Wednesday, March 21, 2012

Science Talk


Listening to Stephen Hawking Talk Science

On November 12, 2005, Stephen Hawking gave a lecture in San Francisco on the birth of the universe. I felt lucky to be in the audience. Though not trained in physics or astronomy, I was eager to hear what Hawking had to say. Cosmology intrigued me, my curiosity having been sparked by the likes of polymath Carl Sagan and science writer Nigel Calder, whose The Key to the Universe I’d found fascinating and fathomable (it was a key, after all). About the Hawking event I remember the lecture room, approximate size of the audience, disposition of the chairs, where I sat and with whom, Hawking’s physical condition, Hawking’s delivery, the duration of the talk, the reception afterwards, and a few conversations I had with people at the reception. Of Hawking’s ideas, however, I remember nothing.

What I recall about Hawking’s talk is the experience of listening to Stephen Hawking. Embarkation was smooth, and the audience exhibited high spirits. Hawking’s digitized voice may have disconcerted at first, but we soon became habituated. I felt like a child who is learning to read, and sails through the first simple sentences of her primer. Of the first 10% of Hawking’s talk – call it the Prologue – I believe I understood every word. Beyond the Prologue, however, I began to struggle. Again like a reader who has not attained fluency, I had moments of successful decoding and comprehension, interspersed with increasingly longer moments in which decoding failed, or failed to yield comprehension. Struggle lasted through perhaps the next 15% of the talk. Struggle at an end, for the next 10% of the talk I floundered. Brief flashes of apparent comprehension were engulfed by waves of disorientation, confusion. Respite came during the next 5% of the talk when knowledge of my limitations surfaced and I recognized Truth: I was completely out of my depth. For the remaining 60% of Stephen Hawking’s cosmology lecture, I was not decoding, I comprehended nothing, I was no longer reading the thoughts of a genius. Instead, my mind wandered, I daydreamed, and I wondered idly about such things as the complicated logistics of being Stephen Hawking.

That experience has intrigued me ever since. Similarly, I’m intrigued by the not infrequent experience of listening to a talk or reading a text of which, in the course of listening or reading, I believe I have full understanding, only to discover, while describing the talk or the text to someone, how partial my understanding really is. Regarding the Hawking talk, I could recap a few ideas from the early part of the lecture during conversations I joined in afterwards. Seven years later, I can’t recap any of those ideas. Did I understand something of what I heard that day, and if so, in what sense?  Did I gain knowledge of Stephen Hawking’s universe, or did I merely recognize ideas I had heard before? And what to make of the pattern of my experience of understanding, from smooth comprehension, to struggle, to floundering, to the near-complete dispersion of my attention?

What did I know and when did I know it?

Can this uncertainty about the acquisition, retention, and recapitulation of knowledge be resolved with reference to the properties of consciousness, or more precisely, to aspects of attention?

Did my difficulty in understanding the Hawking lecture reside in an attention deficit? Was it, in effect, a matter of my not having heeded an internal instruction to pay attention?

Of course not.

The insufficiency of the attention deficit explanation is made clear by posing the question: What would best enable a listener to achieve a state of near-complete attention while listening to Hawking lecture? The question can be answered by listing the necessary prerequisites for optimal attention:

·      Relevant background knowledge/ memories
·      Reliable cognitive processes
·      Extraordinary intelligence/ high IQ
·      Ability to identify and appreciate new and novel patterns

Possessing the above attributes, a listener/ learner could indeed pay attention to nearly every word describing Stephen Hawking’s universe.

Attention, then, is conditional upon the possession of, among other things, prior knowledge. Where does this prior knowledge come from? What process of concept formation leads to our being able to think about and comprehend a complex subject like cosmology?

What do we know and when do we know it?

It seems to me that what we know depends crucially upon the time during which we know it. Knowing depends on memory. To help me clarify my understanding of knowledge and memory, I turned to the Stanford Encyclopedia of Philosophy (a high-level handbook, the encyclopedia offered reassurance and information, not unlike Calder’s Key). Unsurprisingly, I was not the first to associate knowledge with memory. But it turned out that most philosophers who had pondered memory in its relation to knowledge had done their pondering before the modern era. Hence: “The twentieth century saw rather little work on the epistemology of memory.” (Senor, Thomas D., Epistemological Problems of Memory, 2009, p. 12)

Senor notes the extent to which even our current conscious knowledge depends on memory. Our ability to process an immediate impression like ‘the sunset is beautiful’ depends on background knowledge, or background memories, such as that we are looking west and that it is evening. While 20th century philosophers gave little thought to memory’s contribution to knowledge, researchers in psychology and cognitive science probed the function and role of memory in numerous contexts. A current research project on concept acquisition and short-term episodic memory describes the neural pathway connecting knowledge to memory (Hassabis, Demis, Combining Systems Neuroscience and Machine Learning to Build AGI, Future of Humanity Institute, 2011). What about high-level abstract concepts like Time, and scientific concepts like the Big Bang? In what ways might our understanding of these ideas be associated with short- and long-term memory?

John Locke, in An Essay Concerning Human Understanding, describes memory as a “storehouse” of ideas. If I accept this metaphor, I might assume that the immediate issues relevant to knowledge concern what ideas are present in the storehouse at any given time, and how I access those ideas. Additionally, I might want to assume a process of concept acquisition – a way of learning new ideas.

What kinds of ideas do we learn? Some of the abstract concepts we learn very early on are perceptual and kinesthetic, like location and place. Later, we acquire high-level abstract ideas like Right and Wrong. Bertrand Russell’s concept of ideas in Analysis of Mind includes ideas as images as well as ideas as propositions. (Senor, p. 6)

Although the 20th century produced little philosophical thinking about knowledge and memory, epistemologists did develop the concept of memory belief. Memory belief is the memory of an event or the memory of a proposition about an event. Propositional memory might also be based on something that is not an event: on reasoning and analysis. “The objects of propositional memory are propositions; the objects of event memory are events.” (Senor, p. 14) Thus high-level abstract concepts are part of dynamic processes such as analysis. Alas, modern philosophy of knowledge has more or less disintegrated the graphic "storehouse" metaphor; if knowledge is dynamic, it becomes less useful to think about concepts carefully stacked away in a repository.

What kinds of memory beliefs are involved in learning? Learning to read, a near-universal experience in the West, is a complex learning project spanning many years. Learning to read requires memory beliefs that are propositional [background memories of the alphabet, phonics, spelling, the decoding process, and semantics] and event-based [experiences over time of being read to, reading aloud, decoding, subvocalizing, etc.] As human learners, we rely on our experiences in the world as well as our reasoning to gain knowledge.

What happens when we transpose knowledge and memory into machine learning systems? Does the learning capacity of AI come primarily from propositional memory beliefs? A proposition is an expression in language – or signs – of something that can be believed, doubted, or denied or is either true or false. In addition to computer learning systems' reliance on code, certain machines rely on a type of event-based knowledge. These self-improving AI operate in such a way as to achieve their goals while preserving their core utility functions (Omohundro, Stephen, The Nature of Self-Improving Artificial Intelligence, 2007) Self-improving AI might acquire new information from experience and self-improve through access to their source codes.

To our understanding of the limits of knowledge, epistemologists have added the idea of the ambiguity response. (Steup, Matthias, Epistemology, Stanford Encyclopedia of Philosophy, 2005, Section 5.6) The ambiguity response asks whether, when I say I “know” something, I am referring to a high-standards concept ({my} knowledge is infallible) or to a low-standards concept ({my} knowledge is fallible). Even by the low-standards criterion, the assertion that I “know” or “knew” Stephen Hawking’s ideas about Time and the universe from listening to his talk is not credible.

Listening to Stephen Hawking talk about the universe, I wondered if a little knowledge, while not a dangerous thing, might yet be useful. On the assumption that my lack of comprehension was due in part to a deficit in background knowledge and scientific thinking, I described my trajectory from focus to dispersion to a computer scientist who had been in the audience. The computer scientist promptly told me that his understanding of Hawking had followed a similar trajectory, even though he had the necessary background knowledge. It’s a reasonable assumption that, as a scientist, he was also in possession of reliable cognitive processes and a high-functioning intellect. He told me in fact that he had easily comprehended the first third of the talk, struggled through the next third, and abandoned all hope in the final third.   

Did the computer scientist possess the gift of recognizing novel patterns and appreciating their potential meaning? He told me about reading A Brief History of Time. He hadn’t written the book, but he could recount it. What did he know of the universe, and when did he know it?  Bridging the chasm between Hawking’s original theory and my limited grasp of it, the computer scientist knew enough at that moment to advance my understanding, too. 

No comments: