Comments on: HEPP: Human Equivalent Processing Power

By:

Sat, 06 Jun 2009 13:35:10 +0000

It think that AIs that function in non-human ways will be developed long before exact brain simulations – once AIs start to be commercially useful they will outstrip the performance of the human brain without ever coming close to being human, and make brain simulation unnecessary. Simulation of the human brain is just one project within AI – and human brain intelligence is only one type of intelligence (like walking is only one limited form of locomotion) – why recreate something so limited? I want computers to do things I can never do, not recreate the intelligence of an average Joe, who, lets be frank, is pretty dumb (even if he is the peak of evolved biological intelligence).
Dan.

By:

Thu, 04 Jun 2009 09:09:13 +0000

Recently I discovered your website and have been following along steadily. I felt I could give my opening comment. I dont know what to write but that I have really loved perusing. Interesting website. I will keep coming back to this blog now and again. I have also got your rss feed to get any updates.

By:

Fri, 29 May 2009 09:09:16 +0000

I’m skeptical of all the arguments based on operations-per-second, when we seem so relatively ignorant of which operations to perform. Who believes that just simulating a sufficiently large mass of virtual neurons will produce intelligence? Evidence for this position? And copying a brain: This surely requires (a) technology, presumably of the nano-persuasion and not primarily computerish, for reading the detailed structure of a brain and (b) a brain to read, in what seems likely to be a destructive process… Or we’re going to download and simulate the brain of a person who has just died? Controversial *and* dubious, a little?

Did the Wrights have to wait until powerplants evolved to meet their specifications? (OK, yes, a little.)
Did Alexander Bell have to wait for big enough batteries or pure enough carbon? Most inventions are based on a novel combination of existing parts and materials, often based on a counter-intuitive belief, or careful development of new understanding – often enough, the parts and materials have been available for years, or decades, sometimes for centuries. The parts and materials are not enough! Where’s the graph that quantifies our progress toward ‘intelligence’, as opposed to ‘enough gates’?

By:

Thu, 28 May 2009 22:59:33 +0000

When you write “In my” you need to put your name with it.

By:

Thu, 28 May 2009 22:56:52 +0000

Who are you?

By:

Thu, 28 May 2009 03:20:12 +0000

It Is what it is, It A I already Is . has been since aug 47 , artificially inseminated half humanoid half ( lets say visitor with nanobots inserted at time of conception
16 have survived. 4 are somewhat funcional. yes I think they , them . already Exists

at ? between dexter and roswell Hangers # – 120 to # + 19 !!!! army nurses volenteered for lifetime of monertary comfort.

By:

Wed, 27 May 2009 18:56:53 +0000

modeling neurons should have nothing to do with building an AI
Minsky provides the framework in Society of Mind and The Emotion Machine.

By: Nanodot: Nanotechnology News and Discussion » Blog Archive » HEPP: serial or parallel?

Wed, 27 May 2009 13:31:19 +0000

[...] What kind of software will AIs run? This is of some interest, because it will tell us how much the current flowering of parallel hardware will actually get us toward human equivalent processing power. Amdahl’s Law holds: If the task of being intelligent is strongly serial, all those processors won’t help much. If it’s parallelizable, they will, and that means that the hardware for AI is basically here. [...]

By:

Tue, 26 May 2009 21:22:21 +0000

Seems like the Blue Brain project would provide a roadmap of sorts

By:

Tue, 26 May 2009 19:18:53 +0000

Fine, so the hardware for human-level AI is already here or will be soon. What about software? Which architectures are most promising? Is it just a matter of duplicating the brain neuron-for-neuron? Will general AI evolve out of increasingly clever robots with lots of hacked together narrow AI components? This is where I never get a clear roadmap to AI in these short time frames. Would anyone care to enlighten me?