The very first gathering for AI Intel took place all the way back in 1956 in Dartmouth; experts present there all surmised that the working of the human brain could be defined, and understood using computational theory. So, all developments that could take would lead back to AGI.
Not much progress has been made on this theory since, and now people are looking for other routes to achieve this same thing. In neuro-morphic computing, there’s this theory that computer can be used to get a replica of the human brain, and in this sense can figure out how it operates. But, most of these trials have to lead to error and ultimately failure, and as a result, we’re nowhere close to the perfect AGI.
The problem is that we don’t as yet have any real basis to suggest the proper building of something equivalent to a brain, using just some neural networking component. So, Artificial Intelligence Insight Series - Neuromorphic Chipsets, NMC can’t be seen as the first step to the AGI development – we need to understand it and only then can we even begin to create the software used for operating it and then the hardware.
Today companies involved in computer accelerator work that have built similar chips that are biologically inspired. Some define it as a machine learning accelerators – but both are capable of carrying out basic linear algebra at faster speeds than your average CPU, by making use of SIMD (single instruction, multiple data) architecture.
But at the end of the day, they are just computer accelerators – and to get an AGI, researchers still have to work out the theory and science behind it fully. And this part requires intensive research encompassing all areas connected to human intelligence and a range of other disciplines before we come up with workable theories, and then start refining them into combined ones.
Not much progress has been made on this theory since, and now people are looking for other routes to achieve this same thing. In neuro-morphic computing, there’s this theory that computer can be used to get a replica of the human brain, and in this sense can figure out how it operates. But, most of these trials have to lead to error and ultimately failure, and as a result, we’re nowhere close to the perfect AGI.
The problem is that we don’t as yet have any real basis to suggest the proper building of something equivalent to a brain, using just some neural networking component. So, Artificial Intelligence Insight Series - Neuromorphic Chipsets, NMC can’t be seen as the first step to the AGI development – we need to understand it and only then can we even begin to create the software used for operating it and then the hardware.
Today companies involved in computer accelerator work that have built similar chips that are biologically inspired. Some define it as a machine learning accelerators – but both are capable of carrying out basic linear algebra at faster speeds than your average CPU, by making use of SIMD (single instruction, multiple data) architecture.
But at the end of the day, they are just computer accelerators – and to get an AGI, researchers still have to work out the theory and science behind it fully. And this part requires intensive research encompassing all areas connected to human intelligence and a range of other disciplines before we come up with workable theories, and then start refining them into combined ones.
Browse Complete Country Report and Detailed TOC: https://www.researchonglobalmarkets.com/ai-insight-series-neuromorphic-chipsets.html
No comments:
Post a Comment