how_to_start_up_a_startup
Differences
This shows you the differences between two versions of the page.
| Both sides previous revision Previous revision | Next revision Both sides next revision | ||
|
how_to_start_up_a_startup [2020/04/27 06:56] admin |
how_to_start_up_a_startup [2020/04/27 07:06] admin |
||
|---|---|---|---|
| Line 76: | Line 76: | ||
| * Do the right thing, act responsibly and with integrity. Your reputation is more important than your current startup. | * Do the right thing, act responsibly and with integrity. Your reputation is more important than your current startup. | ||
| - | === Example: Neuracore.ai === | + | ==== Example: Neuracore.ai ==== |
| - | ==== What is the company going to do? ==== | + | Neuracore was incorporated on 12 December 2018 and dissolved on 30 April 2019. |
| - | Design and license cores for the efficient execution of neural networks. The efficiently is obtained by having vector and matrix "registers" and hardware to perform the vector * matrix = vector operations (plus associated, such as addition, ReLu, sigmiod and softmax on vector). The engine does complete stochastic gradient descent, not just inference. | + | === What is the company going to do? === |
| - | ==== Why is it unique? ==== | + | Design and license cores for the efficient execution of neural networks. This will enable "AI Everywhere". |
| - | The market is dominated by GPUs with FPGA upcoming. Totally analog (neurpmophic) has been on the horizon for a while but can't be trained. Neuracore will both be trainable with standard software (tensorflow, pyTorch) and achieve massive improvement in TFLOP/watt - it will be the most power efficient NN hardware available. | + | === Why is it unique? === |
| - | ==== How is it going to be successful? ==== | + | Extreme power efficiency efficiently is obtained by a support software enabling low precision integer training and single propagation delay addition. |
| - | PoC - simulation (SPICE) | + | === How is it going to be successful? === |
| - | Patent protection of key ideas | + | |
| - | Partner with ARM, get them to fund joint work | + | |
| - | Sell to ARM, nVidia, Intel for mass production | + | |
| - | ==== Draft a quick business plan so that you have a story to tell others ==== | + | Partner with ARM, get them to fund joint work, use them as first revenues. |
| + | |||
| + | === Draft a quick business plan so that you have a story to tell others === | ||
| Hardware acceleration for Neural Nets is already huge, the whole current wave of Deep Learning happened because GPUs became cheap enough. Google have enough services that need NNs to build their own ASIC, the TPU1. | Hardware acceleration for Neural Nets is already huge, the whole current wave of Deep Learning happened because GPUs became cheap enough. Google have enough services that need NNs to build their own ASIC, the TPU1. | ||
| - | Stage 0: Come up with a reasonable hardware design | + | * Stage 0: Come up with a reasonable hardware design |
| - | + | * Stage 1: DO PATENT REVIEW then Get team together | |
| - | Stage 1: DO PATENT REVIEW then Get team together | + | * Stage 2+ as previous section |
| - | + | ||
| - | Stage 2+ as previous section | + | |
| Aim for getting it out there in 5 years - any sooner and FPGA will dominate, any later and too much risk. | Aim for getting it out there in 5 years - any sooner and FPGA will dominate, any later and too much risk. | ||
| Line 107: | Line 104: | ||
| UseAI index 2018 annual report for evidence of ai gold rush. Neuracore sells the shovels "During the gold rush its a good time to be in the pick and shovel business" Mark Twain | UseAI index 2018 annual report for evidence of ai gold rush. Neuracore sells the shovels "During the gold rush its a good time to be in the pick and shovel business" Mark Twain | ||
| - | ==== Competitors ==== | + | === Competitors === |
| From: https://www.engineering.com/Hardware/ArticleID/16753/The-Great-Debate-of-AI-Architecture.aspx | From: https://www.engineering.com/Hardware/ArticleID/16753/The-Great-Debate-of-AI-Architecture.aspx | ||