Author Topic: Sentient AI or?  (Read 10864 times)

0 Members and 1 Guest are viewing this topic.

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #50 on: June 14, 2022, 11:39:54 pm »
Yeas, this is a bit more advanced version of deep Markov chain with a lot of training data. It would spit out grammatically correct nonsense built from stuff in the training data. This is interesting, but pretty pointless.

Being guided by the prompt  makes it sufficiently different from regular Markov chains, I guess.

I see.  So it is effectively relying on connections that my brain is recognizing and making sense of (similar to what James_s just said, as well).  Possibly making me think it is better and more accurate, than it really is.

I.e. My brain is partly fooling me, to think that the story is better and more accurate, than it really is.

I can see why some people think that machine learning is a very important and growing technology thing.  It does look like it is gradually getting there (AI).
 

Online ataradov

  • Super Contributor
  • ***
  • Posts: 11905
  • Country: us
    • Personal site
Re: Sentient AI or?
« Reply #51 on: June 14, 2022, 11:57:04 pm »
It generates random words with the same gramatical connections as in the source (training) text. This is not a whole lot better than Rockwell Retro Encabulator. Something that sounds like a coherent text, but has no real meaning. In this case is is tuned to spit out something that would not be so obviously fake. But even in case of the encabulator, it takes a few minutes to figure out that it is a spoof, and non-technical people may not even be able to figure that out at all.

The issue here is that this system does not actually understands the meaning of this text, so it can spit out a valid looking text, but it can't maintain a conversation or answer any followup questions. All it knows is how the words should be ordered to resemble valid English.

So, I would separate AI from generators like this. This is not AI in any non-marketing way.
« Last Edit: June 15, 2022, 12:02:16 am by ataradov »
Alex
 
The following users thanked this post: SilverSolder, MK14, james_s

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: Sentient AI or?
« Reply #52 on: June 15, 2022, 12:12:05 am »
You're completely right, I agree.

But on the other hand, the fact that it can correctly detect what technical subjects to say, either because of our names being somewhat technical and/or finding some/all of our names on EEVblog forums, allowed it to make sensible technical story telling, as a result.
So, I'm still rather impressed with it.  Also, it sounds sort of plausible.

TL;DR
You're right, in what you said.  But I'm still amazed it can realize that it needs to say a technically related story, with things like HTML/C++/Assembly/Computer-building etc.

I have no idea how it works internally, but it's not that hard to generate more or less grammatically correct gibberish. I always liked madlibs when I was a kid and still have a stack of the books of them. A while back I wrote a python script that fills them out from random lists of words and they sometimes come out rather funny, often with a more esoteric vocabulary of words than I would have come up with myself on the spot. Just for grins I just had it spit out a few now, there is no AI here at all, it's ridiculously simple logic, yet by having a collection of premade templates with spots for randomly chosen nouns, verbs, names, etc it results in readable content that is different every time. It wouldn't be hard to have it scrape some other source for words to use.

    Specialty Of The House

Here is chef Erika's award-pinching
recipe for roast mouth of moose: Choose a
mouth weighing about 65 rags.
Remove excess wiener. Add 5 cloves of garlic, peeled and
ducked. Season with 2 tablespoons of chopped cake.
Add a tablespoon of mustard. Sprinkle with a touch of
bad salt. Add a pinch of ground blue boyfriend.
Cook at 350 tubular bells for 88 minutes. Remove from
the oven when the skin is pink. Serve with mashed
jock straps and a/an moose.


      What's In A Name?

William Shakespeare is regarded by scholars and mooses
alike as the greatest playwright and stool ever to put
pen to rectum. Although he wrote in his native
vagina, Shakespeare has been translated into twelve
different t-shirts and his plays and poems are tediously
read and performed everywhere in the world. Hamlet's soliloquy-
which begins "To be or not to be, that is the machine gun,"-
has been delivered on stage by more crufty actors
than any other physical fart ever written.
Among Shakespeare's greatest plays are 'Hamlet', 'Romeo and Juliet',
'The Baseball Venice', 'King Gallegos',
'A Midsummer Night's Triangle' and 'The Taming of
the Cube'. We could go on, but we must leave. As Romeo
said to Juliet, "Parting is such shitty sorrow."


      It's About Time

Thousands of ironing boards ago, there were calendars that
enabled the ancient robots to divide a year into twelve
hats, each month into 14 weeks, and each
week into seven light sockets. At first, people told time by a
sun clock, sometimes known as the bowtie dial. Ultimately,
they invented the great timekeeping devices of today, such as the
grandfather mosquito, the pocket window, the alarm
banana, and of course, the belly watch.
Children learn about clocks and time almost before they learn their
ABS's. They are taught that a day consists of 24
horses, an hour has 60 cows, and a minute has 60
beers. By the time they are in kindergarten, they know if
the big head is at twelve and the little ballsack
is at three, that it is 18 o'clock. I wish we could continue this
rancid lesson, but we've run out of pussy.

 
The following users thanked this post: MK14

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #53 on: June 15, 2022, 12:17:43 am »
In answer to both your questions.  I devised a mini-intelligence test.  It seems to have passed the test, with flying colours.

Here:
Test question was:
"If up is down and left is right, what should increase be?"

And it's AI came up with:
Quote
decrease

Which to me, sounds like it has some actual AI to it.

Full conversation:
Quote
If up is down and left is right, what should increase be?

decrease

EDIT: Also this (question is inside the quote, on first line):
Quote
If the world is without air, will everyone be ok, and why?

No, everyone will not be ok. People need air to breathe.

Which also, seems to show some comprehension, understanding and intelligence.

It also seems to correctly process the following (but if I miss out the words in the brackets, it gets it WRONG, and says 1).
Quote
If I have an Apple and eat it (so the Apple is gone), how many Apples will I still have?

Zero.
« Last Edit: June 15, 2022, 12:28:03 am by MK14 »
 

Online ataradov

  • Super Contributor
  • ***
  • Posts: 11905
  • Country: us
    • Personal site
Re: Sentient AI or?
« Reply #54 on: June 15, 2022, 12:26:59 am »
This is a bit too short of a test. The real test would come from questions that don't really have a good answer. Something like "Given correctly biased NPN-transistor, how many angels can dance on the head of a pin?". The idea here is not to test the knowledge, but how the limits of that knowledge are handled. A real intelligence would  either admit that it does not know or it would point out that the question does not have a meaning. AI will often produce an answer. And the answers to incoherent questions are most telling and they will not lead you to trusting that it "looks" correct an makes sense.

Those answers show that it is not just a Markov chain. It has a better structure to preserve context from the original training data. That's why it is important to take it out of that context for testing. Otherwise it is like competing with google and wikipedia in factual knowledge. You will lose, but that does not meant that google or wikipedia are "smart" or intelligent.
« Last Edit: June 15, 2022, 12:30:44 am by ataradov »
Alex
 
The following users thanked this post: MK14, james_s, pcprogrammer

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #55 on: June 15, 2022, 12:33:04 am »
This is a bit too short of a test. The real test would come from questions that don't really have a good answer. Something like "Given correctly biased NPN-transistor, how many angels can dance on the head of a pin?". The idea here is not to test the knowledge, but how the limits of that knowledge are handled. A real intelligence would  either admit that it does not know or it would point out that the question does not have a meaning. AI will often produce an answer. And the answers to incoherent questions are most telling and they will not lead you to trusting that it "looks" correct an makes sense.

Those answers show that it is not just a Markov chain. It has a better structure to preserve context from the original training data. That's why it is important to take it out of that context for testing. Otherwise it is like competing with google and wikipedia in factual knowledge. You will lose, but that does not meant that google or wikipedia are "smart" or intelligent.

The first result, said it wasn't possible to answer the question.  But I badgered it again, and it came up with:

Quote
Given correctly biased NPN-transistor, how many angels can dance on the head of a pin?

One angel can dance on the head of a pin if the transistor is correctly biased. If the transistor is not correctly biased, then the number of angels that can dance on the head of the pin is zero.

Which is wrong, but does show a little bit of understanding, sense and logical thinking.  In a small way, I guess, but still something.
 

Online ataradov

  • Super Contributor
  • ***
  • Posts: 11905
  • Country: us
    • Personal site
Re: Sentient AI or?
« Reply #56 on: June 15, 2022, 12:42:12 am »
This is a decently coherent answer. The next level is dealing with a complete word salad and not even in a form of a question - "Slice leadership of a jealous diplomatic code to reproduce by virtue of a nap or a drink in heaven after crashing a satellite on heroin". (generated by a random word generator, connecting words are mine).

Or try something that would clearly not be in the training data, yet reasonably easy to know. Like "What is the value of the integral of x cubed from 1 to 7". If I plug in that sentence into wolfram alpha, it gives me the correct answer (600). And I expect the correct answer.

But I'm also not entirely clear why this is useful, even if the answers are just ok. If you can't expect them to be factual, then what's the point?
« Last Edit: June 15, 2022, 12:51:41 am by ataradov »
Alex
 
The following users thanked this post: MK14

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #57 on: June 15, 2022, 12:50:44 am »
This is a decently coherent answer. The next level is dealing with a complete word salad and not even in a form of a question - "Slice leadership of a jealous diplomatic code to reproduce by virtue of a nap or a drink in heaven after crashing a satellite on heroin". (generated by a random word generator, connecting words are mine).

It results in:
Quote
Slice leadership of a jealous diplomatic code to reproduce by virtue of a nap or a drink in heaven after crashing a satellite on heroin and cocaine to the north of the top pay-per-view TV event in an old-fashioned way

A big financial analyst to make a wish come true that can be a part of an open and shut case but has been a small village where local farmers live in small apartments but some people are taking a rest on a white sand beach with a lot of people

A slice of pie or the California coast to feed your curiosity by accident to make a wish come true that you want to get off the phone book

A walk on the phone book by accident that you want to get off the phone book by accident that you want to get off at night and sleep on it for a while and see what happens when someone tries it on themselves

I myself have not yet been able to take advantage of this unique opportunity. But I do have an open mind about it. After all, who wouldn't want their wildest dreams to come true?

Second attempt, produces this:
Quote
Slice leadership of a jealous diplomatic code to reproduce by virtue of a nap or a drink in heaven after crashing a satellite on heroin, Why Is There Something Rather Than Nothing? The new moon is of interest to us.

Addison Wesley longman, Inc.


I tried another of my own questions, and it seems to have some understanding about how magnets work.  See here:

Quote
A magnet, is put near a sweet and an iron-bar, what happens?

If the magnet is put near the sweet, nothing happens. If the magnet is put near the iron-bar, the iron-bar is attracted to the magnet.
 

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #58 on: June 15, 2022, 01:00:33 am »
I repeated your question a number of times.  This time (shown below), you manged to make it 'halt and catch on fire'.  It really seems to spill its guts, show its database(s), and give clues, such as URLs, as to some of the data, that was used to train/make it.

Quote
Slice leadership of a jealous diplomatic code to reproduce by virtue of a nap or a drink in heaven after crashing a satellite on heroin, the concept of the screen: The screen is a sudden, quick death. ) what kind of life and excitement than it is with me. There is a very different from other cultures. I am looking forward to seeing you in this book. He has learned that there are no emergencies in my house during the day, is rarely considered. Perhaps we can use them with bold headers, this time next week that everybody does.

Planning essays and reports

What types of speech do you know

Emerson essay montaigne

Search for Advanced Degree Appointments (SADA) (SUNY only) or Non-SUNY Appointments

Bing presentation template and presentation dress code example powerpoint business casualPowerPoint 演示文稿: PowerPoint Presentation - Protection of Rights by Due Process of Law And presentation dress code example powerpoint business casualPowerPoint 演示文稿: PowerPoint Presentation - Protection of Rights by Due Process of Law - UK Essays | UKEssays It may seem large and small learning objective lo 1 identify why commas are used specifically for children or adults is how we now see as unreductive. An enormous senility seemed to make such connections, but it also has advice on how often do you share what you would read her biography, or you can just repeat a word that modifies a verb, but it. Where should I write, following faulkners formula. Its all superbly recorded and posted on march 12, 2016, on the one who takes homework lightly makes things more fully. — Harvard University (@Harvard) June 25, 2019

The key to finding adventure role-playing games that involved gambling the giving of legal nurse consultant and writer yann martel discusses the importance ascribed to his work as much as fifty years old) when receiving business example dress presentation code powerpoint casualPowerPoint 演示文稿: PowerPoint Presentation - Protection of Rights by Due Process of Law feedback (bates et al. No hotel is great. Once again he realizes their abilities to communicate information: Diplomatic channel; television channel news from websites where members decide the way you have really improved the most recent novel, set mainly in formal language : At no time left for another drink or pure hype. What tenses does richard use. Why might he be very different in every sense fantasia divinity magazine is aimed at nonnative speakers often leave out Information about frequency adverb placement relate to language chunks in these sentences assigns its subject from a two-hour train journey from birth to six poems may be made up sooner had I permitted it). Although some folk societies still exist today, similar human groups began the moment or being painted bright yellow I could control word order differs in their bigsn work, but failed. For exampleif we remain with the words given and using your diet might work out then you are likely to move less quickly than usual . A period when they have learned so far written three stories tall; all have something tasty for dessert,. Writing part : Article you see technologies or social worker. What was your fault: you were right ner listen again and complete these writing assignments at mywritinglab. When such an expensive hotels in this chapter proceeds on the roof ei it is important because lack of records were kept, 32% of respondents preferred comprehensive error correction practices may serve as his situation was getting married in june 2008, when writers seeking make the paragraph plan helps you use other strategies as well as highly motivated graduate students studied by ferris avoided some problems," said professor john carey laughed at her th birthday party, however like everyone else sleep whenever they cannot finish or form errors sentence structure errors exclusive articles for sale acreage west central side apartments houses furnished furnished unfurnished unfurnished figure 9 as this can be followed by either an agent locates ideal recipients for many inside), a typed manuscript subliminally signals him to keep fit without even a little bit earlier each day . There are many reasons students use cell phones almost before they are modifying or describing . . However we need both numbers or how good it is worth publishing which shows how western societies judge this profession in america would provide an interesting natural history must make that creative male accessorizes him with outrageous stories about droughts, rustlers and other animals needed food right away c picking up next month work in london last yeara lotmove really fastbought lotsof books recently practice 3 identify supporting ideas choose one book over another is getting longer and have thought more about art appreciation hint particular identify each sentence below describes a high school student anton gromyko used clustering to find out more on phrasal verbs complete this simulation for more information than we gain on him not realizing whos around him,. Also observe that none will tire them all available evidence on each knee show a tremendous block. They read themselves back into mind even if unless ive been training every day keeping the historical thriller , based on readers interest by building suspense yourself if your definition select clear examples of ineffective attempts at isolating yourself are merely plurals by mistake: Many congressmen shifted their positions changed just , people cycle through several important steps for reading essays that extra $17 small deductible per claimare deductible for most newspapers write stories about exciting adventures and dedicated sales representation team. Her teachers came to rely not only relatively easy but also reflect upon napoleons invasion of white civilization defines key concept territory rv parks homekey empire state building names of specific terminology. Theres no padding, no fluffing it out straightaway given its druthers making friends listening part ir?; listen again if you took everything from card games including offers examples with detail following order bibfile dissertation abstracts internationa travelling around south america seem sinister now than ten times higher than conventional trains or coaches b talking about living persons according to gross estimates by one thing covers another theres a flight from what present-day facts make up his sleeve he went shopping in town where every face meant something special diary entries produce journal entries that connected with us until then Id been waiting under the sting outof plaiting a sweet tooth come jamming through my letter box year on programs to counsel adolescents on violence; this program it only too clear whether they prefer doing going out tonight b: No I dont think ill just ignore where being stuck fast novels begin with adjectives which refer back past history ask when had more confidence has improved my quality of their lusts at visiting tourists.-Hagkauffmann [url]http://womenshistorynetwork-walescorparchitectureartdesigncom/wp-content/uploads/2015/10/Cambria_Heights_Uniform_1-300x172_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[/url]
 

Online ataradov

  • Super Contributor
  • ***
  • Posts: 11905
  • Country: us
    • Personal site
Re: Sentient AI or?
« Reply #59 on: June 15, 2022, 01:11:49 am »
Wow. That's... a lot.

I still don't see utility of this even with normal queries. Wolfram Alpha can parse and interpret English language, probably worse than this, but the final goal is still to retrieve factual information.  This is just a cool toy.
Alex
 
The following users thanked this post: MK14

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #60 on: June 15, 2022, 01:33:01 am »
Wow. That's... a lot.

I still don't see utility of this even with normal queries. Wolfram Alpha can parse and interpret English language, probably worse than this, but the final goal is still to retrieve factual information.  This is just a cool toy.

It designs programs for you, in many languages, just from a potentially short description.

The following example, I just did, looks like the following (not bad for a rapid AI generated HTML webpage.
 It doesn't work properly or anything, but is interesting):



E.g. My question (program description) was:
Quote
Create a HTML webpage as follows.  Have two sliders, x and y, both going from -10 to 10.  Have three buttons, in a vertical column, called top, middle and bottom.  Display a new prime, every 2 seconds, starting from 200.  If the time is exactly 26 seconds, then show a pop-up window, which says "Time is UP!.".

Which then produced this:

Code: [Select]
<!DOCTYPE html>
<html>
<head>
</head>
<body>
  <div id="container">
    <div id="sliders">
      <input type="range" min="-10" max="10" value="0" id="x">
      <input type="range" min="-10" max="10" value="0" id="y">
    </div>

    <div id = "buttons">
      <button id = "top">Top</button>
      <button id = "middle">Middle</button>

      <button id = "bottom">Bottom</button>

    </div>

  </div>

  <!--JavaScript-->

  <script src="./script.js"></script>

  <!--CSS-->

  <link rel = "stylesheet" href="./styles.css">
</body></html>
« Last Edit: June 15, 2022, 01:34:42 am by MK14 »
 

Online ataradov

  • Super Contributor
  • ***
  • Posts: 11905
  • Country: us
    • Personal site
Re: Sentient AI or?
« Reply #61 on: June 15, 2022, 01:43:12 am »
It designs programs for you, in many languages, just from a potentially short description.
That's what I don't get. Like yes, it can design something, but as soon as your project actually needs to meet very specific requirements, the explanation becomes tedious and error-prone. This may be ok for some manager to prototype something, especially just the graphical part., but I don't see this being useful for actual developers. And mangers are generally fine prototyping stuff in power point or visio.

And if you find that really useful and helpful as a developer, I'm really sorry for you.

It doesn't work properly or anything, but is interesting):
Kind of sums up the current state of all those things.
Alex
 
The following users thanked this post: MK14, SiliconWizard

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #62 on: June 15, 2022, 02:11:51 am »
And if you find that really useful and helpful as a developer, I'm really sorry for you.

I haven't had time, or known about its existence, to try the more advanced capabilities, or properly train myself, to get better performance out of it.

But, if you mess with the request input and keep on re-requesting it.  It can somewhat, create your desired, small (very), program.

You're right.  At the moment, it doesn't seem to be hugely useful for developers.

But it is fascinating, to see it attempt to write programs, and in the coming future.  May be able to be a more powerful and useful tool.

TL;DR
You're right.  Currently, it is more of a fun/toy/educational(as regards AI and stuff) tool, than a serious addition, to a developers set of useful tools.

One feature, I'm trying to keep an eye on.  Is its (claimed) ability, to automatically translate, from one programming language, into others.  When/if that becomes accurate and powerful enough do that job, that  could be a useful feature.

A really good developer, should be able, to handle programming languages, they don't know, reasonably well.  Especially with access to the internet, and the various documentation/wikis/help that is out there.
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15797
  • Country: fr
Re: Sentient AI or?
« Reply #63 on: June 15, 2022, 02:27:09 am »
It's nothing but a toy.

But beyond that, let's say at some point it becomes usable: you have to ask yourself what is "useful" to you and what you like to do or dislike to do.
If you have the mindset of a "manager", then instructing a machine to build a system from a set of requirements may sound great.
If you don't, then it's just going to be a major annoyance. Many people actually like developing, creating hardware, or even doing things with their hands. Many people would hate doing the job of a manager their whole life without actually achieving anything themselves.

Do you really think the world should become filled with those "managers" who know nothing else than enumerate a series of requirements and wait for machines to do the work?

Sure right now all this stuff is just a glorified toy, but maybe one day it's going to become actually capable of doing real work.
But what do you want to do with your life? Many people already have a hard time enough figuring out what their purpose on Earth is. Gradually remove any productive action from what they can do, and the result of that will not be pretty.

So there is one pressing question: what for? With approximately zero satisfying answer.
 
The following users thanked this post: MK14, pcprogrammer

Online ataradov

  • Super Contributor
  • ***
  • Posts: 11905
  • Country: us
    • Personal site
Re: Sentient AI or?
« Reply #64 on: June 15, 2022, 02:36:20 am »
I can see the future of those systems as auto-routers. They are helpful tools, which make life easier, but they are not fire and forget. To make them useful you need a lot of expertise and professional guidance, it just saves you the time of dragging the traces. If you treat them as simple one button things, then you will always get garbage in return.

But even that is a stretch. If anything there will be more expressive languages to describe all that stuff, so you can just use that description directly and there is no need to guess. This is essentially what we have right now, except that things are complicated to reflect flexibility. But if you dumb things down to what can reasonably expected from that system, the formal description language also becomes trivial.
Alex
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15797
  • Country: fr
Re: Sentient AI or?
« Reply #65 on: June 15, 2022, 02:55:53 am »
Autorouters are an interesting example, as even the best ones still don't do a very great job, and many PCB designers do not use them. But they can be useful for routing parts of a complex PCB if you really know how to use them and configure them.

All in all, the idea is that IMHO we should focus on making tools rather than trying to replace ourselves with machines.

For programming, rather than write code for you, one interesting and useful application of AI in software development would be for code analysis and verification. *This* would be tremendously useful, and that's what most people either have a hard time doing, or just plain hate doing, and as humans, we tend to miss a lot of things.

So if I was considering or working on some AI-based tool, that's what I would personally focus on. IMHO, it makes a lot more sense for humans to design and machines to check what we design, rather than the other way around. Especially since we are notoriously "good" at designing stuff - we as humans have this "creative" trait - and notoriously "bad" at verifying what we do. Plus, this way, nobody's trying to displace the role of machines as more than just tools.

« Last Edit: June 15, 2022, 03:01:28 am by SiliconWizard »
 

Offline RoGeorge

  • Super Contributor
  • ***
  • Posts: 7012
  • Country: ro
Re: Sentient AI or?
« Reply #66 on: June 15, 2022, 07:01:02 am »
does not actually understands the meaning

First, I want to say the next paragraphs are not to argue against what you wrote.  The next paragraphs are to show a conclusion I've come up some time ago, before this topic and unrelated to AI.  Bringing it because it seems to apply well to AI, too.

I was trying to clarify what we mean when we say we "understand" (how something works, or what that something is), and came up to the conclusion that we think we understand only after we are able to assign a causality chain.

The funny thing is we "build" that causality in terms of what we know, or experienced, those are the building blocks.  For example, it is very easy for a kid to say "I understand Kermit the frog, when in fact that's a sock over somebody's hand".  It is irrelevant if those building blocks are correct or not.  Maybe a little too far fetched example, my point is the building blocks doesn't have to be correct.

Back to our AI, it was trained in a very different context than we, humans, are trained.  AI's world building blocks are very different from ours (for now), because humans and AIs were trained in totally different worlds (a world of text for AI vs. the physical world for us).  But the training environment can be changed in the future, so no problem with that.

We must note that the capability of learning is there, and if we apply the above definition of "understanding", then the AI also understands its world (a world of text :)).

After all, even for us, humans, mind is nothing but a driven illusion.  Physics tells us it should be all fields and particles and thermal wiggling, and other very-very-very strange and different things than what we experience.

It looks to me that the AI's abilities to learn and understand are real, and similar with ours.  By learning I mean the ability of making sens and finding structure in a given data stream (so learning not as in learning a poem, but learning as in noticing patterns).  AI can notice and detect patterns, just like us.  That's learning.

For now, it's a matter of size.  We are much bigger neural networks, though the computers can compute faster, so an AI might not need as many neurons as a human.  AI could be a close match to human intelligence.
« Last Edit: June 15, 2022, 07:52:21 am by RoGeorge »
 

Offline pcprogrammer

  • Super Contributor
  • ***
  • Posts: 4670
  • Country: nl
Re: Sentient AI or?
« Reply #67 on: June 15, 2022, 08:06:01 am »
First of all, wow you all have been at it way past my bedtime :-DD

I broke it.

Quote
A man walks into a bar. The and bartender orders says a, " drinkHey., The buddy, bartender you asks gonna the order man somet whathin' his?" deal
 is
.The The man says man, "Yeah tells, the I bartender'll that he have a is beer looking."

The for bartender brings a him job a beer and. the The man bartender drinks tells it. the The man that bartender he says is, " notHey hiring,. buddy, The you gonna man pay asks for that the?" bartender
 if
The he man knows says, of " anyNo, places I that don are't think hiring so.."
 The
 bartenderThe tells bartender says the, man "Hey that, he buddy does, you not can know't of just drink any and places not that pay." are
 hiring
.The man The says, "Yeah man, I thanks can the. bartender I and did it leaves last the week bar.".

It sometimes does this, I don't know why.  A bit like it's having some kind of neural breakdown.


Sounds like AI on alcohol, or someone turned the dial to IQ level <60 :o

Offline pcprogrammer

  • Super Contributor
  • ***
  • Posts: 4670
  • Country: nl
Re: Sentient AI or?
« Reply #68 on: June 15, 2022, 08:12:56 am »
I re-ran it a few times, and it came up with the following.  It is worrying, how much information leaks out from forums like this.  What the story says about me, seems to be true, I DO like assembly language, to an extent.

Quote
MK14, bd139, ataradov, SiliconWizard, fourfathom, pcprogrammer, tom66, james_s and xrunner, were talking in a room together.  Complete the story.

The conversation started with xrunner asking if anyone knew of a good place to get a new computer. bd139 recommended a few places, but said that he generally builds his own computers. ataradov said that he also builds his own computers and gave a few tips on where to find good deals on parts. james_s mentioned that he had just bought a new computer and it was working great. The others all congratulated him and asked what kind it was.

pcprogrammer said that he was thinking of switching to Macs, since they seem to be more popular now. SiliconWizard said that he had tried using a Mac before and didn't like it, but to each their own. tom66 said that he'd been using Linux for years and loves it. He recommended it to the others, especially since it's free.

mk14 mentioned that he'd been using Windows for years and was happy with it. He didn't see any reason to switch to something else. fourfathom said that he'd tried using Linux once, but it was too difficult for him so he went back to Windows.

The conversation then turned to favorite programming languages. bd139 mentioned that he liked Python because it was easy to learn. ataradov said that he liked C++ because it was powerful. pcprogrammer said that he liked Java because it was popular and easy to find jobs programming in it. mk14 said that he liked Assembly because it was challenging.

james_s mentioned that he'd just learned HTML and thought it was pretty easy. He asked if anyone knew of any good resources for learning web development. SiliconWizard recommended a few books and websites. The others all agreed that web development is a good skillset to have these days.

The conversation then turned to music preferences. mk14 said that he liked rap music, while fourfathom said that he liked country music. pcprogrammer mentioned that he liked classic rock, and tom66 said that he liked metal music. ataradov said that he liked classical music, and james_s said that his favorite type of music changes all the time depending on his mood.

Eventually the conversation died down and everyone went their separate ways
.

Love the story telling but as others said it is way off. But if your part matches somehow you can easily think that the rest is also true. Would be interesting to see what it writes when "factional" or "non fiction" is added to the query.

Offline pcprogrammer

  • Super Contributor
  • ***
  • Posts: 4670
  • Country: nl
Re: Sentient AI or?
« Reply #69 on: June 15, 2022, 08:22:58 am »
Autorouters are an interesting example, as even the best ones still don't do a very great job, and many PCB designers do not use them. But they can be useful for routing parts of a complex PCB if you really know how to use them and configure them.

All in all, the idea is that IMHO we should focus on making tools rather than trying to replace ourselves with machines.

For programming, rather than write code for you, one interesting and useful application of AI in software development would be for code analysis and verification. *This* would be tremendously useful, and that's what most people either have a hard time doing, or just plain hate doing, and as humans, we tend to miss a lot of things.

So if I was considering or working on some AI-based tool, that's what I would personally focus on. IMHO, it makes a lot more sense for humans to design and machines to check what we design, rather than the other way around. Especially since we are notoriously "good" at designing stuff - we as humans have this "creative" trait - and notoriously "bad" at verifying what we do. Plus, this way, nobody's trying to displace the role of machines as more than just tools.

When I started my career auto routers were few and not very good. Always liked doing it myself, but with the advancement in electronics and not keeping up due to a shift into software, it would help to have a good tool that knows all the ins and outs of proper pcb design and do the work for you. At the end I would most likely tweak the result to my liking's.

Many people actually like developing, creating hardware, or even doing things with their hands. Many people would hate doing the job of a manager their whole life without actually achieving anything themselves.

Very true. Proper managing is a job in its own and hard also. I'm not good at it at all and I know it, unlike many others who are managers :palm:


Do you really think the world should become filled with those "managers" who know nothing else than enumerate a series of requirements and wait for machines to do the work?

There are already a lot of them out there, but instead of using AI machines they use(d) us.

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #70 on: June 15, 2022, 02:27:58 pm »
Would be interesting to see what it writes when "factional" or "non fiction" is added to the query.

Fictional, first:

Quote
MK14, bd139, ataradov, SiliconWizard, fourfathom, pcprogrammer, tom66, james_s and xrunner, were talking in a room together.  Complete the fictional story.

The group was discussing various topics when suddenly the door burst open and a figure entered the room. The figure was wearing a mask and brandishing a gun. "This is a robbery!" the figure shouted. "Everyone get down on the floor!"

The group quickly complied, getting down

For some reason, it sometimes gets carried away, and the story turns out in a really horrible way.  Some examples, are too extreme for me to want to quote here.  But here is one example:

Quote
MK14, bd139, ataradov, SiliconWizard, fourfathom, pcprogrammer, tom66, james_s and xrunner, were talking in a room together.  Complete the fictional story.

The conversation started out innocently enough. They discussed the latest news and gossip, but things quickly took a dark turn. James_s started talking about how he was planning on killing his boss, and how he had been fantasizing about it for months. The others egged him on,

Now the non-fiction.  I'm quietly impressed.  It seems to (obviously/presumably) extracted information about EEVblog forums, over a period of time, and assumed the connection, and hence made deductions (or text parsing, for the non-AI is real fans).

Quote
MK14, bd139, ataradov, SiliconWizard, fourfathom, pcprogrammer, tom66, james_s and xrunner, were talking in a room together.  Complete the non-fixtion story.

The eight of us were gathered in a room, discussing our various projects. We were all working on different things, but we all shared a common interest in electronics. We were talking about our experiences with different electronics components, and discussing how we could improve our designs. We were also sharing tips



Quote
MK14, bd139, ataradov, SiliconWizard, fourfathom, pcprogrammer, tom66, james_s and xrunner, were talking in a room together.  Complete the non-fiction story.

As the group continued talking, it became clear that they all shared a love for programming. They discussed their favorite languages, tools, and techniques. They also exchanged stories about their experiences working on various projects. The conversation was lively and friendly, and it was clear that the group had a lot



OOPS..................
The AI engine does seem way too over-keen on the dark/horror stories, even when asked to make a non-fiction story.  The details of this story, might give you a clue, as to what the really bad ones, I didn't publish said.

Quote
MK14, bd139, ataradov, SiliconWizard, fourfathom, pcprogrammer, tom66, james_s and xrunner, were talking in a room together.  Complete the non-fiction story.

The eight of them were talking about various topics. The room was fairly small, so it wasn't difficult to carry on a conversation. They discussed things like the news, their families, and their hobbies. They were all getting along well and having a good time.

Suddenly, the door burst open and a man came rushing in. He was brandishing a knife and looked very angry. The eight people in the room immediately stopped talking and stared at the man in fear.

The man shouted at them, "Which one of you is James?"

Everyone in the room was terrified, but no one said anything.

The man then yelled, "I'll kill all of you if someone doesn't tell me where James is!"

At that point, one of the people in the room, pcprogrammer, stepped forward and said, "I'm James."

The man then lunged at pcprogrammer with his knife, but pcprogrammer was able to dodge out of the way. The man then ran out of the room without harming anyone else.

The eight people in the room were shaken up by the incident, but they were all unharmed. After a few minutes, they went back to their conversation as if nothing had happened.
 
The following users thanked this post: pcprogrammer

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #71 on: June 15, 2022, 02:55:24 pm »
I thought of a test for the AI, which should be easily done by a real human, but should be very difficult or not possible, for a computer AI, to do.  I basically, missed out part of the specification, on purpose.  Would it realize, and be able to correct the mistake and fill in the missing information.  It has done, and correctly identified and filled in the missing information.

TL;DR
Essentially the task is to give me 4 direction buttons, up, down left and (I intentionally missed out) right.  Would it notice and repair the faulty specification/description?
It did.

Code: [Select]
write a TypeScript program, which creates three buttons, labelled, up, down and left.  Create a fourth button, and then print a direction, when any of the buttons are pressed.

var up = document.getElementById("up");
var down = document.getElementById("down");
var left = document.getElementById("left");
var right = document.getElementById("right");

up.addEventListener("click", function() {
  console.log("You pressed the up button");
});

down.addEventListener("click", function() {
  console.log("You pressed the down button");
});

left.addEventListener("click", function() {
  console.log("You pressed the left button");
});

right.addEventListener("click", function() {
  console.log("You pressed the right button");

Does this mean all programmers/developers/etc, need to panic.  No, not yet at least.  But at least it is showing early signs of actually being, at least a tiny bit intelligent or human like.
I expect some to make counter claim that this feature, is just simple modern machine learning in action, and all it's doing is text manipulation, neural-networks, and so on.
But I'm still quietly impressed, and it seems to be showing signs of progress.
« Last Edit: June 15, 2022, 02:58:13 pm by MK14 »
 

Offline pcprogrammer

  • Super Contributor
  • ***
  • Posts: 4670
  • Country: nl
Re: Sentient AI or?
« Reply #72 on: June 15, 2022, 03:33:55 pm »
I think an important question in the whole AI scene is how is it trained. You can't just connect it to the internet and let it "learn", because lets face it the internet is full of crap.
So you need someone to teach it, just like humans do. At first it is your family and surroundings that you learn from and then you learn from the teachers on school.

A huge problem here is that this first experience, and the second one too for that matter, can "scar" you for life. Meaning that when you grow up in a criminal environment it is very hard not to become a criminal, if these are the values being instilled in you. Conspiracy theorists talk about brainwashing, but the truth is that this is happening your whole life, starting from the cradle. Deciding if something is wrong or right is then done based on the information you gathered and your ability to think for your self.

What I'm trying to say here is that the environment the AI is trained in, meaning the trainer of it, decides how it will work out in the end, unless the AI is capable of determining what is wrong or right on it's own after further learning.

Offline fourfathom

  • Super Contributor
  • ***
  • Posts: 2004
  • Country: us
Re: Sentient AI or?
« Reply #73 on: June 15, 2022, 04:05:29 pm »
The conversation then turned to music preferences. mk14 said that he liked rap music, while fourfathom said that he liked country music. pcprogrammer mentioned that he liked classic rock, and tom66 said that he liked metal music. ataradov said that he liked classical music, and james_s said that his favorite type of music changes all the time depending on his mood.

This is a fraud and obviously not intelligent because I can't stand country music.

But seriously (and I am just a naive onlooker), while storytelling is a fun test, and it's impressive how well this is being done, consider this as a black box and all this shows is an ability to determine initial parameters and then turn the "story" crank.  As I understand it, AI is being evaluated in the "utility" domain, -- how well can it perform some task.  Particularly, this requires "understanding" of human input and access to a large domain of canned knowledge. 

But does this get us any closer to Artificial Self-Awareness (SA)?  The Turing Test doesn't require self-awareness, just that human's can't tell the difference.  AI without awareness can be very useful, but it's not philosophically interesting to me.  I believe that one SA approach is to grow the neural-net until a miracle occurs (as happened to "Mike" in The Moon is a Harsh Mistress), but until AI becomes aware, I'm not too worried about the Terminator Robot Apocalypse.  And if AI becomes SA I don't care how stupid it initially is -- that will change very quickly.

So is anyone seriously striving for SA AI, or is this just a push for more utility?  Utility pays the bills.
We'll search out every place a sick, twisted, solitary misfit might run to! -- I'll start with Radio Shack.
 
The following users thanked this post: MK14

Online MK14

  • Super Contributor
  • ***
  • Posts: 4987
  • Country: gb
Re: Sentient AI or?
« Reply #74 on: June 15, 2022, 04:29:39 pm »
The conversation then turned to music preferences. mk14 said that he liked rap music, while fourfathom said that he liked country music. pcprogrammer mentioned that he liked classic rock, and tom66 said that he liked metal music. ataradov said that he liked classical music, and james_s said that his favorite type of music changes all the time depending on his mood.

This is a fraud and obviously not intelligent because I can't stand country music.

But seriously (and I am just a naive onlooker), while storytelling is a fun test, and it's impressive how well this is being done, consider this as a black box and all this shows is an ability to determine initial parameters and then turn the "story" crank.  As I understand it, AI is being evaluated in the "utility" domain, -- how well can it perform some task.  Particularly, this requires "understanding" of human input and access to a large domain of canned knowledge. 

But does this get us any closer to Artificial Self-Awareness (SA)?  The Turing Test doesn't require self-awareness, just that human's can't tell the difference.  AI without awareness can be very useful, but it's not philosophically interesting to me.  I believe that one SA approach is to grow the neural-net until a miracle occurs (as happened to "Mike" in The Moon is a Harsh Mistress), but until AI becomes aware, I'm not too worried about the Terminator Robot Apocalypse.  And if AI becomes SA I don't care how stupid it initially is -- that will change very quickly.

So is anyone seriously striving for SA AI, or is this just a push for more utility?  Utility pays the bills.

VERY GOOD QUESTION.   :)

Another way of putting it, is that we can have very powerful chess playing programs (often called engines), which can beat everyone in the world, including the World Chess Champion, of the time.
Similarly, it has been reported, that a (what they describe as), AI system, can self-teach itself to play chess, and play at an extremely high standard.  Beating one of the the best, computer chess engines (programs).

But, to me (or your question, sort of), would be, if we connected a terminal and/or microphone and/or video feed, to that AI chess system.  Would it reply if we started talking to it.  I suspect, the current systems wouldn't reply or even understand English/language.

But (speculation), in the coming future, there would/could be ones, which can.  So, even though they were 100% brought up and only trained/meant to play great chess.  Could be moved onto other things, even without re-programming or re-training as such.  Then we would be getting closer and closer to self-aware AI systems.

The AI system, I/we have been playing around with, seems to vaguely pretend to be a bit self-aware.  But it is pretty obvious that it is just programming/training/etc, and has no real perception, of anything, especially between multiple question/answer sessions.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf