The mysterious 98%: Scientists look to shine light on the 'dark genome'

  • I think it's mainly a very complex translation process for 1) useful functions, 2) frameworks for patterns, geometry, compression, etc. 3) shared abstract memories shared at the species level or a sub scope of some sort 4) check systems 5) a merging system that's specifically adapted for the potential of mutations and maybe tries to harness "mutational noise" 6) ... I could think about this more but I got work to do.

    What do you all think?

    I wonder if A.I. systems & genetic algorithms incorporate this today.
  • Well, all "dark genome" (a terrible term in and of itself) means is that those genes don't make protein(s). Current research points towards these being regulatory genes that control the expression of other genes. Mapping these portions is an ongoing task, however. 

    If I'm interpreting your statements correctly, then point 5 happens with all DNA regardless of function, though "harness" is a stretch. It'd be more accurate to say genes randomly change all the time: if it's not deleterious, it stays in the population. Points 2 and 3 sound ripped out of a Deepak Chopra book; I'm not sure what they mean, and I'm unsure anyone else does.

    AI and genetic algorithms are two entirely different ball games with totally independent rules and goals from both biology and each other. There is no correlation between those concepts and regulatory genes. 
  • I say "harness"  just like we harness randomness for our genetic algorithms.

    It is a stretch. it's assuming that our codebase actually has segments that are intentionally & intelligently/strategically sensitive to randomness.

    I disagree with you. There's a heavy inspiration from nature for what we call "genetic algorithms". And even a heavy influence from nature for how we came up with neuro networks.
  • In that case, it kind of harnesses random mutation, but it does so without any order whatsoever. Any base pair or length of base pairs can be randomly altered during replication. There's no intelligence or strategy behind any of it.  

    While genetic algorithms and neural networks have biological origins, neither are good models for the natural world. Genetic algorithms are designed to solve optimization problems such that one arrives at the best possible solution. Regular genetics and selection doesn't care about optimization at all; instead only what works matters. 

    Neural networks share the same distinction, but the divergence instead comes down to the function of individual neurons. A neural network's simulated neuron is useful for determining a specific value or set of values and then passing it along to the next layer for further processing until you have an output. A biological neuron by contrast carries signals, yes, but do so in a more nuanced and flexible manner that allows singular neurons to handle multiple types of information.