I'm not aware of a single AI project who views solutions to problem in terms of just 'writing more code'. Many AI systems work with remarkably small amounts of code.

And these remarks are amazingly fat-handed in dealing with the huge array of questions that fall under 'AI'. Everything from Automated Theorem Provers to 'Intelligent' Tutoring Systems to the sorts of perceptual systems he whines about are often thrown together under the auspices of 'Research into Artificial Intelligence'. The problem with perceptual systems may have more to do with the limitations of our tools (digital serial computation and above nano-scale machinery may be like augers and sledge hammers compared to the sophistication of biological systems ).

Meads claim that the conceptually hard questions concerning AI are not 'being asked' because of the tenure system I suspect arises form the fact that he is just not well read. On the philosophical level, understanding what we mean when we say 'intelligence' has been a thriving industry for the last 50+ years- indeed the first attempts to formally and rigorously address the question began with Turing himself! Plenty of people are getting tenure by publishing work on the 'hard' or 'deep' questions. The point is: these _are_ both hard and deep questions.

I am not an AI expert, but I do use (and work on) 'expert systems' in my work on a daily basis, and my 'ancient history' grad work was very bound up in the field of AI. As limited as the value of my insight into the problems and hard questions of AI is, I suspect they are far more interesting than Dr. Meads. Meads Fruit Fly Saw strikes me as exactly like the 'BumbleBee issue' that is a favorite old song of pseudoscientists: That science can not explain the flight of the bumblebee. In fact it can- but the materials and tools we have on hand can't build a bumblebee. Given what I know of visual systems, we probably can go a long way towards articulating how the fruitfly operates, but of course we don;t have to tools to build anything near as sophisticated as even a fruitflys perceptual system, especially on fruit fly scales.

Its 'sexy', 'hip' and 'iconoclastic' to haughtily throw such 'thunderbolts' at academic research from the outside. Wolfram, for example. But in more cases than not, such critiques are clumsy silly shallow affairs that sound good to the beer drinking common man, but roundly miss most of the points.

These things are damn hard, and very smart people spend their entire lives stuck at full throttle to just push us a few inches down the road. Worse, these fields are inherently messy with ill-defined questions. AI as a field is arguably 60 years old at best, and we're still trying to get the definitions, questions and techniques right- we have had a few 'great leapers' like Turing, Simon, Newell and etc... But I don't think this field has yet seen an Einstein. And compared to the structure of our only reference point, the embodied human mind, even our best multi-core systems are like flakes of chert stacked up against the space shuttle.

Yeti