În Germany it might stand, but you are forgetting the American legal system is very much moneybased, so Amazon can just fight it in court an win, or at worst settle, which I have no doubt it's overall still a win
Companies were never supposed to pay taxes on their profits, they were supposed pay dividends which are then taxed as income. The desire to tax that profit twice (or 3 times if you count consumption tax) is what creates this whole mess, because you fundamentally can’t force a company to make a profit.
The outcome of trying is a tax system that only applies to SME.
I concur. I'm not a coder but I do review things every 2 months as part of the release cycle, and whenever I see mistakes, especially silly mistakes, I don't feel smart at all, it doesn't even cross my mind.
I feel, in order: disappointment - because how could they make this stupid mistake, then anger - why the f do I have to waste time to correct stupid mistakes, then a bit hateful - we need to fire this person and hire someone competent, then sometimes a bit scared - what else has this person got wrong and I missed, then a mix of all of the above.
What would make me feel good, as far as reviews go, would be to say "hey, this person did this way better than I would have".
Pasting this reply I found elsewhere. Should be accurate to my knowledge.
Yes to Slack! Ask whoever manages your organization to enable XMPP or IRC gateway connection, then add the appropriate type of account to Trillian. You won't have access to rich cards or shared files from within Trillian, but you will be able to read and send messages to channels and individuals.
Doesn't sound too bad. But can you elaborate on the last part, about retrieval requiring memory and not scaling to more than a couple of thousands images?
How much memory would you need for ~2000 images, how slow does it get, etc.
Retrieval using CNNs requires computing a cosine similarity matrix. So, for 'n' images, a matrix of size n x n would need to be stored in the memory. As you can see, the storage requirements blow up quadratically as 'n' increases. We have already made some optimizations to reduce the memory footprint but there would be clear upper limits to it (We haven't experimented). As for the numbers, the cifar10 dataset example present in the repo was carried out on google colab notebook. The dataset has 60k images and ended up consuming about 6G of RAM using CNN. However, since there hasn't yet been a principled benchmarking done on the package, I would consider these numbers as only marginally indicative of the capabilities. A better way to figure out if it works for you would be to try it out with your own dataset, starting out at a small scale and increasing the dataset size gradually.
"The tricky part is the retrieval of duplicates, which on the same 10K dataset takes a few minutes"
"Doesn't sound too bad."
I doesn't? As a word of caution: If a 10k dataset takes a few minutes I would be careful how long 10MM pictures take. I predict it does not take 10MM/10k times a few minutes.
