Re: [agi] AGI questions

https://www.youtube.com/watch?v=U1WtwfUm368 When the fuel is ignited, it decompresses and extracts free hidden energy in the engine and turns the wheels of the truck. A truck driver knows where he wants to drive to, so he knows what road to decide to Pick Next or 'predict'. He understands the

Re: [agi] Re: AGI speedrun continues

I spent some time with Lenat's crew in Austin circa 2000 as part of my deal with HP's "Internet Chapter 2" project ala "eSpeak". In tha tdeal I rather insisted that I be permitted my own resources to pursue some fallback plans in the event that their primary "vision" failed. What I was

Re: [agi] AGI questions

Show me how you can use a data compressor to drive a truck. On Thu, Feb 20, 2020 at 10:27 PM wrote: > > All life on Earth seek food/shelter/mates so they can immortally sustain the > species and repair missing data or employees. We go to antibiotics for that > reason. Survival. All labor tasks

Re: [agi] Re: AGI speedrun continues

On Thu, Feb 20, 2020 at 7:40 PM wrote: > > On Thursday, February 20, 2020, at 7:02 PM, Matt Mahoney wrote: > > Cyc was a failure. But it is interesting that in 34 years we haven't found > something better. > > > How so? Is Cyc that good? Show me, I've never seen Cyc in action. I have, and it

Re: [agi] AGI questions

All life on Earth seek food/shelter/mates so they can immortally sustain the species and repair missing data or employees. We go to antibiotics for that reason. Survival. All labor tasks are for that reason. Survival. We will make AGI achieve its own survival by forcing them to enable our

Re: [agi] AGI questions

The goal of AGI is to automate human labor. It requires solving hard problems like vision, language, robotics, art, and modeling human behavior. We don't need to define intelligence to solve these problems. If you want to define intelligence as survival, then that is not what we want to build. We

[agi] NC girl born without eyes or nose graduates from college

https://www.youtube.com/watch?v=F3JDumpJfPg So you don't need vision or smell to make AGI pass College. The only catch is vision is basically touch. Except you can't feel a painting :-), only shape; the real thing, not a drawing of it. It is same thing.

Re: [agi] AGI questions

Have you tried telling OpenAI to use Lossless Compression as the evaluation for GPT-2 etc? They seem to have not used it yet got GPT-2. But with it, perhaps they can achieve something better. -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] AGI questions

So the survival test is best but takes too long, Lossless Compression is best but too narrow (a bit). The turing test would also look for something, namely human-like outputs that could be predicted. So yes passing a *questionnaire (not a mere conversation) test would work as a good AGI test,

Re: [agi] AGI questions

Furthermore, you can do a turing test on a visual compressor if you want to. It would need to have the AGI components though and not just be a compressor. So possibly Lossless Compressor isn't a full evaluation (maybe!, because it does use Online Learning etc, which is profoundly surprising!).

Re: [agi] Re: AGI speedrun continues

On Thursday, February 20, 2020, at 7:02 PM, Matt Mahoney wrote: > Cyc was a failure. But it is interesting that in 34 years we haven't found > something better. How so? Is Cyc that good? Show me, I've never seen Cyc in action. -- Artificial General

Re: [agi] AGI questions

First of all, I don't like the 'turing test' because fooling a human doesn't imply intelligence, survival should be the evaluation. The next best thing to test is Lossless Compression (because testing survival requires watching long term, inventions...). Of course copying humans will lead to

Re: [agi] Re: AGI speedrun continues

On Wed, Feb 19, 2020, 3:17 PM WriterOfMinds wrote: > No, to my knowledge the Winograd challenge has not been solved (at least, > not to the point of a program getting correct answers on all the sentences > in a test set). The only other hobbyist/independent researcher that I've > seen openly

Re: [agi] AGI questions

How would you design a visual Turing test? Draw pictures and test if you can tell if they were drawn by a human or a machine? On Thu, Feb 20, 2020, 5:44 PM wrote: > There is a visual turing test (I mean you can make one), you'd use > prediction to complete the rest of an image or video, you'd

Re: [agi] A comparison between dolphin sounds and neuron signals

Linguistic structure is undoubtedly present in the firing patterns of some subset of neurons in the human neocortex. Before I dismiss the plausibility of dolphin sounds being relatively direct transduction of neuron firing, I'd want to see what linguistic structure looks like in human neuron

Re: [agi] A comparison between dolphin sounds and neuron signals

A human is a human, and a team of them can only be a human. The goal steps can be split among their arms, eyes, brains, data. The more the global swarm brain knows, the better. It also needs redundant clones. When you have many humans working on sub tasks that are each different domains, these

Re: [agi] A comparison between dolphin sounds and neuron signals

On Wed, Feb 19, 2020, 10:57 PM James Bowery wrote: > What is the next best thing to telepathy? > Interesting they sound alike but it doesn't mean anything. -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] AGI questions

There is a visual turing test (I mean you can make one), you'd use prediction to complete the rest of an image or video, you'd use Lossless Compression as the evaluation criteria the same as the Large Text Compression Benchmark contest does. I admit it seems no same patch in an image appears

Re: [agi] AGI questions

Visual rotation invariance is not absolute. It is harder to recognize faces upside down. Compression doesn't work well for evaluating vision. First, there isn't a visual equivalent of the Turing test. Second, the information content of video is dominated by noise. Long term memory capacity for

Re: [agi] Re: AGI speedrun continues

Solving the Winograd Challenge, Pt. 2: Making a multi-stage processing framework and a top-down parser. If I find the energy for it, I'll make part 3 tonight. We are entering the "logic engine" phase where it actually derives true statements & then answers the question.

[agi] Heavily Typed Conceptual Language

I have not been able to come up with a way to overcome p!=np in logic, so I am thinking about developing a heavily typed logic (logic-like references) as a way to get around the bottleneck of exponential complexity. However, I have run into some difficulties there as well. I would like an object

Re: [agi] Re: A comparison between dolphin sounds and neuron signals

On Wed, Feb 19, 2020 at 10:33 PM wrote: > ... > Yes dolphins are a full AGI agent, they don't need others. The team > "brain" also is 'there' but not fully functioning yet in different regards; > ants do pretty good. We could do better though > Current understanding of eusocial evolution is

Re: [agi] Re: AGI speedrun continues

Cool, thanks -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T887757e45bfd1342-M0c70a46956eca8d68d4cdb46 Delivery options: https://agi.topicbox.com/groups/agi/subscription