It's so hard to produce anything with neural networks, and the result is always so static (can't be retrained easily).
My personal alternative to NN is slowly taking shape. Pretty much, these are the components: Coarse procedures <http://code.botcompany.de/1023890> Genetic optimization Natural language Human help Basically we make *human-understandable procedures* which also optimize themselves. Like, if there is an image recognition task, a human throws in a suggestion like "look for the round red thing on th left", and from this coarse description, a recognition procedure is quickly learned (few seconds). On Fri, 6 Sep 2019 at 03:15, Secretary of Trades <costi.dumitre...@gmx.com> wrote: > or CV intelligence at 100+FPS > > https://davheld.github.io/GOTURN/GOTURN.pdf > > Grant: Toyota (automotive) > Political unit: SAIL (academic) -- Stefan Reich BotCompany.de // Java-based operating systems ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tc2213439607535e8-M7bf1d6af00b63ec5500bd426 Delivery options: https://agi.topicbox.com/groups/agi/subscription
