The Way AI Behaves

[bnewman]

In "Computing Machinery and Intelligence", Alan Turing suggested that we will know whether machines can think, roughly speaking, by the way they behave — just as, in "Prince Lir's Song", the prince will know when he loves "by the way I behave".  Thus begins a brief survey of the history and philosophy of machine intelligence.

lyrics by Benjamin Newman, after "Prince Lir's Song" by Peter S. Beagle

ttto: "The Ash Grove", trad.

 

/: D - Em A / D G A D :/

 

When I was a student of cognitive science,

I studied the classics, and studied them well,

Among them, of course, Turing's seminal paper:

"Machines, can they think, and just how could we tell?"

 

/ " / Bm - E7 A / " / " /

 

For this was the insight that got the field started,

A thesis so simple, yet shockingly brave:

"The only true test is to watch them at work,

And we'll know if they think by the way they behave."

 

The first pioneers, like McCarthy and Minsky,

Laid groundwork from which all AI would evolve,

But, in spite of the efforts of Newell and Simon,

The general problem was never quite solved.

 

For since then AI has produced machine experts,

Whose "thoughts" run quite well on the paths that we pave,

But just don't apply to more general cases,

Which shows they don't think in the manner we crave.

 

Robotics has some asking "Can we still learn from

Abstract toy domains such as blocks-world and chess?"

But Winograd's research, and Hofstadter's later

Suggest that the answer's a qualified "Yes".

 

Analogy making and pattern detection

Are central to thought even stuck in a cave;

Our models, though toys, will face intricate tasks,

And we'll know if they think by the way they behave.

 

We still hear from Nagel and Searle, contra Dennett,

That thought and computing are different in kind;

And some, such as Penrose, still hope to unearth

Some unheard-of new force at the heart of the mind.

 

But we who pursue artificial intelligence

Heed not that grim philosophical sphinx:

No more must we seek some mysterious spark,

But must build a machine that behaves like it thinks!

 

Consider, as well, the imagined creations

Of authors like Asimov, Capek, and Lem;

The thinking machines in most fiction are evil,

But I worry more what we might do to them.

 

And so, in conclusion, I offer this warning

To makers of future mechanical slaves:

Will they willingly serve?  What rights will they deserve,

When we know that they think by the way they behave?