Cognitive computing: IBM uses phase-change material to model your brain's neurons IBM scientists claimed – for the first time – to have created artificial spiking neurons using a phase-change material, opening up the possibilities of building a neural network that could be used for AI. The brain is the biggest inspiration for researchers working in cognitive computing: the exact mechanisms that describe how …
This post has been deleted by its author
That's 10^8 of 10^-9 seconds? Most of us would just say a tenth of a second, rather than adding needless complexity.
Not quite. It is to indicate the least significant digit and tolerance (sorry, I learned this in a different language so I may not use the right English terms for it).
If you express 0,1 as 0,1000, it means your measurement can be anywhere between 0,0999 and 0,1001. If you express it as 0,1s, it means "somewhere between 0,0 and 0,2", which is different.
The key to learning is in the identification of correlates.and then the 'preservation' of the ability to match further correlates to those already found... and cross-correlating correlates so as to build up a multi-dimensional 'correlation web' (aka. database?) of past experiences. Outside of ICT, in my other field of interest, of behavioural science, we might see in this 'correlation web' the development of the norms and values by which organisms arbitrate external stimulii into actions by applying a recursive process of C-D transforms (aka if-then-else decisions... broadly speaking) through which the stimulii, or a set of stimulii as the neural net becomes more sophisticated in its operation, are effectively assessed against a record of action outcomes as to which is advised in this case.
Interestingly it is the only way we know of for handling the never ending optimisation problems that every organism faces, sometimes thousands of times a day. Mathematically they are generally NP-hard, and cannot be reasonably computed or cannot be solved at all. We daily use Gantt Charts to solve scheduling challenges for example... to make a solvable problem computable in real time while giving us the option of accepting a sub-optimal 'solution' when it is 'good enough'.
From this perspective... just the idea of how the synthetic neurons filter and identify correlating signals in a physical 'mechanistic' analogue to the mathematical identification of correlate points in a neural network has to be seen as incredibly important. It has the potential to lead to the design of many physical alternatives to the synthetic neuron described by the IBM team and a host of optimisation capabilities, simplified to be suited to specific target tasks within the design goals, of a whole range of specialised AI-enabled tools.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
