Linda clinging to a funny-looking guy
Stephan Hawking is a smart guy. You know who he is: the Cambridge professor
of physics, cosmology and all things difficult who has been confined to a
wheelchair most of his life by ALS (Lou Gehrig’s disease) and now communicates,
somehow, through a machine. Hawking has
warned strongly against AI (artificial intelligence), sometimes known as
machine learning. This seems to be a
method for computers to “educate” themselves; learn new tricks by observing the
world around them. Hawking believes that this may allow machines, which
after all are amoral fabrications made from inorganic stuff, to take over the
world and dispense with humans. I tend
(somewhat) to agree; after all, I remember (Terminator)
when Skynet became “self-aware” and initiated a war that loosed the likes of
Arnold Schwarzennerg on humanity. There
is a (rather bad) book, The Fear Index by
Robert Harris, which explores that same theme.
I tend to believe that considerable caution should accompany any attempt
to facilitate computer self-education.
So, what does any of that have to do with cancer? Well, the Director’s Blog by Francis Collins
(NIH) recently featured an article
titled “Using machine learning to understand genome function”:
about a smart mathematically inclined biologist, or perhaps
more accurately biologically-inclined mathematician at Stanford, who is using
machine learning to understand how the genome works. His name is Anshul Kundaje, and he is
originally from India. I don’t fully
understand the article, but it appears that Dr. Kundaje is applying machine
learning to identify significant patterns in huge genomic (and/or epigenomic)
data sets. For instance, artificial intelligence
already has been used to identify patterns in brain scans that predict whether
a child will develop autism. AI also has
been used to identify the abnormalities associated with Alzheimer’s disease.
Genomics and epigenomics involve huge data sets, far larger
than our little brains can wrestle with efficiently, so I guess it makes sense
to sic self-educating computers on them.
But, seriously, I think we ought to be cautious in doing so. I don’t want the Terminator walking into my
living room and solving all my health problems, permanently.
No comments:
Post a Comment