So if I’m understanding this correctly you have a relativistic understanding of information in which the zero state is observer dependent? Just for my understanding, consider for example the spin of an electron. It can be in one of two states, up or down. In which scenario am I unclear about the absolute information content of knowing the spin state?
Shannon wasn’t the first to think about entropy. It’s a general concept that is hugely important in thermodynamics and statistical physics. Information theory itself becomes increasingly more important in fundamental physics, see the firewall problem of black holes. It’s even harder to overlook these ideas and claim the human brain and our ears are somehow relevant to the definition of information and entropy.
No it does not contradict it. You can have random processes induce a reduction in entropy of a system (e.g. random bit flips can lead to the sequence 111111 which has 0 entropy) and in particular a random process can increase the ability of a system to maintain low entropy over time (e.g. a mutation that makes dna replication during cell division more accurate), which is the more relevant point in the proposed definition.
It isn’t obvious to me that mutation, in general, would increase entropy. There are mutations that would. For example a mutation that would lead to a random bit flip during cell division would certainly increase entropy of the system. In my understanding such a mutation is also pretty bad for your organism. Its actually a somewhat trivial observation that systems that tend to higher entropy over time, “die” because they tend to a maximally unordered state over time.
Maybe I misunderstood you here but I don’t think information theory and in particular entropy has anything to do with “our brain”. It’s simply a statement of the number of micro states in a system. There is no requirement for reading them out or having a brain look at it. This value exists without observer. In that sense there is indeed an intrinsic quality to living beings that leads them to have a reduced number of micro states. They spend energy to keep it that way over time.
I’m not sure about this. Define your “erroneous” state as “halt”. Now the question becomes, for a systems that halts, find out how it reached this state. The mathematical answer to this is simply the description of the Turing machine that produced this state. Whether you can understand this description or not isn’t relevant.
This came up since the beginning in January. The number of people that were reported to reinfect like this seem to lie well within the expected rate of false positives. I don’t understand how this claim is made repeatedly, without a thorough statistical analysis of what is expected.
The best “test” for this I’ve heard is to ask people how many windows they have in their apartment. They might say 5 and then you ask them how they came to this conclusion. A normal thing to do is to walk through your apartment in your mind and count. Someone with a Aphantasia either remembers the number or cannot answer this question.
Ha, funny seeing this mentioned on HN, this student was my Professor for Astrophysics during my time in this field. He was a great mentor and taught me (and others) quite a lot, in particular about outreach and science communication. The Zooniverse project is amazing and citizen science in general is a great idea. However, you mention the crux of the issue already: the tasks need to be enjoyable. Looking at galaxies that potentially no other human ever laid eyes on is pretty much as exciting as it may get. A former colleague of mine (in the group of the mentioned student) worked on this interesting structure https://en.wikipedia.org/wiki/Hanny%27s_Voorwerp that was discovered by one citizen scientist in the Zooniverse project (hence the name). It's one of the coolest success stories of crowd-sourced research I've heard about.
See my reply to your original comment. Exponential growth isn’t a valid reason for this to not work. The factor of recovered vs currently Ill patients is constant, exponential growth or not. This factor may be high and I assume this is what you mean but given the (relatively small) percentage of critically ill and the fact that some countries are further along this is not a clear disqualifying argument against serum.
e^(ct)/e^(c(t - 14days)) = e^(c*14days) = k = constant, with c the growth rate. While you are right that, in an exponential growth scenario, the number of recovered cases is smaller than the number of currently active cases, the factor is proportional to the growth rate which is worst case constant, but likely going down over time due to quarantine measures. As such we have theoretically enough recovered cases if n_infected/n_critical > k. For the most critical age group an estimated 10% of infected require ICU care [1]. Thus we need k < 10. With a 1:2 ratio for serum we get k < 20. Likely not everyone will donate though and constant factors in this equation will end up mattering but the exponential nature does not kill the idea in the manner you indicate. However, it still requires scaling production exponentially. This is true for all other drugs as well though.
Anyone knows why serum is not sufficient to completely solve all our issues? I applaud this trial but this confuses me. It seems like we already have the capacity to do this on large scale, it's considered very safe and as far as I understand is highly effective. There was an article about it here a few days back, but besides that I don't hear much about it in the general discussion, why?
I wonder how vulnerable this system is to adversarial attacks. This will depend on the details of the implementation and is probably easy to guard against (e.g. by using an additional, non-NN based tracker or out-of-distribution detection), but it is an interesting scenario. Would you get arrested for entering a shop with a sweater that features a (seemingly) random noise print, making the payment system malfunction?
What you are stating is the original Lottery Ticket Hypothesis. What they prove in this paper is the stronger version, empirically noticed here https://arxiv.org/abs/1905.01067 and referred to as "supermasks". To quote from the paper posted here: "within a sufficiently overparameterized neural network with random weights (e.g. at initialization), there exists a subnetwork that achieves competitive accuracy".
I think you do misunderstand. They do not add “correlated variables” to a model. The idea is that if you have an overparameterised model for a specific problem, this model contains a smaller model, that has similar performance to the trained large model, without training! That means gradient descent is in fact equivalent to pruning weights in a random network. There is no algorithm for how to do this efficiently (as they show) but that does not mean that there are no (so far unknown) heuristics out there that would get you close. This is exciting as it means a potential alternative for backprop is out there. This would be cool because it might mean more efficient algorithms and something I haven’t seen mentioned in the paper, an alternative to backprop that might be easier to understand in a biologically plausible way.
If that were the case I am wondering why anyone would buy GPUs? I invite you to retrain a state of the art model of your choice on a CPU and see how far you get.
So you are saying the system memory is 300GB and you can train your model on the cpu instead? Well yeah you can always do that but training will be slow because your model is not trained on the GPU. What’s the point?
The concept of ownership is not necessarily the same for digital content. Streaming services for example extended this idea and I don’t believe this is a bad thing. It gives us access to more content for less money. Yes, someone can take it away from you again but there are many advantages to this type of distribution that you could not enjoy if there was a guarantee for lifetime ownership.
Yes and they enjoyed them and probably forgot about them already. If someone would buy my old books I'd be thrilled for most of them. For some selected others devastated, but these I would copy. If they wanted to chip in with another 25$ to pay for my sweet annotations I probably wouldn't even accept it.