Fact Finder
Top 20
Hi MiloSorry FF but I think point 5 in above list is either incomplete or misleading to generalise it like that.
Akida 1.0 will outperform those in some aspects but can it do everything that GPUs and CPUs do? They are made for different purposes.
For me this is like comparing motor bikes to cars to buses and saying the motor bike outperforms the other two without saying in what aspect. They all have their own strengths.
Happy for you to point to the source for me to learn more if I'm incorrect.
My opinion only.
As you are emerged you might not know that I am a self confessed technophobe and as such I do not claim any technology expertise so I rely upon the engineer/s in our midst to keep me on track and accurate.
What I do, do, in a haphazard way is remember what is said and 95% of my posts are from memory so you need to do your own research but in making point number 5. I relied upon the following:
1. The Screen shot posted of the Edge Impulse presenter standing in front of the graphic comparing AKIDA technology to GPU's and CPU's and as Brainchip had a much bigger bar by some measure he was suggesting what we have been told by Brainchip;
2. The first thing we were told by Brainchip was at the 2019 AGM presentation by Peter van der Made where he stated that 100 AKD1000 chips could provide all of the computing power necessary to achieve full autonomous driving;
3. The second thing we were told by Brainchip is that AKIDA 1.0 can process all five senses including vision as well as Lidar, Radar and Ultrasonics on chip and can do sensor fusion;
4. The third thing we were told by Anil Mankar is that others (I assume now to be Edge Impulse) were bench marking AKIDA against GPU's and it was coming up favorable to AKIDA; (Just reading your post again to make sure I did not miss anything out and it occurs to me that this statement by Anil Mankar makes your suggestion about cars, buses and motor bikes totally incorrect I think.)
5. The fourth thing we were told is that using the Meta TF platform and the CNN2SNN converter you can keep all your existing compute and run it through Meta TF and the converter and achieve all the benefits of spiking neural network processing;
6. The fifth thing we were told is that AKIDA can do on chip convolution;
7. The sixth thing we have been told at the same 2019 AGM by Peter van der Made is that AKIDA can do maths;
8. The seventh thing we have been told is that AKIDA 1.0 is capable of performing regression analysis for vibration detection; and
9. Finally the ability of AKIDA technology to process multiple video streams with state of the art performance was proven with Studio and Accelerator which both won industry awards both of which were based upon exactly the same but earlier versions of AKIDA technology.
So if I have been misleading I apologize but it was not with any intent as I have simply accepted all of these known facts.
I will leave it to the engineer/s here to take these facts and explain why point 5. is or is not misleading. Perhaps @Diogenese can assist or perhaps you can also as I am assuming you have engineering or data science qualifications though you have not stated so.
My opinion only DYOR
FF
AKIDA BALLISTA
Last edited: