[Oberon] Accessibility of FPGA configuration to V5 system.

Skulski, Wojciech skulski at pas.rochester.edu
Mon Aug 30 16:47:14 CEST 2021


This is indeed a side trip not quite aligned with the topic. Let me chime in nevertheless. The man made neural nets are composed of connections and weights (coefficients). The connections are fixed. The weights are adjusted to the situation by training with known patterns. The trained net can then recognize arriving patterns according to the similarity to the patterns which were used for training. Nobody can make sense of the coefficients themselves or generate the coefficients according to some theory. Training is all that we can do to get these working sets of coefficients. Presumably, the natural neural nets work in a similar way, but the natural nets can grow new connections, which the artificial nets cannot do.

In my humble opinion, part of the excitement was generated when the unassuming name "neural nets" was changed to a new exciting "artificial intelligence". I have a nagging impression that publicity invested into AI is motivated by hopes that abundance of artificial intelligence will substitute for lack of natural intelligence in the populace. I have read claims that electing politicians using the AI rather than with voting will decrease the number of morons getting to the top. I am not making it up. Some highly educated people are making such claims. IMHO, those who put such hopes into the AI should rather work harder to improve their own natural version of it. 

I am not sure how all this is related to the V5 system configuration.

W


More information about the Oberon mailing list