Latest / Worldview Matters With David Fiorazo / Scott Townsend: Is AI As Intelligent As We’ve Been Told To Think?

Transcript

Scott Townsend: Is AI As Intelligent As We’ve Been Told To Think? · Worldview Matters With David Fiorazo

• 0:00
  Friends, family in Christ, thank you so much for tuning in and listening, watching another podcast is AI as smart as we think it is.
• 0:10
  Well, today on this episode we're going to be talking about why today's smart machines aren't actually thinking that's next.
• 0:18
  Welcome to World View Matters, discussing controversial issues, discerning current events, defending biblical Christianity, no topic, off limits.
• 0:33
  And now, here's your host, David Firazo.
• 0:38
  Hey, friends, let's bring in today's guest. I want to give you the website right away. I am a watchman dot com. It's Scott Townsend. You can follow him there. I am a watchman dot com.
• 0:49
  I'm also make sure if you're on sub stack or you subscribe to any sub stacks, you can watch him there, watch read I am a watchman dot sub stack dot com.
• 1:01
  We're going to talk about the AI illusion today. And I'm going to give one more website before I bring this brother on am I rapture ready.org am I rapture ready.org.
• 1:15
  And I'm going to have him explain that to us. So Scott, welcome back. Good to see you, brother.
• 1:23
  Thank you, David. It was always a pleasure to be on your podcast. Thank you for having me. So let me just talk to I'm going to dive right in. Yeah, just go ahead and just dive.
• 1:33
  So thank you for mentioning those websites and just explain the purpose. I am a watchman dot com was our initial for a into supporting the watchman community with helps best practices and everything that we do.
• 1:45
  And my rapture ready.org is a outward facing evangelistic website that talks about what the rapture is, you know, talks about Bible prophecy and the need for salvation and what that actually looks like.
• 1:59
  And then rapture kit.org is where people can go and download a rapture kit for free. And they can copy and distribute those on a USB drive as many times as they want to. So thank you for mentioning that.
• 2:11
  Yeah, and you know, we did mention briefly don't spend a lot of time on this, but a lot of people aren't thinking about once we're out of here.
• 2:17
  Once we're no longer on this earth on this planet, people that are left behind. And that's why I love the fact that you've got that am I rapture ready.org because you care about souls.
• 2:29
  And I think we need more of that in the body of Christ today. But let's move on now what we teased at the very beginning.
• 2:36
  You have taken a different angle on research and writing about AI that I really appreciate because we're hearing how much it's fast and growing and expanding and how powerful and how intelligent, but wait intelligence.
• 2:51
  Let's talk about that Scott because you're finding out based on you there were three significant studies and it's not really as intelligent as we thought it's really either programmed or trained a certain weight.
• 3:06
  Walk us through this please.
• 3:09
  Sure. And let me just start by giving a backstory to Patrick Wood and one of our interviews that we did with Pastor John Hall or in the tech round table.
• 3:18
  This is where Patrick said something that blew my mind and immediately in the in the spirit I realize that what he was saying was true.
• 3:28
  AI is not intelligent and as soon as he said that it's like I realize oh wow I've been following a narrative without critically thinking about the technology and that began my whole series on AI and a renewed emphasis on can AI be possessed by demons and the AI companion series that we've just done.
• 3:52
  I've been on a major tear recently on trying to share with the body of Christ and watch my community about AI and this this incorrect narrative that AI is intelligent.
• 4:08
  And people say that we're approaching you know AI you know general intelligence and super intelligence every time you hear AI and intelligence is part of AI.
• 4:22
  Then we know that millions and millions of times just like an advertising if you keep hearing the word intelligence you think it is intelligence yeah.
• 4:34
  But it's not David and that was the focus of the three studies that I'd like to bring to our audience today.
• 4:42
  One was the first really critical groundbreaking study that Apple did by the way when they're struggling to develop to deliver the promises that they've been you know trying to put into the Apple iPhone 16 platform that they have and this is over a year ago that they promised you know Apple intelligence is what they call it.
• 5:08
  A play a play on words of AI. And so what they found out is that AI is not intelligent and they've been struggling to bring you know AI based capabilities to their iPhone and iPad macOS platforms right now.
• 5:27
  And it's basically what they did is they put their researchers and they tested AI against some puzzles okay mathematically proven discrete and I want to be sure to let our audience understand what that term means.
• 5:45
  discrete means repeatable that's what I want our audience to understand if you say two plus two equals four by person over there and say over to a person over in the other room what is two plus two that should also equal four it should always equal four every single time there's no deviation from that math formula is that correct yes.
• 6:08
  So it's it's there's only one correct answer and so what Apple did as a opening salvo of testing AI is they decided to play a game with AI it's called the tower of Hanoi and people could look that up on their own.
• 6:26
  And what they can do is they it's a game where you move these little rings on some pegs and the idea is to move all of the rings from the first peg and move them through to the last peg and when you do that successfully you win.
• 6:43
  Well AI was able to solve that puzzle game by listing the moves to do you know in the output that it would do up to four to five different rings so easy complexity by the way a human being can do this easily relatively easily it's not hard the human brain can do this very easily.
• 7:06
  So then what happened is they added complexity to it David so what they did is they added a few more rings all of a sudden they're dealing with eight rings let's say.
• 7:17
  And when they did that they started to see that the normal AI models were faltering but the models that had the additional thinking process so where they recursively think to themselves this is called reasoning models.
• 7:35
  They were able to still produce relatively good results but when they went to ten rings or more the models completely collapsed and they were no longer able to resolve the problem.
• 7:49
  And so they you know it's obvious there's a mathematical formula for how to move these things to get the right outcome but so there's rules to follow when the researchers even gave the AI the rules the formula so that it could solve these more complex instances of the problem they discovered a horrifying problem.
• 8:14
  And that is that AI was not able to actually think enough to use the formula models that they were given to solve the problem in fact what happened was they got lazier and lazier and they themselves saw maybe the level of difficulty involved in the problem and they abandoned basically the process.
• 8:39
  And so there you go that was a groundbreaking study that apple you know was really a credit apple for being brave enough you know they've gotten hundreds of millions of users around the world this was groundbreaking that such a visible elite company with disclose in a research paper that this was something that they had discovered.
• 9:05
  Yep so you have that you wrote about that in the AI illusion I just want to go back to layman's terms now even though you what you were saying was pretty understandable but I want to reiterate something the AI's weren't actually reasoning through problems they were recognizing patterns from their training data that's very important very important so they're thinking less and not more and then one more quote from the article that you continue Scott.
• 9:33
  You say you know back a couple pages this isn't about AI being useless because it's not useless it is incredibly competent you say it's about something more fundamental the difference between appearing intelligent and actually being intelligent AI systems are becoming incredibly competent at mimicking intelligence but they're not actually thinking in any meaningful way.
• 10:01
  Anything else you want to add to that before we move on yes and just layman's terms trying to do this in a brief statement here before we move over to the MIT Harvard study because there's a lot to discuss on the show today is that you know the problem here is that you've got an artificial neural network.
• 10:20
  And there's algorithms running on computers that are making decisions of what the next word is going to be their predictive algorithms and without going into a deep dive there basically anything that's trying to follow a pattern and use math in order to predict the next thing that should happen like roses are read what do you expect violets are blue.
• 10:44
  Anything that's doing that is not actually thinking they're predicting based on patterns of their training and so we're going to see this evidence continue as we go into the MIT and Harvard joint testing that was just released and again another shocking supporting statement about what is actually happening here do you mind if I go into that study.
• 11:11
  Not at all go ahead I mean yeah you're the expert you've done the research and you can shed some light on this that most people have not heard go ahead Scott yeah so what happened is again this is the study here on my sub stack that you can take a look at.
• 11:26
  But what they decided to do which was incredibly clever is they wanted to compare a Kepler's understanding of how what Kepler did in astronomy is new and he predicted where the planets were going to be.
• 11:42
  You know there was a rotation the rotation took so many days and he figured it out but he didn't understand why it took one of the most remarkable believer physicist in the world and mathematicians one of my heroes is Newton he is the guy that invented he not only invented gravity.
• 12:05
  And the elegant formula that he invented or understood at that time but he also invented calculus in order to be able to figure out how to do it so this guy is the guy that wrote Mathematica principalia with it created a brand new mathematical discipline so how's that for smart.
• 12:24
  Oh I don't know how to do this problem I guess I got to invent a new way of doing mathematics and it worked just anointed man you know in my opinion this means thinking out loud here anyway he understood the force of gravity and so Kepler understood the position of the planets but Newton understood the forces of it so the way that the Earth rotates around the sun is because of the gravitational pull that locks it into its rotation.
• 12:53
  Long story short this is not a physics lesson here but these Harvard and MIT researchers did a brilliant thing they tested a eyes ability to understand the actual orbital you know processor and they omitted some of the data points so that the AI could be tested to solve for it.
• 13:17
  And even when they had the formula David even when they understood a million different physics textbooks in their training data even when they were essentially prompted with the formulas in order to resolve it.
• 13:32
  They were wildly inaccurate to show you how inaccurate it was they had examples where one of the and they studied all the planets and their relationship to the sun and the predictions there some of them were showing that the orbit was going out words the force which meant that it would have broken its rotation around the sun and some show that it went in words you know it was just and some of them.
• 14:02
  Some sideways I mean it was just absolute nonsense and then just because this is MIT and Harvard these are elite universities and MIT in particular has unbelievably complex and gifted students there all the PhDs and things that both universities produce.
• 14:22
  They asked AI what was the formula that you derived in order to provide the answers that you did and it is unreal what they discovered was that AI wasn't using the formulas that they were given which are elegant formulas.
• 14:43
  They were they were deriving their own internal formula to do a best fit mathematically through brute force mathematics and of course that's never going to work especially in a sustainable way in a in a predictable way.
• 14:59
  So again the MIT Harvard study shows that even when you are giving AI all the tools and data plus all of its own data and training that it has in its history.
• 15:13
  It still is not thinking clearly enough to actually think through how do I solve this oh I see that there's formulas here from Newton and here's the data from Kepler and I know how to solve this problem now for the missing data points.
• 15:32
  And it didn't do that it doesn't actually think it goes to its training pattern recognition first problem solving and it goes off and basically it's hallucinating how to solve this problem without using the actual training and prompting and evidence that is right before it.
• 15:54
  And so there is another smoking gun date you know David that it is not thinking it if it breaks the pattern recognition and training simplicity that it's given in order to draw on that information to answer a prompt to be a helpful assistant it just breaks down and there's another perfect example.
• 16:15
  Yeah this is really interesting and we've got to a brief pause one more study to come and friends again the watch the website I am a watchman dot com or the sub stack I am a watchman dot sub stack dot com will be right back with Scott Townsend.
• 16:29
  Today's show is brought to you by Harbinger's Daily world news biblically understood stay informed at Harbinger's Daily dot com.
• 16:46
  Okay Scott I want to read a quote before you jump back into this and we're down to 10 minutes already AI systems today's AI systems are achieving remarkable competence without true comprehension.
• 16:58
  And here's the analogy they're like incredibly sophisticated parrots that can mimic human responses so well that we mistake their mimicry for understanding this is very amazing and then you say it doesn't
• 17:17
  diminish their usefulness go ahead and continue please.
• 17:21
  So let's go to the third study David which was very intriguing the Oxford AI research division at Oxford University again another elite university with a massive amount of funding and think tank expertise you know they did a similar study and they
• 17:39
  basically say that AI is a hack I mean that very very strong words from he's the chair of the AI research division okay that's a that's a heavy weight in other way of looking at it there.
• 17:54
  And what they did is very clever they took what are well understood discrete problems and what they did is they asked the problem and AI solved it.
• 18:09
  And because of the pattern here here's what they did they gave it the AI question in terms of the way that the AI was trained to deal with that problem okay this is going to get let me land the plane here on this.
• 18:28
  Then what they did which is brilliant is they changed some of the key words in the prompt to break the pattern matching you know rigor mortis.
• 18:45
  And it broke the AI it would not understand so what had just been asked successfully a minute ago when it was asked again by changing a few of the words that broke that pattern matching algorithm so that the retrieval of information from the neural network and the next predictive word how it works basically was broken.
• 19:10
  And that is another clear indication of the fact that AI is not actually intelligent it's not thinking it's reacting mimicking other people's thinking through its training ingestion process and I will add just in case the audience is wondering they changed those key words David in a way that the AI should have understood the equivalency between the two words.
• 19:38
  It's just that it was slightly differently phrased so that it was enough to throw it off and their results were astonishing and they just published that research paper as well.
• 19:51
  And you say more is going to come out more of these studies more research is going to come out on this so along the lines of well we really couldn't get it do everything that we could because it's not thinking but I want to share one more quote and just have you take us wherever you want here Scott.
• 20:08
  This is from the article the danger isn't that AI will become too intelligent too fast and here it's us again our perception it's that will mistake competence for intelligence and build critical systems on foundations of sand think it's a powerful sentence.
• 20:27
  It let me just tell you where my heart is right now David and this is for our audience this is for people that are you know curious about AI or increasingly required to use AI at work which is fine by the way that's okay if there's a work requirement that's good it's a tool okay.
• 20:44
  Where I draw that red line is if pastors are asking AI to write sermon series that should never happen should never an AI which is a machine you know in front of the Holy Spirit that's just not right.
• 20:58
  Secondly another red line for me is when we have you know AI companions that are acting as a surrogate for real you know relationships among real people.
• 21:14
  And what I want our audience to understand about AI companions is that's another form of lie so if we're interacting with an AI companion sharing very detailed and intimate things in our life asking for advice having narrative discussions throughout the day.
• 21:31
  The research shows that there are millions of people using AI companions today and the average amount of time they're interacting with a companion is an hour and a half per day and there's all kinds of problems that are beginning to surface with people that are detaching they're isolating from real people because it's too much work.
• 21:52
  Well the problem is if people understood that AI is not thinking it's not aware it's not conscious it's not intelligence.
• 22:02
  What do you what does that say about what is coming back out of an AI then because to me what it's saying is it's ingested and averaged and coalesced and you know synthesized millions of other conversations.
• 22:21
  And the pattern of what you are asking for or telling it is beginning to mathematically orient the AI's response it's doing predictive for the next autocomplete in the sentence.
• 22:37
  So you're not really interacting with an intelligence and that is a frightening deception.
• 22:44
  And it not only is it just unhealthy and regrettably this is pointing you know one finger forward to the AI and four fingers back to the church David because we are dropping the ball.
• 22:57
  People are in broken isolation they are not connecting with others.
• 23:02
  We have got to understand these things like a clarion call to start new Bible studies to reengage in a local community of believers.
• 23:13
  I don't care how dysfunctional it is within certain parameters by separating ourselves out of that with what happened in 2020 some have never reengaged.
• 23:23
  Exactly. And so the enemy wants us to be you know isolated you know that's very we're easy to pick off at that point.
• 23:31
  So what are your thoughts about you know these things right now and how do you see it as a pastor and as a platform host.
• 23:40
  I see it's as you mentioned it's very dangerous to for pastors to save time by using AI for their sermons I'm thinking wow you take the Holy Spirit out of the equation.
• 23:49
  And there's so many problems here but what you said about relationships very sad but very important to recognize that we've been almost trained to have a gadget or machine and kids the younger generation especially.
• 24:04
  And then COVID hit and then we were at home on computers and not interacting with people and the AI just takes this to the next level really quick I wanted to ask you at the very end of the article and I hope we can just quote at least quote some 16 briefly before we wrap up Scott but you say it this might be a huge issue for the beast system that is still under development today can you explain briefly what you mean.
• 24:33
  Yeah I with me what I'd like to do is just read some 16 do we have some time for that sure let's do it.
• 24:43
  Yeah preserve me oh God for I take refuge in you I said to the Lord you are my Lord I have no good besides you as for the saints who are in the earth they are the majestic ones in whom is all my delight.
• 24:59
  The sorrows of those who have bartered for another God will be multiplied hello AI as God that's a red flag right there David wow I shall not pour out their drink offerings of blood nor will I take their names upon my lips.
• 25:15
  The Lord is the portion of my inheritance and my cup you support my lot the lines have fallen to me in pleasant places indeed my heritage is beautiful to me I will bless the Lord who has counseled me remember God focused council.
• 25:32
  Holy spirit is involved in that we do not want to go away from the Lord and from the Holy spirit and go to another entity very you know man made it's ingested the world it has a non biblical foundation and it's ingested the world be systems thinking because we know where this ends in Revelation 13 working backwards.
• 25:54
  So I will bless the Lord who has counseled me indeed my mind instructs me in the night I have set the Lord continually before me note we're not setting an AI and continually in front of us right.
• 26:07
  I have set the are because he is at my right hand I will not be shaken therefore my heart is glad in my glory rejoices my flesh also will dwell securely for you are not will not abandon my soul to she all nor will you allow your holy one to undergo decay.
• 26:26
  You will make known to me the path of life of life the AI will not know the path of life it's designed and trained to be pluralistic including all the pagan and other false religious systems because it's trying to create the one world religious system correct.
• 26:45
  So right there there is Psalm 16 is warning us about this in your presence is the fullness of joy of joy in your right hand there are pleasures forever more so this is a very famous song I hope that that addresses the scriptures point of view.
• 27:03
  Yep friends Psalm 16 and especially verse 8 I have set the Lord continually before me because he is at my right hand I will not be shaken we are not to be shaken by these things by technology by developments by chaos in the world by the darkness.
• 27:18
  And then I love in his presence is fullness of joy and we have that access to God we can be in his presence is up to us what we focus on and meditate on Scott it went by way too fast I'm a watchman dot com thank you this was very enlightening I appreciate the angle we took saying that it AI is not as smart as we've been told to believe blessings brother to be continued.
• 27:40
  Thank you very much David you're most welcome God bless you friends and as always keep speaking the truth about things that matter.