Follow the links therein to read the three subsequent parts, or click here: two, three, four.
 |
| Rodney Brooks |
I fleetingly met Rodney Brooks once, at an AI conference in London. I guess this would be in the 1980s when he was already rather famous for his controversial 'subsumption architecture'. He was a small guy who reminded me of Paul Simon. He politely asked me what I did and I pompously replied that I was an AI theoretician. He looked at me as I would have looked at someone from, say, Andorra who claimed to be an 'AI theoretician'.
The eminent roboticist gracefully made his apologies and moved on.
My subsequent encounter with Rodney Brooks was indirectly via our purchase of several vacuum cleaners from his highly successful company, iRobot. They were excellent.
The essays too are uniformly excellent.
Nigel,
ReplyDeleteThese were interesting and good posts from Rodney Brooks (modulo a few typos). The best parts were on his analysis of the history of AI, and clearly he is acting to subdue enthusiasm for ANNs as a panacea.
Nevertheless there are comments to be made on these articles (and I might risk sending him one). For your part as an "AI Theoretician" one might want an updated view of your analysis of this AI history.
A possibly striking fact is that after 60 years of expectation AI is not yet delivering the capabilities of a two year old. Now this has some similarities to concerns in Theoretical Physics about the sense that more is now needed to complete the project, and that some very basic assumptions that have been made these past 40 or 60 years might need re-examining.
One possibility, not really discussed by Brooks, is whether there are indeed limitations to the capabilities of (this type of - or all types of) computational AI in terms of generality.If so these limitations will not be obvious in the sense that by adding another thousand test cases (ANN) or another thousand lines of code (in Symbolic AI) more real-world conditions can be taken account of. As this extension process can go on indefinitely, it might be assumed that with some years more work eventually realism will be achieved. In the process the limits to these approaches are masked.
So it might be time for AI Theoreticians to step up to the mark and clarify what is actually going on.
Yes, elsewhere Brooks has critiqued the idea of success by the finite enumeration of edge cases. There is a concept of 'fake design' whereby the emulating artefact can *sort of* reproduce a simulacrum of the phenomenon of interest (eg conversation, or affective behaviour) but replication is inherently limited by something essential lacking in the underlying architecture.
ReplyDeleteAn example might be a flapping automaton which seems to emulate a bird's activity but can't fly because the essence of flight (the requirement above all to do it!) is nowhere present in the replica.
Brooks is an engineer. In enlarging upon his critique I would focus on the consequences of social agency. Can we get anything important right if they're not capable of being one of us? Which means subject to the same innate drives and evolutionary imperatives. Humans who are deficient in this department are frequently diagnosed as mentally impaired, and are not effective social actors.