[jbritely]

I did not see a thread about applications.

As pointed out in the book predictions about applications are bound to be wrong but I still found the last chapter very interesting. My prediction is that current computer technology will handle that problem fairly well before true AI is completed (both 10 to 15 years off).

It may sound mundane but I think analyzing historical data will be one of the biggest payoffs. This can be started with the simpliest AI and continued with each improvement. Even an AI with very limited and focused intelligence on pattern recoginition but large and near perfect memory could find many new discoveries with historical data. For example, feed all historical financial data and maybe we will figure out recessions/depressions. Feed historical weather data to find new patterns. This will be much quicker than hooking up new sensors as suggested in the book (not saying we souldn't do both).

What applications did you think of?

[jbritely]

Sorry for the bad editing, the third sentence should read: My prediction is that current computer technology will handle the smart car problem fairly well before true AI is completed (both 10 to 15 years off).

[danshawen]

You are right. They (computers) already did solve the problem. It happened on or about October, 2005.

The problems remaining however are not particularly suitable to these glorified adding machines. This is because true intelligence has so little to do with what a typical computer does.

The Turing model of intelligence is also an apt one. The punch and tape reader arrangements has many problems that make it sub-intelligent, including problems with getting enough paper and memory to do what you ask it to do. But the very MOMENT a Turing machine becomes capable of experiencing its environment directly through some means other than a keyboard, and to affect its environment (including communicating in some manner with its own memory and with other similar beefed up Turing machines) it instantly acquires the status of "intelligence", although admittedly, not to a very great extent. To give it more intelligence from that point on does not mean "giving it more memory", or "faster processors", but "better sensors", and "better manipulators".

Many of the entries in the DARPA Grand Challenges (post-2005), including DEXTER were actually intelligent because they demonstrated that they could sense and interact with the environment to accomplish a goal.

Paper tape punches and readers were, for the first time, replaced by the open road and whatever the machines could maneuver around or over. These were no longer Turning machines.