W. H. Heydt
Posts: 12642
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Tongue so far into cheek...

Tue Jun 23, 2020 2:08 am

...that's coming out an ear. In this https://www.theregister.com/2020/06/22/ ... gy_update/ article on a future wish list for very large hardware at CERN, the following paragraph shows up:
The Register imagines some of our readers have advanced prototypes of most of the above under that pile of unfinished Raspberry Pi projects out in the garage.

User avatar
Gavinmc42
Posts: 4526
Joined: Wed Aug 28, 2013 3:31 am

Re: Tongue so far into cheek...

Wed Jul 01, 2020 1:02 am

Ha, ha, Raspberry Pi users are all Emmett Brown's?

Mind you I have been wondering lately if HB11 fusion is possible.
One small enough for individual households.
500W-1KW should be enough for most people.
Need some better VCSEL diodes first.

A 500W- 1KW thermal CO2 super-critical Stirling Generator design would do in the mean time.
Run it from stored solar thermal or Thorium thermal?

Playing around with antimatter sounds a little more dangerous.
This sort of money is pocket change for some plutocrats these days, let them pay for it?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

W. H. Heydt
Posts: 12642
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: Tongue so far into cheek...

Wed Jul 01, 2020 1:14 am

Gavinmc42 wrote:
Wed Jul 01, 2020 1:02 am
Mind you I have been wondering lately if HB11 fusion is possible.
One small enough for individual households.
500W-1KW should be enough for most people.
Need some better VCSEL diodes first.
I wouldn't even consider anything less than 5KW.

Some years ago, GE (IIRC) came out with a standby or off-grid household-sized fuel cell power system. It ran off natural gas or propane (it had a built in reformer, and the waste heat from that could be used to heat water). It was, if memory serves, 7KW continuous, with a 10KW 30 minute peak, and 13KW 30 second peak output.

pidd
Posts: 567
Joined: Fri May 29, 2020 8:29 pm
Location: Birkenhead, Wirral, UK
Contact: Website

Re: Tongue so far into cheek...

Wed Jul 01, 2020 1:25 am

Very nearly 70 years of fusion reactions and not once has anyone managed to produce any power! Thank goodness computers progressed a bit faster.

This winter I will be turning my heating down a bit further and using my Pis to warm my hands, thereby increasing the efficiency of the whole of the world's power systems while fusion rectors continue to use more power than they generate.

W. H. Heydt
Posts: 12642
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: Tongue so far into cheek...

Wed Jul 01, 2020 2:35 am

pidd wrote:
Wed Jul 01, 2020 1:25 am
Very nearly 70 years of fusion reactions and not once has anyone managed to produce any power! Thank goodness computers progressed a bit faster.

This winter I will be turning my heating down a bit further and using my Pis to warm my hands, thereby increasing the efficiency of the whole of the world's power systems while fusion rectors continue to use more power than they generate.
It's much like AI (the real thing, not what gets touted as "AI" all the time).... We're 5, or 10, 20 years from having an actual AI (or fusion reactor)....and we have been for the last 50 years.

(Personally, I think the first real AI will happen by accident and it will turn out that all the ideas about how to achieve one will are wrong. The same might be true of fusion reactors as well.)

Heater
Posts: 15948
Joined: Tue Jul 17, 2012 3:02 pm

Re: Tongue so far into cheek...

Wed Jul 01, 2020 3:31 am

W. H. Heydt wrote:
Wed Jul 01, 2020 2:35 am
It's much like AI (the real thing, not what gets touted as "AI" all the time).... We're 5, or 10, 20 years from having an actual AI (or fusion reactor)....and we have been for the last 50 years.
That is so confusing.

Where is this "AI...the real thing" that you like? Or should I say what is this "AI...the real thing" that you like?

How can you like this "AI...the real thing" if at the same time you say it does not exist: " We're 5, or 10, 20 years from having an actual AI"

I have always thought calling whatever these guys do "AI" was really silly. For a start we don't have any rigorous definition of what "intelligence" is or how to measure it. So it makes little sense to me for engineers to claim they are trying to make one.

And then, if what they make is actually intelligent then why call it "artificial"? It would just be "I".

Then there is this confusion over "Deep Learning". It's not about some philosophically deep, meaningful and profound learning. It's just a description of using multiple layers of so called "neurons" instead of just one.
Memory in C++ is a leaky abstraction .

W. H. Heydt
Posts: 12642
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: Tongue so far into cheek...

Wed Jul 01, 2020 3:46 am

Heater wrote:
Wed Jul 01, 2020 3:31 am
W. H. Heydt wrote:
Wed Jul 01, 2020 2:35 am
It's much like AI (the real thing, not what gets touted as "AI" all the time).... We're 5, or 10, 20 years from having an actual AI (or fusion reactor)....and we have been for the last 50 years.
That is so confusing.

Where is this "AI...the real thing" that you like? Or should I say what is this "AI...the real thing" that you like?

How can you like this "AI...the real thing" if at the same time you say it does not exist: " We're 5, or 10, 20 years from having an actual AI"

I have always thought calling whatever these guys do "AI" was really silly. For a start we don't have any rigorous definition of what "intelligence" is or how to measure it. So it makes little sense to me for engineers to claim they are trying to make one.

And then, if what they make is actually intelligent then why call it "artificial"? It would just be "I".

Then there is this confusion over "Deep Learning". It's not about some philosophically deep, meaningful and profound learning. It's just a description of using multiple layers of so called "neurons" instead of just one.
We may not have a definition, but we have a test...

pidd
Posts: 567
Joined: Fri May 29, 2020 8:29 pm
Location: Birkenhead, Wirral, UK
Contact: Website

Re: Tongue so far into cheek...

Wed Jul 01, 2020 4:05 am

They have already achieved AI, intelligence is only associating to things you have learnt, the same definition as machine learning.

Once we get outside our knowledge zone we soon flounder just the same as a machine does, the machine is more intelligent because it doesn't dither, if its got to make a random choice it will do it without hesitation, we will replicate in an attempted avoidance.

We have hope which is irrational, the opposite of intelligence. We will expose ourselves to risk based on hoping we will be rescued - total madness! If you need to swim to shore, do it straight away while you have more strength.

And, of course, we are hypocritical, I'd "hope" :wink:

God rest us if they manage to make a machine thick enough to duplicate our "intelligence".

User avatar
Gavinmc42
Posts: 4526
Joined: Wed Aug 28, 2013 3:31 am

Re: Tongue so far into cheek...

Wed Jul 01, 2020 4:09 am

I think He3 Fusion might be why so many are going to the moon and intend to stay there.
If it will be viable they probably don't know yet.
There are a number of ways to do fusion without making neutrons
https://en.wikipedia.org/wiki/Aneutronic_fusion
I wouldn't even consider anything less than 5KW.
Depends if you have battery storage for peak demand times.
NASA have a 4KW nuclear thermal Stirling generator the size of a chair.
24/7 of 4KW is more than most people would need.

While a nuclear thermal source would be handy for space power generation and Mars habitation a generator for earth thermal would find a use.
Multi-fuel would be best, but solar thermal like solar heated hot oil/salt are CO2 neutral.

Spending effort to figure out how to make $100 500W multi purpose generator for everyone's use seem like a better use of money.
We do need to find something for the boffins to do, just to keep them off the streets.
I just object to my taxes going to expensive basic science that has no immediate benefit for me and my family.
off-grid household-sized fuel cell power system
The word I don't like in that is "fuel", that means a fuel supply system.
That's not 100% sustainable.
Mind you a little nuke generator making electrical power can take air and water and turn it into any fuel.

This "little" CERN type project has got to be the least efficient way to use up money, resources and energy I have heard of outside of ITER, even worse than wind generators.
Probably why some people want to do it?

Re AI - we will probably be Cyborgs before real AI comes along.
Still worth learning about AI, Cyborgs might not be that far away.
Being more than human appeals more to me more than the less than humans we have running around now.

The current Algorithms seem even more biased and not even close to being as logical as Molyneux.
I wonder who programmed them and why? Any source code?
Are they purposely restricting the training data?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

W. H. Heydt
Posts: 12642
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: Tongue so far into cheek...

Wed Jul 01, 2020 5:53 am

Gavinmc42 wrote:
Wed Jul 01, 2020 4:09 am
off-grid household-sized fuel cell power system
The word I don't like in that is "fuel", that means a fuel supply system.
That's not 100% sustainable.
I've long thought that using solar and wind to drive cracking water to get the Hydrogen and then using fuel cells for the sustained power generation would be a good scheme. That way, your energy storage system is the Hydrogen.

User avatar
Gavinmc42
Posts: 4526
Joined: Wed Aug 28, 2013 3:31 am

Re: Tongue so far into cheek...

Wed Jul 01, 2020 7:41 am

That way, your energy storage system is the Hydrogen.
Energy density.
High pressure or liquid H2?
Takes power to compress it.
https://en.wikipedia.org/wiki/Energy_density

Check out how energy dense stored hydro is, but probably cheaper than batteries or super capacitors.
Oz Gov spending lots of money on that with Snowy 2.0.

Using wind to make Hydrogen is the least energy dense way to make fuel.
Nice and clean but wind generators need replacing and servicing etc.
Ok small scale but scaling up to planet of 7 billion?
Imagine debt levels many, many, many times current bad levels.
Who pays? Who profits?

Australia is probably one of the few places it could be done.
But we only have 25 million people.

These Green Deals?
Good for making the green $tuff for the suppliers and investors.
Notice they rarely give cost$ or cost to the environment to make the equipment.

Thorium reactors have high temps good for Hydrogen production.
Thorium is cheap too.
https://www.thmsr.com/en/costs/
But design and licensing etc not going to happen.
Too much invested interest in existing FF and renewables, the last thing they want is cheaper power.

Antimatter is really good for power density.
Too bad we cannot just mine it.
How expensive is it? $$$ makes more profit for some.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Heater
Posts: 15948
Joined: Tue Jul 17, 2012 3:02 pm

Re: Tongue so far into cheek...

Wed Jul 01, 2020 7:57 am

pidd wrote:
Wed Jul 01, 2020 4:05 am
They have already achieved AI, intelligence is only associating to things you have learnt, the same definition as machine learning.
I suspect I'm too dumb to understand what intelligence is. Is it than ability to score well on those IQ tests as W. H. Heydt pointed out? Or something else? Everyone who states an opinion about this seems to have a different idea.
pidd wrote:
Wed Jul 01, 2020 4:05 am
Once we get outside our knowledge zone we soon flounder just the same as a machine does, the machine is more intelligent because it doesn't dither, if its got to make a random choice it will do it without hesitation, we will replicate in an attempted avoidance.
As far as I know there is no random element in AI. The algorithms are just computer programs and as such they are as deterministic as any other computer programs.

There randomness in the way they are trained. Training involves exposure to thousands, millions of example images or whatever and tweaking the millions, billions of parameters of the algorithm. But once trained there is no randomness in their output, the same input will produce the same output every time.
pidd wrote:
Wed Jul 01, 2020 4:05 am
We have hope which is irrational, the opposite of intelligence. We will expose ourselves to risk based on hoping we will be rescued - total madness! If you need to swim to shore, do it straight away while you have more strength.
Well there is a thing.

Humans claim to be rational and have some kind of intelligence. But I would argue that is only a small part of what we are. We have all kind of motivations for doing or not doing all kind of things that are not rational and often downright bonkers.

Whilst we seem to have made a start on creating a mechanical intelligence I don't see that we are close to all the other behaviors humans have.
Memory in C++ is a leaky abstraction .

User avatar
Gavinmc42
Posts: 4526
Joined: Wed Aug 28, 2013 3:31 am

Re: Tongue so far into cheek...

Wed Jul 01, 2020 8:36 am

Humans claim to be rational and have some kind of intelligence. But I would argue that is only a small part of what we are. We have all kind of motivations for doing or not doing all kind of things that are not rational and often downright bonkers.
Makes you wonder if we do make AI's will they do bonkers things too?
Not sure you could say humans are intelligent till older than 50 or 60?
This just got confirmed for me watching Sargon's take on the Scottish Parliament taking a knee to BLM.
Whilst we seem to have made a start on creating a mechanical intelligence I don't see that we are close to all the other behaviours humans have.
That might be a good thing?

Specific machine intelligence seems to be here.
General AI is still some ways away.

Tesla's autopilot is now probably better than average human drivers.
It is the edge cases that are not trained for that are the issue.
Will a General AI emerge if trained for everything?
And how could anyone tell the difference?
As far as I know there is no random element in AI. The algorithms are just computer programs and as such they are as deterministic as any other computer programs.
An interesting point.
Rule of Law requires Government decisions to be based on facts determined from the evidence.
When an algorithm makes a decisions then the facts and evidence should be available and known.

In Oz we had the RoboDebt AI go nuts and basically fine people based on predicted averages from historical data.
Now something like $700M must be paid back to people.
Was it the AI or the data or the programmer?

If the RoboDebt system cannot print out the evidence and facts then it is an arbitrary government AI decision.
All that legislation Parliaments make?
It is supposed to be there to protect people from arbitrary gov decisions.
When AI's are arbitrary, have we achieved parity with humans?

Now we have IBM, Amazon etc holding off on face recognition software and the first US legal case of face recog being used as the sole evidence.
Drawing the line for AI is getting interesting.

AI's and humans are now bumping heads with the laws in Rule of Law societies.
We can no longer think about AI's without considering laws and society.
This is actually a good thing.

Now if only we could reprogram some humans.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Heater
Posts: 15948
Joined: Tue Jul 17, 2012 3:02 pm

Re: Tongue so far into cheek...

Wed Jul 01, 2020 10:20 am

Gavinmc42 wrote:
Wed Jul 01, 2020 8:36 am
Was it the AI or the data or the programmer?
It was none of the above.

The AI is just a bunch of convolutions, matrix multiplications etc strung together that happen to do something interesting.

The data, well that just data. Assuming it was actually correct.

The programmer is just a programmer. They make the this stuff work to some degree or other.

No, the fault lies with the dumb nuts that purchase, install and use all of this and blindly trust that it works correctly. Then and the guys who sold those suckers their over hyped product.

Similarly I would say the problem is not facial recognition AI. No it's the massive power of surveillance that it provides to those that own and control it.

People are the problem. Not the AI.

And that is why we likely need laws governing it's use. Same like guns and such.
Memory in C++ is a leaky abstraction .

User avatar
Gavinmc42
Posts: 4526
Joined: Wed Aug 28, 2013 3:31 am

Re: Tongue so far into cheek...

Wed Jul 01, 2020 11:55 am

No, the fault lies with the dumb nuts that purchase, install and use all of this and blindly trust that it works correctly. Then and the guys who sold those suckers their over hyped product.
Oi, that's my government you are talking about ;)
Yes they are a bunch of idiots that listen to "experts", pretty much like any government it seems at this point.

I used to think the government knew what they were doing, some probably do but it is not what we want them to be doing.
Will AI overlords do any better?
Going to need some Law reforms, I wonder if AI can help?
I am interested to see if Democracy 3.0 will emerge out of this NWO mess.
With Oz going on a military spending spree, my gov seems to think bad times are ahead.
No it's the massive power of surveillance that it provides to those that own and control it.
Well you don't even need to own it or control it, it looks like any old company can track people now.
Mobilewalla tracked 17,000 protesters. Makes you wonder what the real spooks can do?

It has been a good news week, I found out about Graphene OS
https://grapheneos.org/
Will I need that 8GB Pi4 to compile it?
Will try tmr on the 4GB?

Even better news, Warzone2100 3.4 is out and compiles and works on Raspberry PI OS 32 and 64bit :D
https://wz2100.net/
Now there is an issue with the sound in 5.4 but the graphics bug that was in Raspbian is fixed.
3.4 even has Vulkan maybe, https://wz2100.net/news/vulkan-directx- ... -progress/

Wonder if I can test my own AI's in Warzone2100?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

pidd
Posts: 567
Joined: Fri May 29, 2020 8:29 pm
Location: Birkenhead, Wirral, UK
Contact: Website

Re: Tongue so far into cheek...

Wed Jul 01, 2020 3:18 pm

Heater wrote:
Wed Jul 01, 2020 7:57 am
As far as I know there is no random element in AI. The algorithms are just computer programs and as such they are as deterministic as any other computer programs.

There randomness in the way they are trained. Training involves exposure to thousands, millions of example images or whatever and tweaking the millions, billions of parameters of the algorithm. But once trained there is no randomness in their output, the same input will produce the same output every time.
It does come down to want you call intelligence but there is basic requirement to be called intelligent

An "intelligent function" will attempt to create an answer from inadequate information, if there were adequate information then it would be nothing more than a look-up table which can't be called intelligent. Whether you call the attempt a guess or a gamble the thing is, the answer can never be correct, there is only one correct answer and that is "I don't know".

With an intelligent system you want to force (commit to) an answer based on the perceived balance of probabilities, this is often trivial where there is overwhelming evidence (known pattern), however we want our systems to go beyond that and start guessing the un-guessable (or what appears un-guessable to us).

Lets say it has guessed an answer and through other methods it is found that the answer would be correct 60% of the time and wrong 40% of the time. In that case, if the machine was truly unbiased intelligent, if you asked it that question 100 times, it would not give that answer 100 times. Yes, this is disputable/challenge-able .... but keep reading .....

We as humans make thousands of decisions a day based on inadequate information, most of those decisions have virtually no impact so we are quite free and easy with the answer, yesterday I walked 2m away from the lamp-post, today I walked 1m away from it, it is a random freedom-of-choice and that is what makes us intelligent, we not only can give a variety of answers, we also commit to different answers on different occasions because it is inbuilt to try variety to learn more information (that flagstone 1m away from the lamp post rocks and squirt water up your leg).

You can't possibly be intelligent without creating self-learning yourself, you can only self-learning if you don't do the same thing all the time, there HAS to be a random element, even if you know that you are committing to a non-optimal answer, you know it wastes an extra second if you walk 1m from the lamp post BUT you will sometimes still do it.

What the current experts call AI may well be nothing to do with the above but I've seen little more than machine learning, I've yet to see anything that could be associated with intelligence. i formerly worked in AI (in a specialised field), I lasted about a year until I realised what a con it was. Yes, computer can out-perform us in many aspects but that does not make them intelligent.

I think it could be shown that a digital machine cannot be intelligent, it starts off with the basis that it is finite, an analogue computer however simple is infinite from the off and therefore already contains every answer whereas a digital machine doesn't and can't.

Analogue by its own nature is always random, an analogue machine will never give the same answer twice, its impossible.

W. H. Heydt
Posts: 12642
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: Tongue so far into cheek...

Wed Jul 01, 2020 4:31 pm

Heater wrote:
Wed Jul 01, 2020 7:57 am
pidd wrote:
Wed Jul 01, 2020 4:05 am
They have already achieved AI, intelligence is only associating to things you have learnt, the same definition as machine learning.
I suspect I'm too dumb to understand what intelligence is. Is it than ability to score well on those IQ tests as W. H. Heydt pointed out? Or something else? Everyone who states an opinion about this seems to have a different idea.
Sorry. I wasn't referring to "intelligence tests". I was referring to the Turing Test.

lurk101
Posts: 56
Joined: Mon Jan 27, 2020 2:35 pm

Re: Tongue so far into cheek...

Wed Jul 01, 2020 5:04 pm

I seem to recall reading that so called AIs can now pass the traditional Turing test.

Heater
Posts: 15948
Joined: Tue Jul 17, 2012 3:02 pm

Re: Tongue so far into cheek...

Wed Jul 01, 2020 6:03 pm

pidd wrote:
Wed Jul 01, 2020 3:18 pm
It does come down to want you call intelligence but there is basic requirement to be called intelligent
That is a nice long essay. At the end of it I'm not sure if you are agreeing with what I said or not.
pidd wrote:
Wed Jul 01, 2020 3:18 pm
An "intelligent function" will attempt to create an answer from inadequate information, if there were adequate information then it would be nothing more than a look-up table which can't be called intelligent.
I'm no AI/Deep Leaning expert but it's clear to me that the typical multi-layer neural networks, perhaps with simple or convolutional layers, we read about are in fact equivalent to look up tables.

For the same numbers put in the same numbers will come out. Every time.

Of course the size of the input numbers is huge, as many bits as the size of an image for example, so building an actual lookup table big enough is not possible. But functionally, conceptually it is the same.

In that respect current AI is the same intelligence as a light switch. It's just a matter of degree.
Memory in C++ is a leaky abstraction .

W. H. Heydt
Posts: 12642
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: Tongue so far into cheek...

Wed Jul 01, 2020 6:34 pm

lurk101 wrote:
Wed Jul 01, 2020 5:04 pm
I seem to recall reading that so called AIs can now pass the traditional Turing test.
On some subjects when engaged with some people (generally *not* experts in the subject). I haven't heard of one that can pass the Turing Test in general.

pidd
Posts: 567
Joined: Fri May 29, 2020 8:29 pm
Location: Birkenhead, Wirral, UK
Contact: Website

Re: Tongue so far into cheek...

Wed Jul 01, 2020 7:48 pm

Heater wrote:
Wed Jul 01, 2020 6:03 pm
pidd wrote:
Wed Jul 01, 2020 3:18 pm
It does come down to want you call intelligence but there is basic requirement to be called intelligent
That is a nice long essay. At the end of it I'm not sure if you are agreeing with what I said or not.
pidd wrote:
Wed Jul 01, 2020 3:18 pm
An "intelligent function" will attempt to create an answer from inadequate information, if there were adequate information then it would be nothing more than a look-up table which can't be called intelligent.
I'm no AI/Deep Leaning expert but it's clear to me that the typical multi-layer neural networks, perhaps with simple or convolutional layers, we read about are in fact equivalent to look up tables.

For the same numbers put in the same numbers will come out. Every time.

Of course the size of the input numbers is huge, as many bits as the size of an image for example, so building an actual lookup table big enough is not possible. But functionally, conceptually it is the same.

In that respect current AI is the same intelligence as a light switch. It's just a matter of degree.
I can say that we are in rough agreement on the overall concept of AI.

pidd
Posts: 567
Joined: Fri May 29, 2020 8:29 pm
Location: Birkenhead, Wirral, UK
Contact: Website

Re: Tongue so far into cheek...

Wed Jul 01, 2020 7:57 pm

lurk101 wrote:
Wed Jul 01, 2020 5:04 pm
I seem to recall reading that so called AIs can now pass the traditional Turing test.
That would be difficult as there is no peer reviewed consensus on the implementation of a Turing Test.

Turing's own suggestions and examples of tests were subjective and how to apply them are again subjective.

It is also flawed as a test of intelligence, if you find a professional interrogator that consistently wins better than 50% at Turing's games and store every question and answer over a number of years, the result would be a database that would probably also win the game more than 50% - no intelligence involved, a simple look-up table.

Heater
Posts: 15948
Joined: Tue Jul 17, 2012 3:02 pm

Re: Tongue so far into cheek...

Wed Jul 01, 2020 11:27 pm

Wait a minute.

In the Turing test as I know it, and as described by Wikipedia, it is not about a human chatting to something and having to decide if that something is human or machine.

In the Turing Test there are three participants, the human chatting to the machine, and a judge who watches what happens. The judge knows one is human and one is machine. If the judge cannot tell which is which, the machine passes the Turing Test.

Which seems grossly unfair to me. Now the machine has two humans to contend with to prove itself.
Memory in C++ is a leaky abstraction .

pidd
Posts: 567
Joined: Fri May 29, 2020 8:29 pm
Location: Birkenhead, Wirral, UK
Contact: Website

Re: Tongue so far into cheek...

Thu Jul 02, 2020 12:11 am

Turing described a number of variant of "games", the format you described is one of the better ones but is a long way from scientific proof of anything nor are all the criteria specified.

Surely the skill of the Judge comes into play and his judgement is subjective - is this more a test of the Judge or the machine?. What pre-knowledge does the judge have? Is the machine supposed to answer honestly or is it programmed/taught to purposely deceive? What role does the other participant have, are they supposed to deceive or not? Are they supposed to try and beat/expose the robot's flaws or not? Do they know the other person is a machine? Who knows what the purpose of the exercise is? Who takes the lead in the conversations? What are the conversations about etc etc an endless list of incomplete specifications and a lot of subjectiveness.

There are three possible answers by the judge, the robot wins on two of those answers, are the odds stacked in favour of the robot, especially when you consider the Judge is guessing?

I don't think fooling someone into believing you are human demonstrates anything to do with intelligence, fooling someone into believing you are an intelligent human would be a more realistic target but then you are still stuck with defining intelligent. Turing's criteria was to show something that could think, not something that was intelligent.

Thinking is yet another abstract, especially if you exclude randomness. If we picture something, is that thinking? Obviously a computer can replicate that, in both cases it would normally be a recall not a pure random thought.

Furthermore you don't know what Turing was thinking, he may have been considering that a human is faster than a computer when it involves conversation as opposed to calculation and his tests would expose that the computer couldn't respond fast enough in normal conversation. His test would have been applicable then but I doubt he could foresee a time when computers are as fast as they are today.

Heater
Posts: 15948
Joined: Tue Jul 17, 2012 3:02 pm

Re: Tongue so far into cheek...

Thu Jul 02, 2020 6:09 am

pidd wrote:
Thu Jul 02, 2020 12:11 am
Surely the skill of the Judge comes into play and his judgement is subjective - is this more a test of the Judge or the machine?. What pre-knowledge does the judge have? Is the machine supposed to answer honestly or is it programmed/taught to purposely deceive? What role does the other participant have, are they supposed to deceive or not? Are they supposed to try and beat/expose the robot's flaws or not? Do they know the other person is a machine? Who knows what the purpose of the exercise is? Who takes the lead in the conversations? What are the conversations about etc etc an endless list of incomplete specifications and a lot of subjectiveness.
It's easy then.

If all three parties in the Turing game are having a raging debate about how to play the game, the meaning of it all, and all those questions. Say in a forum thread much like this one. Then nobody could tell which one is the machine and the machine passes the Turing Test.

I win :)
Memory in C++ is a leaky abstraction .

Return to “Off topic discussion”