I did already have the opportunity to test drive a Tesla here in Belgium and liked it even if it was too short to my liking (couln't test lane keeping feature). But at 58.800 Euros (chepeast in Belgium), I will pass:( At 35.000 Euros I could think about it and compare with what other automakers will have at that time (ie: If they get their 35.000 version before the other automakers get their fingers out of their collective arses, Tesla will have my money !)
What could be that unicorn real work number crunching application that couldn't be written in 1/10 of lines of code of python in some numpy with a pinch of DAAL or tensorflow? Why most AI research who is not know to be low on the number crunching departement is done in Python? There is even Games developped in Python (Sims 4, Eve Online, Civilisation IV, Pirates of the Caribbean Online...).
In my career, I have written my share of Assembly and C code for performance reasons when a 486 DX 400 was the must have of the shinny machines. But now, compilers and 'interpreters' can do a better work than I due to the complexity of the processors and their different caches sizes. Right now you can write some highly optimized assembly code that work at nearly the theorical speed limit of a processor with a specific cache size, but who will work like a molasse on a similar architecture with double the cache size. An interpreter could adapt its realtime compilation to the architecture it run on and even for the same algorithm to the type of data you process.
Now, if your Python program is still too slow, you can update your slowest function into cython and compile it to have a speed boost that C can get you without losing the time needed to develop everything in C and by that saving a lot of time to go to the next project.
In the beginning of my carreer, I would have used the best language for the job from a toolbelt composed of Assembly, C/C++, Fortran, Delphi, Lisp and Prolog. Now, I use most exclusively Python for my projects.
As long as it take more space and more weight than an IC performing an equivalent task, this will stay a nice research subject. We replaced hardware radio receptor by software ones, I don't see why we would replace software neural network by hardware ones... Plus, they miss the 'plasticity' of software ones and a scratch would make it 'dead':(
Trigger happy cops will now be happier. The dead will not resist that caps use its fingerprints to unlock their phone... Proof gathered this way should be invalidated or else cops will be more inclined to kill the suspects to access more easily their phones:(
Don't blame the satnav for that, but the map providers. A lot of maps miss the informations about the category of the road, max speed, limitations (height because of a bridge, max wheight for lorries...).
Why does these API have been created for? I don't really se a use case for them:( What other API are being put in without any really interesting use-case?
If there is only 1 bus every day, even if the bus is free, I will not take that bus. I live 6km from my work, there is a bus that stop just in front of my house every hour. I still take my car to go to work. Why? By car it take me 15-20min to go to work in the morning and 8-10min to go back in the evening. By bus with a change for a metro it take me ~20min in the morning and between 45min and 1h15 in the evening. Even if the bus and metro were free, I still value the time lost way higher than the price of riding with my car.
2 times I have ordered takeout from a 1 star michelin restaurant without any problems. Very few restaurant refuse takeout. The only starred restaurent who refused me a takeout was because it say that he doesn't want to sell food that doesn't look as good as presented on a plate.
Most of the time takeout order are a bonus for restaurants because it doesn't take a seat in their dinning room and then they can sell one more serving to a customer.
Like said before very few restaurant promote that they can do takeout because for most people, takeout mean cheap and low quality, even if it's not true. Restaurant doesn't want that idea sticking in their customers head.
When I played MtG, I bough at least 3 BOX of boosters when a new series of card were available. And then did I do some cards shopping if really I needed them. But I always started by buying a lot of boosters. When you play competitively, its a good way to not pay an arm and a leg even if it still costing you the alf of an arm:)
I have actually rewritten my algorithms (written in python) to run them in OpenCL to gain that 100x speedup. And I really need that speedup for the learning process but not for the 'playing' process.
More than 97% of the time is used in the computation of matrix product, matrix sums, applying an activation function on a vector and do some max-pooling on these same vectors. All these functions have some kind best algorithms or implementation that take advantage of cache and instruction sets. For the matrix product, I don't think I will find by myself a better algorithm than winograd. Any advancement I could put on the algorithmic side would be in the 3% left coded in python. Even if I double the speed of the rest of my program the speedup will be less than 2% in the end.
The quote I will get from the teacher is mostly dependant on how well I followed the specifications, the clarity of my code, and how my bot fare against the other student bots. Having a more complex algorithm or implementation to shave some CPU cycles here and there will only make the source code of the bot less readable for my teacher and make him raise more concern and thus more scrutinity on my code. This is why my code for the bot will not use my OpenCL implementation who is around 600 lines of OpenCL code. It use about 30 lines of numpy code and is running under 0.5s (ie: mostly instant for the end-user and in the specs for the project). This is good enough for the part of my project that I will present to the teacher.
Has AMD (ATI) video cards do some speculative execution, Spectre is most probably also a problem for them.
From there, there is some possibilities: - They tried to let it go under the rug by not talking about it for their video card products - They are still in the process of working around - They don't think that problem is a problem for video card workload (non sensitive) - They are greedy and won't lose money working on it - They don't have the problem - They are incompetent
Most of your programming assignments: yes. All, certainly not! I have coded some assignments that put my i7-6700HQ with 16GB of ram on its knees (salesman problem, perceptron...), running them on a 486 would be madness:( Right now, my current project is a program that could play othello reversi. I am doing it with a neural network without tree search. The trainning of the neural network is done in a similar way of Alphago Zero (self taught with reinforcement learning) and is running since 2 weeks on my GPU (Nvidia 960M). The programm I will deliver to the teacher will only use the CPU, but for trainning I use my GPU with OpenCL to get an >100x speedup for the computations. The IA right now is good enough to beat me on each game, but not yet at the level of programs like Wzebra.
Think a little about it. Does your camera is at exactly the same place? Did your camera CMOS chip have exactly the same defects as the one they use? Do you take your still images at exactly the same time? And there is so much parameters that make your easy hack not so easy. The easy hack would be to put some kind of spying apparatus behind their camera that send the information back to you, not putting another camera.
I write code in C, PHP, Python, Pascal, VBA, Common Lisp, Prolog, Assembler (z80, 6502, x86). I can help people with their code in C#, C++, Java, Smalltalk. From a language user point of view I am a Jack of all trades, master of none. I am very reliant on the compiler/interpreter to highlight me where I made syntactic errors or passed the wrong parameters to a function or API because in that language, the order of the parameter is different than all the other languages.
But you will not get me on the evaluation of the complexity of an algorithm. I have written programs/libraries from a wide range of domains (robotics, biology equipment, domain specific compilers, pay, stock management and ordering, automatic packing of orders, display of industrial process parameters, expert system, machine learning). You can really thrust me in software design and algorithmic, but I am still very reliant on my IDE and the compiler/interpreter to help me about typos and showing me that the construct I use is not the correct one for the language I am currently using (my black sheep is the switch/case/select construct).
Some composer were bad at playing instruments like Mahler, Berlioz, Schoenberg, Ravel. Mahler and Berlioz were just good enough to play their own composition to help them in their creative process but not in public. Schoenberg and Ravel used the help of assistants to play them what they have composed.
Internet access in our society is now a need because more and more official documents are sent by mail, official information is only available through internet and job offers are more and more only on internet. Now, you say that a family who can barely buy enough food to feed them should throw 29.95 a month (~1 week of food for a poor family of 4) to be able to find jobs and communicate? Get out of your mom basement and look out what it mean to being poor. Go to your nearest association who help poor people and ask them and you will see that even a 9$ internet connection is weighting a lot in some poor family budgets.
Slashdot Mirror
But the factory default contain the install of these bloatwares. And when it is Facebook that take over your phone your are skrewed :/
I did already have the opportunity to test drive a Tesla here in Belgium and liked it even if it was too short to my liking (couln't test lane keeping feature). :( At 35.000 Euros I could think about it and compare with what other automakers will have at that time (ie: If they get their 35.000 version before the other automakers get their fingers out of their collective arses, Tesla will have my money !)
But at 58.800 Euros (chepeast in Belgium), I will pass
But their are a bullet in the arsenal against bugs...
What could be that unicorn real work number crunching application that couldn't be written in 1/10 of lines of code of python in some numpy with a pinch of DAAL or tensorflow? Why most AI research who is not know to be low on the number crunching departement is done in Python?
There is even Games developped in Python (Sims 4, Eve Online, Civilisation IV, Pirates of the Caribbean Online...).
In my career, I have written my share of Assembly and C code for performance reasons when a 486 DX 400 was the must have of the shinny machines. But now, compilers and 'interpreters' can do a better work than I due to the complexity of the processors and their different caches sizes. Right now you can write some highly optimized assembly code that work at nearly the theorical speed limit of a processor with a specific cache size, but who will work like a molasse on a similar architecture with double the cache size. An interpreter could adapt its realtime compilation to the architecture it run on and even for the same algorithm to the type of data you process.
Now, if your Python program is still too slow, you can update your slowest function into cython and compile it to have a speed boost that C can get you without losing the time needed to develop everything in C and by that saving a lot of time to go to the next project.
In the beginning of my carreer, I would have used the best language for the job from a toolbelt composed of Assembly, C/C++, Fortran, Delphi, Lisp and Prolog. Now, I use most exclusively Python for my projects.
Like a woman taking a pareental leave to take care of her newborn and comming back to see that she was fired in the meantime... We know how it work :/
But not the curator when he will nedd to find some money to bail the company
when a company gets sold your data may be sold too
As long as it take more space and more weight than an IC performing an equivalent task, this will stay a nice research subject. :(
We replaced hardware radio receptor by software ones, I don't see why we would replace software neural network by hardware ones... Plus, they miss the 'plasticity' of software ones and a scratch would make it 'dead'
No, no ethic problem here. If you get sick with 'monsanto' products, they will provide you with some aspirin :)
Trigger happy cops will now be happier. The dead will not resist that caps use its fingerprints to unlock their phone... :(
Proof gathered this way should be invalidated or else cops will be more inclined to kill the suspects to access more easily their phones
Don't blame the satnav for that, but the map providers. A lot of maps miss the informations about the category of the road, max speed, limitations (height because of a bridge, max wheight for lorries...).
Why does these API have been created for? I don't really se a use case for them :( What other API are being put in without any really interesting use-case?
If there is only 1 bus every day, even if the bus is free, I will not take that bus.
I live 6km from my work, there is a bus that stop just in front of my house every hour. I still take my car to go to work. Why?
By car it take me 15-20min to go to work in the morning and 8-10min to go back in the evening.
By bus with a change for a metro it take me ~20min in the morning and between 45min and 1h15 in the evening.
Even if the bus and metro were free, I still value the time lost way higher than the price of riding with my car.
Russia seems ready to go up to 2028 if they find other countries to follow them.
https://www.space.com/36356-ru...
2 times I have ordered takeout from a 1 star michelin restaurant without any problems. Very few restaurant refuse takeout. The only starred restaurent who refused me a takeout was because it say that he doesn't want to sell food that doesn't look as good as presented on a plate.
Most of the time takeout order are a bonus for restaurants because it doesn't take a seat in their dinning room and then they can sell one more serving to a customer.
Like said before very few restaurant promote that they can do takeout because for most people, takeout mean cheap and low quality, even if it's not true. Restaurant doesn't want that idea sticking in their customers head.
When I played MtG, I bough at least 3 BOX of boosters when a new series of card were available. And then did I do some cards shopping if really I needed them. But I always started by buying a lot of boosters. When you play competitively, its a good way to not pay an arm and a leg even if it still costing you the alf of an arm :)
I have actually rewritten my algorithms (written in python) to run them in OpenCL to gain that 100x speedup. And I really need that speedup for the learning process but not for the 'playing' process.
More than 97% of the time is used in the computation of matrix product, matrix sums, applying an activation function on a vector and do some max-pooling on these same vectors. All these functions have some kind best algorithms or implementation that take advantage of cache and instruction sets. For the matrix product, I don't think I will find by myself a better algorithm than winograd. Any advancement I could put on the algorithmic side would be in the 3% left coded in python. Even if I double the speed of the rest of my program the speedup will be less than 2% in the end.
The quote I will get from the teacher is mostly dependant on how well I followed the specifications, the clarity of my code, and how my bot fare against the other student bots. Having a more complex algorithm or implementation to shave some CPU cycles here and there will only make the source code of the bot less readable for my teacher and make him raise more concern and thus more scrutinity on my code. This is why my code for the bot will not use my OpenCL implementation who is around 600 lines of OpenCL code. It use about 30 lines of numpy code and is running under 0.5s (ie: mostly instant for the end-user and in the specs for the project). This is good enough for the part of my project that I will present to the teacher.
Has AMD (ATI) video cards do some speculative execution, Spectre is most probably also a problem for them.
From there, there is some possibilities:
- They tried to let it go under the rug by not talking about it for their video card products
- They are still in the process of working around
- They don't think that problem is a problem for video card workload (non sensitive)
- They are greedy and won't lose money working on it
- They don't have the problem
- They are incompetent
Pick your best guess
Most of your programming assignments: yes. All, certainly not! I have coded some assignments that put my i7-6700HQ with 16GB of ram on its knees (salesman problem, perceptron...), running them on a 486 would be madness :( Right now, my current project is a program that could play othello reversi. I am doing it with a neural network without tree search. The trainning of the neural network is done in a similar way of Alphago Zero (self taught with reinforcement learning) and is running since 2 weeks on my GPU (Nvidia 960M). The programm I will deliver to the teacher will only use the CPU, but for trainning I use my GPU with OpenCL to get an >100x speedup for the computations. The IA right now is good enough to beat me on each game, but not yet at the level of programs like Wzebra.
Think a little about it. Does your camera is at exactly the same place? Did your camera CMOS chip have exactly the same defects as the one they use? Do you take your still images at exactly the same time? And there is so much parameters that make your easy hack not so easy. The easy hack would be to put some kind of spying apparatus behind their camera that send the information back to you, not putting another camera.
The 1-click patent is a "purchase now" button skipping the virtual basket and confirming your purchase directly...
I write code in C, PHP, Python, Pascal, VBA, Common Lisp, Prolog, Assembler (z80, 6502, x86). I can help people with their code in C#, C++, Java, Smalltalk. From a language user point of view I am a Jack of all trades, master of none. I am very reliant on the compiler/interpreter to highlight me where I made syntactic errors or passed the wrong parameters to a function or API because in that language, the order of the parameter is different than all the other languages.
But you will not get me on the evaluation of the complexity of an algorithm. I have written programs/libraries from a wide range of domains (robotics, biology equipment, domain specific compilers, pay, stock management and ordering, automatic packing of orders, display of industrial process parameters, expert system, machine learning). You can really thrust me in software design and algorithmic, but I am still very reliant on my IDE and the compiler/interpreter to help me about typos and showing me that the construct I use is not the correct one for the language I am currently using (my black sheep is the switch/case/select construct).
Some composer were bad at playing instruments like Mahler, Berlioz, Schoenberg, Ravel. Mahler and Berlioz were just good enough to play their own composition to help them in their creative process but not in public. Schoenberg and Ravel used the help of assistants to play them what they have composed.
On a Sharp PC 1500A.
Internet access in our society is now a need because more and more official documents are sent by mail, official information is only available through internet and job offers are more and more only on internet. Now, you say that a family who can barely buy enough food to feed them should throw 29.95 a month (~1 week of food for a poor family of 4) to be able to find jobs and communicate? Get out of your mom basement and look out what it mean to being poor. Go to your nearest association who help poor people and ask them and you will see that even a 9$ internet connection is weighting a lot in some poor family budgets.
A 5$ one-time reduction on their bill for checking the opt-in of this fabulous promotion !!!
It's been a week that Firefox implemented it (ie: Firefox 52)