You’ve curated the sources. You’ve researched everything. You know exactly what you want to say.
You just can't get it out of your head and onto the page. We are building Almanac exactly for this.
Experience the beta version here: https://t.co/F7Q7Pfgz7y
Here are a few things you can do with Almanac👇
@flipkartsupport@Flipkart 2/2
2. no email was sent about cancelled return
3. return was cancelled incorrectly. I called customer cares many times & they say yes their mistake and not mistake but still can't return. U didn't WORST CUSTOMER CARE. SCAMMING PEOPLE for money. #worst
@flipkartsupport@Flipkart 1/2 i raised return for dining table, technician visited, validated with defect/damage & return was approved. Now return got cancelled & denying refund even though u did 3 mistakes - 1. return status changed from approved to cancelled, who does that.
@flipkartsupport@Flipkart i raised return for dining table & it was approved. Now return cancelled & denying refund even though u did 3 mistakes - return status changed from approved to cancelled, no cancel email and incorrect return cancel. WORST CUSTOMER CARE. SCAMMING PEOPLE
No nation should have to accept terrorist attacks being launched against it from land controlled by another country.
India is justified in striking terrorist infrastructure. There can be no impunity for terrorists.
Mayura is our new translation model available as API from today.
It supports 10 languages: Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Oriya, Punjabi, Tamil, and Telugu.
Beginnings of a new semicon industry in India!
P.S. InCore has only one investor- Peak XV (formerly Sequoia). InCore is a one-stop shop for all your silicon needs. To know more, please visit https://t.co/UlBCo7WyZl
The analogy between GPTs of today to the CPUs of early days of computing are interesting. GPT is a funny kind of programmable text computer. Have to think through it more 🤔 but e.g.:
## Memory
GPT-4 RAM is ~log2(50K vocab size)*(32K context length)/(8 bits/byte) ~= 64kB, roughly a Commodore64. Just as then, optimizing this precious resource is critical.
GPT registers are the residual stream. There are d_model of them, e.g. GPT-3 has ~12K registers. VLIW architecture vibes.
## CPU
The LOAD instruction is the Attention mechanism, except it can address by both location and/or content.
The STORE instruction is forced every n_layer number of clock cycles.
The ALU are the MLPs + LayerNorms. Awkwardly, as their params are not shared across layers, the ALU changes at each clock cycle. Optionally the MLPs may also be interpreted as supporting a kind of fixed knowledge database lookup.
The programs always takes the form [[LOAD, ALU]*N, STORE]*M, where N is n_layer and M is num_tokens.
## Architecture
GPT feels closer to a fixed-function than stored-program computer because the number of parameters is so large. In contrast, the description length of a CPU is very low and all the action is in the memory configuration.
Another way to look at it is that GPT is a much more bloated/complex computer. Which is fine because it is not engineered but optimized and the upshot is that the programs can be shorter.
Yesterday I ordered from zomato, And one of the item was missing, when i raised the complaint to zomato customer care support, their response was restuarant was not responding so i have to check with them today, so i mailed again, my refund was rejected @zomatocare@zomato