Why we don't have more progress in AI

1) Deep ambivalence in the public, elected officials and the elite (IE everyone) about whether it's good to have AI move forward in the first place. The diversion into talk about 'morality' and 'ethics' whenever AI discussed reveals this ambivalence. You don't get the same kind of discussion at a conference devoted to materials science. People are not afraid of a stronger beam in a bridge they way they are of 'better/smarter/more capable' AI.

2) Many vested interests would have a lot to lose from better/stronger AI so they can be counted on to be ambivalent at best towards the prospects of better AI.

Examples are the professions of:
Accounting
Law
Medicine

and many more.

Take Law for instance. Would it possibly be a benefit for many people to have some sort of a system that could understand natural language and answer (or at least start to answer) routine legal questions? Of course. And how would our legal profession feel about that. We can be quite certain that the legal profession would be ambivalent at best - and probably hostile to the whole enterprise.

Professions like law, medicine, and accounting are regulated and in order to practice in these fields you need some sort of standing (typically a degree and then you must pass one or more exams). How exactly would some AI entity gain the the standing to practice in these areas. This has not happened yet and is not likely to happen soon. Hence for AI to operate in these fields it has to be under the control (used as a tool) of a human provider. IE a law firm can use AI for legal research but you or me cannot. See:
http://www.nytimes.com/2011/03/05/science/05legal.html

Note that automating discovery has cost some firms lots of billable hours. How exactly would the legal profession feel about an online agent available to answer routine legal questions. Currently I would have to go to a law firm and pay a retainer (hundreds of dollars typically) just to attain the right to get a question answered.

To show how tenaciously (and effectively) professions can fight automation (much less AI) look at the Real Estate field. Ten to fifteen years ago the the prevailing feeling was that the traditional methods were doomed. Who would need/want the MLS system after the arrival of the internet. What would happen to 90% of the real estate agents out there charging 7% when people could just go online. Surely a website or a set of websites would disrupt the whole thing. But it's 2011 and the real estate field operates pretty much the way it did in 1980 and still charges 7%, the MLS system is not just alive but as dominant as ever and can only be accessed by a licensed real estate professional. Most tellingly, the typical real estate commission is the same as it has been for decades. See:
http://www.city-data.com/forum/real-estate-professionals/7423-whats-typical-realtor-commissions-5-6-a.html

7 percent (or maybe 6 if you really negotiate hard) was standard in 1980.

The MLS system was closed in 1980; you had to be an agent to get access. As of today (May 14, 2011) it still is: "Most MLS systems restrict membership and access to real estate brokers (and their agents) who are appropriately licensed by the state (or province); are members of a local board or association of realtors; and are members of the trade association" source: http://en.wikipedia.org/wiki/Multiple_Listing_Service#Limitations_of_access_and_Criticisms_of_MLS

Any entity (human or AI entity) needs interaction and feedback in order to progress but in most real world cases AI entities lack standing and hence cannot interact and get feedback directly. They have no 'agency' to act on their own and this in my opinion hobbles progress in AI.

Maybe corporate person-hood could be helpful. Perhaps a corporation could be formed with some sort of AI system as it's main component and then perhaps this entity could be given the license to practice law (or accounting or medicine) - given that a corporation is legally a person in some respects.

Some in the AI field may be thinking of a more general solution - something that could 'do it all' so to speak. My response would be to look at life: do we have one ultimate life form or do we have finches with specialized beaks, algae, killer whales and bighorn sheep. Living things evolve, adapt and fill in whatever space/niche they can. My feeling is that AI must do the same to make real progress. When it comes to 'higher order' human affairs that means practicing law, medicine, doing real estate transactions and so forth.