Turing is dead, long live Turing!


God!  It had to happen.  Life becomes Science (as opposed to Art).

An Israeli mob is working with AI to provide a coding assistant that does more than complete function names – it is to scan Stack Overflow for code snippet suggestions!

I wonder if they are going to have it write sardonic blog entries as it discovers, over and over again, that the code examples at Stack Overflow do not work for the most part.  Or at least they are so narrow a “solution” they cannot be generalised.

What, by the way, does the AI do in my situation where perfectly good init.d script runs in console, does not run in service startup, and there is no solution provided (currently) on the Internet?

Sort of also reminds me of the AI goons who, at a party (and in a very Hitch Hikers Guide to the Galaxy kind of way), connected two AI conversation agents together.  The AI agents had been “trained” by hooking them into social websites (twitter, facebook et cetera) – to “train” them on human conversational style.  Within 5 minutes there was a tirade of abuse going both ways.  The AI scientists had “trained” Trolls.

Now, here comes the deep and philosophical.

The trend to “free” software has incurred a huge technical debt through the utter lack of software quality attributes.  I am, for example with the problems setting up my OPiZ for node-red and emqttd, suffering at the hands of lack of portability, usability and a smidgen of functionality.

What I am going through is old skool Through Life Blues.

The cost of Through Life aspects of software systems is about 80% of the full life cycle cost.  That is we spend 20% on development then a shit load more on keeping it going.

Much of that cost comes from un-codified knowledge.

So, the propensity to write code, comments are a “code smell”, produce no documentation, all of that increases the cost of Through Life Support.  Though the cost is in time an efficiency for me, it must be burning someone’s bottom line in the commercial sectors.

So, since there is no code comments, design documents or useful manuals the question begged is should there be?  The AI push for recovery of codified knowledge says yes since we are not denying its necessity.   However, we still feel the generation of the technical debt incurred by developers, but repaid by everyone downstream, is acceptable – somehow.

The problem is, of course, a quality attribute Adoptability (not in the ISO standard).  Adoptability is not to be confused with Attractiveness.  Adoptability might the decision point, based on the aggregation of all the facets of the software quality.

So, where is the benefit of AI trawling the coffers of sloppy code?

Where is the cost trade off against the waste going on versus the relatively small number of successes in the scheme of things?

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: