11 November 2009

Mind Uploads, Emulations, and Machine Intelligence

For some very interesting speculations concerning the uploading of human consciousness, development of human level machine intelligence, and the emulation of human minds in machines, check out the Singularity Summit 2009 videos that are now online. A wide variety of ideas and viewpoints will help to spur your imagination and perhaps open a few mental pathways.

Al Fin has very low expectations for these technologies in the near term. They are very hard problems -- much harder than the "gee whiz" futurists comprehend. Worse than that, the underlying society is rotting quickly, and may be unable to sustain the effort long enough for success to occur.

If the underlying societal substrate breaks down, large portions of nascent science will tend to break down as well. Some will be exported to other continents or countries. Some will be lost, at least for a time.

If these technologies are developed in a time of war or great instability, under a dictatorial regime hungry for conquest or hegemony, they can contribute to massive hardship. Unfortunately, all of the important decisions currently being made by the leaders of the western world will have the ultimate effect of weakening the west.

The west will eventually recover, somewhat.  But during the recovery period the world will be extremely vulnerable to falling into a scenario described by Neal Stephenson in "The Diamond Age".

Hope for the best, prepare for the worst, and be ready to exploit any opportunity that may arise.

Labels: ,

Bookmark and Share


Blogger Sword S said...

If human level machine intelligence actually manages to open the door to super-intelligence, it won't matter what regime creates it. I doubt it can be controlled or contained, it will be what determines the future.

Wednesday, 11 November, 2009  
Blogger al fin said...

One hears a lot of that sort of thing around singularity websites. But it is the details -- and the time span of the necessary evolution to human level AI and beyond -- which are critical.

Just as it mattered that the nuclear bomb was developed by the US rather than Nazi Germany, so it will matter whether human-equivalent or better AI is developed by a liberal democracy rather than by an ambitious dictator.

Very stupid leaders, bosses, and mob thugs are able to control very intelligent genius scientists and tacticians every day.

It is likely that stupid humans will find ways to control ingenious machines for at least a finite period of time. Long enough to matter who is in control.

Wednesday, 11 November, 2009  
Blogger Sword S said...

People may be able to control it for a short while, but can they trust what it offers for that short while. Amidst the strategies, code and hardware designed|offered by it, may be the means to its escape or goals.

Also, if the necessary theoretical work turns out to not be exotic enough[ consider that a working example the brain can be studied by anyone.], it may not be the case that one group can expect its discovery to go unmatched for long. If that were the case attempts at constraining the first true ais may lead to rival groups surpassing them.

Though, the scenarios where things turn out negatively for most are not inconceivable. I believe it wouldn't be easy for frail humans to be at the center of such for long, at least not while still remaining human.

Wednesday, 11 November, 2009  
Blogger al fin said...

Yes, it is something of a crap shoot. Kurzweil wants to hurry up and get to superhuman AI because he thinks it will let him live forever.

What if superhuman AI doesn't want Kurzweil to stay around? Goodbye Kurzweil.

Kurzweil's projections for SHAI by 2029 are very irrational. They betray a lack of understanding of how consciousness arises from a glob of neuro-glial tissue (plus a body).

Kurzweil's vision will be superceded by better informed models of human consciousness which should put human - level AI somewhere around 2063, around November 23 at 6:27 am PST.

It will take 19 years, 3 months, and 2 and a half weeks longer to achieve true super-human AI.

During that 19 years plus or minus, the controller of human-level AI will have amassed a fortune in intellectual property rights, and will have nano-built an army of AI controlled vehicles patrolling land, water, undersea, air, and space.

Okay. Clearly no one knows how long it will take to get to any given level of AI. So far, the entire enterprise stinks to high heaven with 50 years of unfounded claims and predictions.

One Al Fin engineer who has backgrounds in both neuroscience and computer science, has told Al Fin that he is singularly unimpressed with current achievements in the field.

Thursday, 12 November, 2009  

Post a Comment

“During times of universal deceit, telling the truth becomes a revolutionary act” _George Orwell

<< Home

Newer Posts Older Posts