Huge fan of all the work you do. Wanted to understand something fundamental and whom better to ask than you: Whats so special about the transformer architecture that its able to predict the next token so beautifully understanding all the intricate previous token relationships? I understand Attention but what so special about this architecture that no other architectures are able to "attend" appropriately to previous tokens? Being a CS guy, its really hard for me to fathom that we have not yet created another architecture which can perform similarly.
From the white paper: "If additional shadow headers are used, their suggested locations are (where possible) at the sectors starting at offsets 128MiB, 128GiB, and 128TiB"
Is such repetition of header (in a supposedly random-wiped disk) not suggestive of existence of a PUREE partition? This can be figured by scanning the entire disk.
We will soon add more pages explaining how we are doing the encryption. Few details are already mentioned in the homepage as of now. Essentially we take every 64 kb of your file and encrypt it with AES-256 encryption algorithm with GCM authentication using your VyomKey as the encryption key. We haven't rolled out any new encryption algo. This is just a start, will add more pages as we go forward. Our main USP is that we use an encryption key of user's own choice - the VyomKey - to encrypt the files which is not there in any other competing product.
Thanks bigiain for the suggestion. Just mentioned so since Dropbox is in the news. We are essentially an alternative to Dropbox, Google Drive, Microsoft OneDrive.
We will be adding lot more information in the coming days. We will publish a white paper on the backend tech to be completely transparent on how data is being handled. Please bear with us. Please do signup so we can send you early invite.
As mentioned in the other comment, cloud may be involved if personal machines are behind firewall/NAT (all data to cloud will be encrypted and passed through with no persistence what so ever). Although the product tries to punch holes with UPnP, it may not always be successful. The other reason is for us to make money. Any alternate monetisation ideas?
The product enables you create a seamless personal cloud among your own computers. Although there are no servers involved for storing the data, there are servers involved for enabling seamless connection between your computers when they are behind firewall/NAT.
I think there is more pessimism than required here. We need not start with ATC directly sending instructions to the iPad thats controlling the aircraft (although something similar happens in UAVs). We can start with it displaying the required instructions along with the HF radio. Also the final judgement of obeying an instruction is with the pilot and he will responsible for the results.
In the era of having artificial hearts and pacemakers that are helping better lives, I don't think its as hard to even build sound and safe automated flying systems.
> In my experience, if you love technology and tinkering, you should absolutely not work at a startup.
This need not be essentially true. Although delivering results is of highest priority, delivering "right" results is even more important, results that create your USP, results that set you way ahead of your competitors. And that requires right amount tinkering, if you want to call it that. And that tinkering is a lot of fun - to deliver world class results in the least amount of time possible.
I think this is way too direct. You atleast need to qualify if they really have the PROBLEM (even if you know they have the PROBLEM, it will atleast involve him in the conversation). Probably after that you can explain very briefly how you can solve the PROBLEM and give out few good names from the current clientele that have solved same/similar PROBLEM and then ask for a demo day. No?
Huge fan of all the work you do. Wanted to understand something fundamental and whom better to ask than you: Whats so special about the transformer architecture that its able to predict the next token so beautifully understanding all the intricate previous token relationships? I understand Attention but what so special about this architecture that no other architectures are able to "attend" appropriately to previous tokens? Being a CS guy, its really hard for me to fathom that we have not yet created another architecture which can perform similarly.