Artificial Intelligence

Out of context: Reply #478

  • Started
  • Last post
  • 1,341 Responses
  • Nairn3

    ^
    I was talking to someone who's going to Denmark to work on his PhD which revolved around using an AI network to analyse genomes to work out what the markers are for susceptibility to various types of cancer. Very interesting, and saved my Christmas lunch as it was a topic I actually had an interest in and was able to feign competence in enough to make his side of the conversation not utterly painful or in turn condescending.

    Suffice to say, it's currently not a million miles away from how the sort of plebian 'AI' we've been tinkering with is - wherein it analyses existing datasets and tries to order them based on queries which are made and then checked by humans.

    It's very clever stuff, but what it is not and necessarily cannot be (yet), is a magic black box that can take ALL inputs (ie. everything) and somehow respond to queries that it was not specifically trained on. ie. It can't (yet) generate truly novel information. For current purposes, that doesn't much matter, as it's akin to magic, but for truly novel discovery, we're a way off yet. The gist of it being, we'd need to develop gAI AND then understand how the resultant 'brain' works. Given we don't understand how even our brain works.. well.

    Oh, and the annoying thing about his research - even if they discover all the cancer markers and check you for them.. all they can then do is work towards non-genome-based prevention. It's not like we can re-write our genomes in any fundamental sense to make ourselves immune yet, irrespective of how cool CRISPR can be.

View thread