either confused or out of depth. It starts producing gibberish, which is very hard to separate from the quality answers. Especially if it would just straightforwardly
304.
by LeanderK 4y ago
I think MLflow is a good idea (very) badly executed. I would like to have a library that combines: - simple logging of (simple) metrics during and after training - simple logging of all arguments the model was created with - simple logging
305.
by LeanderK 4y ago
I wonder how long it will take until you will see a lot AI assisted services pop up. There is a lot of low-budget work that's constantly in need, from thousands of flyers for events to small websites for independent business. I think i
306.
by LeanderK 4y ago
which lack of progress? The few encounters with logic (in a philosophical sense, although it feels more "mathy") I had recently were all more recent works, certainly more recent than a large part of math I use!
307.
by LeanderK 4y ago
That hasn't changed, you are still expected to pick one up yourself except (for me) Java, the first and only language that really got "taught". Maybe that was C++ a few decades ago or something. We had to do a little bit of C
308.
by LeanderK 4y ago
Highly doubt this, but I am no expert. A human is just very flexible and the rovers up there look super complicated while only been capable of very simple tasks. A lot of scientific progress happened because somebody took a risk, not in th
309.
by LeanderK 4y ago
> If normalized, it's part of their expected compensation package. But it already is normalised. I think this is morally difficult because it is customary, expected and if you refuse you will not get anything done but also won'
310.
by LeanderK 4y ago
> very long 2.5 hour commute imagine doing nothing else besides working and commenting. I can't. It sounds like a recipe for a lonely life.
311.
by LeanderK 4y ago
This is a big difference I think. Humans will get distracted but that's not such a problem if you have a reasonable amount of time to get back into control of the situation, but that's not possible in a car.
312.
by LeanderK 4y ago
Well, I was serious in my argument. It is a somewhat realistic next step in my point of view but I consider it absurd, hence the next sentence that we move slowly to a world entirely managed by smartphones and apple honestly just can't
313.
by LeanderK 4y ago
how far will apple take their in-app purchases? It's really getting ridiculous. Will they also request their share if I transfer some money from my bank account via app? Buy a new sofa? A share from airbnb if I rent a spare room? I bet
314.
by LeanderK 4y ago
I wonder whether this is due to unreliable data. I am not sure where this is sourced from since tracking all the solar installations seems like a lot of work.
315.
by LeanderK 4y ago
ah well, then it makes sense. If you can insure your home then the taxpayer effectively subsidises you for your risk, but you don't have to worry.
316.
by LeanderK 4y ago
in this example it is wasteful but it also decreases risks in the long term. The free market does not have to produce the most efficient solution, it just often does.
317.
by LeanderK 4y ago
Shouldn't, in principle, insurance costs reflect risks? If they are comparably expensive but still affordable, then living in these places is still possible.
318.
by LeanderK 4y ago
> 20+ years? Economic growth happens despite degrading social indicators? Where is economic growth in europe happening despite degrading social indicators? Are there really countries with significantly degrading social indicators & e
319.
by LeanderK 4y ago
If universities are the main indicator, why is cambridge with Harvard & MIT not more of a player? I don't really hear much from the startup-scene/tech-scene in Massachusetts.
320.
by LeanderK 4y ago
This sounds like it has an incredible tail-risk. A sound plan if everything goes to plan but if anything serious happens you might loose big time. Everything is essentially placed on one card. I am not sure how much you can reduce it with i
321.
by LeanderK 4y ago
O’Reilly 2042
322.
by LeanderK 4y ago
i would be happy with an app notification and an assessment what my status is :)
323.
by LeanderK 4y ago
I was recently thinking about a similar hardware (although I have no hardware background). I have a lot of plants, some shared with my roommates, and many frequently die. Taking care of plants together is hard and sometimes we are all away
324.
by LeanderK 4y ago
I think this post is simplistic. Chess does not have anything to do with what most middle-managers do and I don't see AI manage humans with more depth, it has to become sentient first essentially. I think for AI to replace most middle
325.
by LeanderK 4y ago
nobody would buy such an expensive iPhone. I bet they will just switch to usb-c, like they did with the iPad.
326.
by LeanderK 4y ago
> I’m aware of your sarcasm, but it did bother me that my Tesla used a proprietary port that made it more difficult to use other charging locations, such as the charger next to my apartment. Electric vehicle charging ports need standardi
327.
by LeanderK 4y ago
Is it getting done? In germany progress is mixed because some cities are way to cautious to introduce bike lanes. They are sometimes planned to death, little strips of a few hundred meters of bike lane takes years (different proposals, inpu
328.
by LeanderK 4y ago
I think it is all a game of perception. If you think you can muscle your way through, either quickly enough so that no harm is done to your economy/population until your goal is reached or if you perceive a weakness in the opposition,
329.
by LeanderK 4y ago
The aurora supercomputer is delayed 4 years...given the progress in compute I wonder whether it will be obsolete on arrival? In 4 years a lot of new competitors should have emerged, right? If my GPU would come 4 years late I would not be ge
330.
by LeanderK 4y ago
I think the problem is that you can't really restrict IEEEs values. I would like to have an error for infinities unless explicitly constructed (so number/number = infinity is an error but number * infinity = infinity is not). So I

More ›