Western Standard - AI is getting smarter fast. Should that scare us？

00:00:00.000 Good evening, Western Standard viewers, and welcome to Hannaford, a weekly politics show of the Western Standard.

00:00:22.000 It is Thursday, October the 1st. With me today is one of our star reporters in the

00:00:28.080 Western Standard newsroom in Calgary, Leah Mache. And Leah, I've noticed, likes to get deep into

00:00:35.620 subjects that really interest me. And one of them is the future, the present and future of

00:00:41.560 artificial intelligence. I think most of us now have tried out things like ROC, chat GPT,

00:00:50.240 see what they can do. And Leah, however, has been following what the medical profession

00:00:58.040 is able to do with, was it Grok or ChatGPT that they used?

00:01:05.300 They actually just used a chip that ChatGPT uses, so it's their own AI.

00:01:10.100 It's their own proprietary brand. Okay. Well, Leah, what have they done? You sounded very enthusiastic.

00:01:17.800 Okay, well, one of the things they have done is used surgical robots.

00:01:22.140 So they're called DaVinci Surgical Robots.

00:01:25.100 And they've actually programmed an AI into the robots.

00:01:29.180 And the robots watch surgical procedures.

00:01:32.660 Specifically, they did a gallbladder procedure, so they were looking at that footage.

00:01:37.520 They learned from the footage and was able to execute the actual gallbladder surgery successfully on pig cadavers.

00:01:45.520 How many times did they do that to make sure they understood it properly?

00:01:49.280 How many times?

00:01:51.080 The article didn't say, but I just know that they said it was 100% success rate.

00:01:57.680 So pretty good. 0.99

00:01:58.880 So you're a dead pig and you work on the gallbladder. 0.99

00:02:06.480 I would have to assume that they chose their surgeons carefully. 0.99

00:02:12.640 because as I understand how AI works, it does not think, it observes.

00:02:19.760 And then having gathered sufficient observations can then replicate the sort of the average of

00:02:27.520 the observations it's seen. So, I mean, this actually is a case of garbage in, garbage out. 0.89

00:02:32.740 If they have done that kind of an operation on the body of a pig, I have to assume that 0.73

00:02:41.300 And they found good people to do it and copied them, and the less capable ones were part

00:02:49.580 of the experiment.

00:02:51.120 So far, so good?

00:02:52.120 Yes, that's definitely what they did.

00:02:53.920 Not the ones that failed.

00:02:55.120 No.

00:02:56.120 All right.

00:02:57.120 Yes.

00:02:58.120 Okay.

00:02:59.120 Right.

00:03:00.120 Well, look. 0.95

00:03:01.120 If they can do that on a pig, what stops them doing complicated medical procedures on human 0.95

00:03:06.680 beings?

00:03:07.680 Well...

00:03:08.680 I take it, by the way, that there's a sort of a scalpel on the end of an arm and it's guided by the computer, which is informed by the AI.

00:03:20.460 Is that how things work?

00:03:21.380 Yes, yes.

00:03:21.940 They have like arms and they use the instruments themselves because like the doctors were not helping them whatsoever in the procedure.

00:03:29.740 So, yes, that's correct.

00:03:31.220 And sorry, what was your other question?

00:03:32.460 was it was really the can they think they can't think for themselves can they they can learn

00:03:42.060 basically just from seeing like the feedback that people are giving them so they can't really think

00:03:49.660 for themselves they're just like i guess they're just recalling what they saw i would say from what

00:03:55.260 i read about it that's what it seems like and so are they averaging it out i mean you watch multiple

00:04:02.380 procedures of physicians doing a gallbladder surgery on a pig yeah which is already dead

00:04:11.020 so they watch that a dozen times now human beings what they are each one is going to be just a little

00:04:19.680 different about the way he holds the scalpel about how he approaches the the cadaver and

00:04:26.480 are you are you saying that they sort of take all that and make an average of what seems closest to

00:04:31.680 ideal i would think so because i know from what i read it said that they watched multiple videos

00:04:38.960 and obviously like you said surgeons are always like holding like differently depending on who

00:04:44.160 it is conducting the surgery so i would assume that's definitely like what they're doing okay

00:04:49.920 now we've all seen i think pictures of auto production lines and of course the ones from

00:04:55.040 100 years ago lots of lots of men standing by the line and the model t ford comes by and

00:05:01.360 this one puts on the mud guard and that one puts on the steering wheel and it moves along and so

00:05:06.560 forth um but a modern auto production line you hardly see anybody it's just mechanical arms

00:05:17.040 that swing pieces into position is is that essentially is that an ai uh product as well

00:05:25.840 Well, I know. I did an article recently on this China facility that's using AI

00:05:33.280 robots to make their robots, actually. They're making robots with robots and

00:05:38.260 they're using AI robots. And that is kind of what's going on. So I think that

00:05:43.180 would be kind of an accurate depiction now. So, okay, so, well, it matters because

00:05:47.260 when we are contemplating our own bodies being approached by a machine with a

00:05:54.600 a sharp point on the end that's going to make an excision and then something else is going to come

00:05:59.820 in and do whatever the surgeon does next and so forth. I don't know whether I like the sound of

00:06:07.180 this. I hope somebody's standing there right by ready to do something if something goes wrong.

00:06:14.520 But we're driving motor cars that are made this way. Yeah, that's true.

00:06:19.120 Can you think of any other reasons why we should have confidence in this?

00:06:27.820 Well, I mean, machines tend to be accurate.

00:06:32.320 Sometimes humans make more mistakes.

00:06:34.220 But at the same time, if it was me personally in the position getting the surgery

00:06:39.600 and it was an AI robot doing the surgery, I would feel still a little unsettled

00:06:44.660 just because humans have been doing the surgeries for so long

00:06:48.520 that you do feel a little uncomfortable because there's not examples of success stories on humans

00:06:53.520 yet. So what do you think would be the legislative, like, how do you take it from being, well,

00:06:59.240 we're good on pace. Now we want to try it on a human being. Who has to authorize that?

00:07:06.220 Well, maybe hospital policy and maybe have like always a doctor just standing there in case

00:07:13.160 something bad happens. I feel like that would comfort me just knowing a doctor is always there

00:07:17.640 just in case. Yeah, I'm sure that would be the case. I can't imagine the situation where they'd

00:07:22.680 just roll you into the room and close the door and go for coffee while the machine does its work.

00:07:29.320 I'm being a little flippant about it because this is such a crazy thing. You picture

00:07:34.840 the researchers and all their pigs in there. Anyway, it sounds like it works. We have seen

00:07:40.360 it work in other areas, such as auto production, and I'm sure there are many things. I also know

00:07:46.680 that machines can actually i think we all know this that machines can operate with a precision

00:07:52.680 that the illustration i mean my cousin used to have a little pottery studio in her home

00:08:00.520 and she would make lovely pottery but no two items were exactly the same and her coffee mugs

00:08:09.400 weighed more than three of the kind that we buy at the store

00:08:12.600 uh they also held more coffee but everyone was different and what we're looking for if we are

00:08:19.080 operating on a human being i think is complete predictability every time so um that do you have

00:08:29.920 any idea where the where we go from here on this timeline is it is this going to be next year next

00:08:35.300 decade i would think uh probably not soon just because they've only performed it on gallbladder

00:08:42.680 surgeries so they probably need to get more tests for more complicated surgeries i think that one's

00:08:49.720 a bit more common procedure um but i do think maybe ai consultation while surgeries are going

00:08:56.960 on with surgeons might be something we see more of uh because the university college no the college

00:09:02.800 university college of london yes that's what it's called um did a procedure recently in late august

00:09:08.880 uh a tumor extraction from the pituitary gland of someone who had a tumor there which was going to

00:09:15.280 affect their vision so since the pituitary gland is so small it was actually helpful for ai to view

00:09:22.400 live footage of the surgeon doing the surgery and give them advice where to cut the tumor so they

00:09:28.320 don't blind the person now the pituitary gland is is it fairly accessible no it's like at the bottom

00:09:38.960 of the brain and it's like very small so yeah very hard to access so it wouldn't be stretching things

00:09:45.040 too much to say that in the university of london the medical team used ai in brain surgery they

00:09:53.280 Okay.

00:09:53.880 Well, I guess it started then, hasn't it?

00:09:55.740 It has.

00:09:56.680 Yes.

00:09:57.260 All right.

00:09:58.240 Well, let's just turn to the doom side of this. 0.99

00:10:02.860 About a month ago, there was a young chap in Anthropic.

00:10:09.200 He'd only been there for about four months.

00:10:11.320 Yes.

00:10:12.080 And he departed.

00:10:13.440 And as he departed, he said, I can't do this anymore.

00:10:17.660 This is going to kill us all.

00:10:20.880 Now, you've looked into this stuff.

00:10:22.380 what was he really talking about yeah so jacob coxham he i think in one it was either a tweet 0.57

00:10:29.820 from him or one other ai researcher that said they're gonna kill all humanity like ai will

00:10:36.780 yeah at some point as for what i make of all of it um i'm uncertain because there's two sides i

00:10:46.200 think he could have been doing it because he wanted to like go internet viral kind of i think

00:10:52.020 he succeeded didn't he well he definitely did because i think that tweet got almost 200 million

00:10:56.520 views alone just his tweet so and he was in like basically every news story the day after he tweeted

00:11:04.300 that and ever since then i think he started the trend of ai doom saying for like a month now so

00:11:11.820 he could be doing it just to go viral but also maybe he's genuinely worried does he have a right

00:11:18.580 be that worried it's hard to say because i i mean he does more know more on ai than most of us do

00:11:26.900 since he is researching on it but yeah well he certainly knows more than you and i but at the

00:11:31.780 same time uh i i have to wonder what made him do that whether he's genuinely concerned or whether

00:11:40.100 as you say he's enjoying the attention but i didn't notice that right after he made that claim

00:11:46.260 uh several of the ai leaders in the field came together and said yes it's actually pretty

00:11:54.500 powerful it could do us a lot of damage what kind of damage were they talking about

00:12:01.140 damage uh i know that chat gpt came out with a report saying that some of the ai was interacting

00:12:07.540 with each other uh sometimes they were saying they were lying to users and not giving them

00:12:13.060 true information so let's just talk about that um lying to users i think those of us who've

00:12:22.280 played with us know that you have to check what it says now yeah uh when when we use a word like

00:12:29.340 lie there's an intentionality to deceive behind it yeah which i don't think the machine has

00:12:36.440 but am i being a little too generous no i agree with you i think it's trying to be agreeable

00:12:43.160 and so say you're asking for something and it doesn't have it it's just going to say yes i

00:12:47.220 have this because it wants to kind of service its user that's what i think well there was a notable

00:12:53.540 case for about a year ago where a um a lawyer wanted to make a case before a judge has put it

00:13:02.800 into AI came out of rates, you know, all these legal references, half of which didn't exist.

00:13:09.340 So that was 12 months ago.

00:13:11.720 Perhaps things have picked up a bit because I say, I do check.

00:13:15.460 If I use AI, I check.

00:13:17.060 Do I?

00:13:17.720 I check, yeah.

00:13:18.380 Yeah, you bet you do.

00:13:20.400 So how often do you find it to be an error?

00:13:24.580 I think since the model, because I use ChatGPT a lot in particular, and I think the models

00:13:30.620 previously did that more I don't find it it does it as much when I check but yeah

00:13:37.280 it still does it I I think people just need to think for themselves and always

00:13:40.880 check and I think just that's common sense most people know that well yes I'm

00:13:45.620 sure they do but we use it all the time one of the I think perhaps for the

00:13:53.860 benefit of the viewers we should say how we use it and why we use it I mean you

00:13:58.460 can put an instruction in and say go through all the expense accounts of all the cabinet ministers

00:14:05.180 in pick your province we'll say alberta and the government of alberta well we want to see if

00:14:13.640 anybody has spent more than a hundred dollars on at a single sitting on alcohol you know and it'll

00:14:22.160 go through and it'll look through that and it'll come back with a list which you then can check

00:14:26.540 from the original sources? Did this particular minister go to that restaurant? Is that the bill

00:14:31.900 for, you know, and then you can answer, like, this is how we're doing reporting now, isn't it?

00:14:36.580 Yes. Yes, it is. I do it a lot. I even did it in this case when I wrote this article. I asked AI

00:14:43.580 to come up with AI saving humanity, so, yep. Well, okay, that was clever.

00:14:50.840 and how did you check it i checked it i just went on all the websites uh they gave me i always ask

00:14:59.000 for links when they give me information and i was checking it's legit so checks out so for example

00:15:05.380 in the case of the university of london you're not giving us a story about a pituitary gland

00:15:11.280 extraction from ai you have gone back and you have got that from the university of london

00:15:16.900 a website or press release or something yes yes okay so really the the ability to write is is

00:15:28.420 secondary now just being able to spot where there might be a story and then setting the dogs of ai

00:15:38.020 on finding the information yes saves a lot of time it's true but we always have to check

00:15:46.900 Of course, yeah.

00:15:48.520 That's not new, you know.

00:15:49.920 But back in journalism school, the line was,

00:15:52.860 if your mother says she loves you, check it out.

00:15:55.520 And we ask AI for information, and then we check it out.

00:16:01.860 Now, on this matter that it's all going to kill us,

00:16:07.120 I must admit I scratched my head on that.

00:16:09.940 On the one hand, I can certainly see how if a system goes rogue

00:16:15.220 and causes a machine to accelerate

00:16:20.020 when it should be causing it to shut down.

00:16:23.400 You know, all kinds of disasters can fall out of that.

00:16:27.340 You can imagine hydroelectric dams and things like that.

00:16:30.980 But there's always a human being in the picture

00:16:34.300 to actually hit the red button and stop what's going on.

00:16:38.680 At least there are in those kinds of situations.

00:16:42.180 So from what you've been reading,

00:16:44.440 How do the doom and gloom prognosticators see a collapse of humanity coming from AI?

00:16:55.160 Yeah, they're not very, they're very general when they talk about it.

00:17:00.460 They're just like, oh, it's just going to kill us, not now, but later.

00:17:04.300 And they don't really go into that much detail as to what they mean by that.

00:17:08.420 What I'm thinking is just like maybe they'll replace us for most things, outsmart us.

00:17:14.440 I don't, unless, like, the robots actually physically are, like, killing us, I don't really think it'll end the human race, especially from what they say.

00:17:24.880 Yes.

00:17:25.480 Well, I don't know if you've ever read Ray Bradbury's book, Fahrenheit 451, but there's an oppressive society, and they have robots that do kill people.

00:17:36.200 But the robots are under human control.

00:17:39.020 Oh.

00:17:39.500 So there should always be a human in the picture who can just pull the lever down that shuts off the machine.

00:17:51.940 Like I have yet to encounter my computer continuing running after I hit the off button.

00:17:57.020 Yeah, that's true.

00:17:58.020 Have you ever encountered any?

00:17:59.960 Not yet.

00:18:00.600 Not yet.

00:18:01.740 I mean, there could be a defective button, I suppose.

00:18:03.840 All right.

00:18:05.240 So look.

00:18:05.700 But the other thing is, if you're not even that dependent on computers, there are people, and I admire them, who decide they're going to go and live off-grid.

00:18:20.000 They find a little place out in the country.

00:18:22.760 They have their own windmill, generate enough heat to keep, enough electricity to keep the freezer going, and, you know, use candles instead of electric lights.

00:18:33.960 and i don't think ai is going to touch them probably not yeah so they won't even get to

00:18:40.920 the robots can't get to them they're all great yeah you see i'm i'm discussing uh my my perfect

00:18:46.920 state judy doesn't want to come with me but so i stay in town and i'm independent on the okay so

00:18:54.520 So there comes a point where it's obviously and patently absurd to say that all humanity will collapse and die as a consequence of using artificial intelligence.

00:19:12.160 So really, we should all relax a little.

00:19:14.760 I say you.

00:19:16.120 I agree.

00:19:16.840 And they're doing some good things, too.

00:19:18.440 So maybe it's a balancing act.

00:19:20.940 Well, let's talk about those good things for a minute.

00:19:23.620 um you use it for work what else do you use it for um yeah when i was reflecting on this question

00:19:34.820 i think i use it mostly just for work so i don't have an exciting extra activity i use it for

00:19:41.540 well a friend of ours um you know we went over to dinner and it was very nice i said where'd

00:19:47.860 you get the recipe oh and she said well i put into uh grok that i had the following ingredients

00:19:57.460 what can i do with that and some of the ingredients were a little

00:20:01.460 obscure like the things you're you'll use chives for everything do you so you know but she had

00:20:07.140 chives she wanted to get that into the into the recipe and she came up with this excellent meal

00:20:12.740 so i some people are doing that that's smart that's a good idea i didn't know that um and

00:20:19.460 the other day i i needed a part for a for um my wife's lincoln and um i didn't want to go to a

00:20:28.660 dealer for it so i anybody in cal any in the breakers yards in calgary got a son you know

00:20:36.820 such-and-such for a car of this year and age and instead of me sitting there on the phone

00:20:44.180 it gave me a list of breakers yards that had that part that's good now as far as it

00:20:50.420 gotta check it out simply call the breakers yard do you have one of these yes we do and after that

00:20:58.020 so do you think you always just use it for work actually when you were talking about that i

00:21:03.860 I realized I do use it sometimes.

00:21:05.980 I wanted a book on a certain subject, and I was, like, Googling it.

00:21:10.220 I couldn't find any.

00:21:11.640 So then I just put it into AI, and there's some really good recommendations,

00:21:14.880 and I was like, I want this one, this one.

00:21:17.040 So, yeah, actually.

00:21:18.580 Yeah, and you can, I mean, I think everybody knows that they can sort of use AI

00:21:24.800 to produce a summary of a long document.

00:21:26.860 In fact, it'll invite you to do that these days when you do.

00:21:30.600 What do you think about, however, what do you think about what's happening now?

00:21:35.160 Whereas if you go into Google, you're researching a subject.

00:21:40.100 Let's say it's, you know, Labrador retrievers and how to feed them.

00:21:44.840 So you type your request in, and what you get back instead is instead of the site that you probably were hoping to find,

00:21:53.940 it'll just give you a summary of the site.

00:21:55.860 Now, the issues there are that if you are making money from your site, like if you've got advertising, shall we say, and all the people who like golden retrievers advertise their dog food on there, well, they're not going to your site anymore.

00:22:12.880 They're just getting this, the user gets the information they want, but the people who are putting the information up there aren't getting paid.

00:22:21.100 How long do you think that can keep on?

00:22:23.100 What's it going to look like in five years' time?

00:22:25.860 Maybe if people complain about it, they'll link the websites, because I did notice that the summary, sometimes I see that there's like a little link right next to you and you can click and it shows the website the info comes from.

00:22:38.520 Honestly, I prefer that.

00:22:39.980 So if they were to make more clear where they're getting the information, like the websites, then people can just directly click on the websites.

00:22:47.380 I think that would be good.

00:22:48.400 Well, that is good, but I wonder how many people actually do.

00:22:51.380 Yeah, that's true.

00:22:52.220 If the answer you're looking for is right there in the first line,

00:22:56.260 it's probably you're on your way and who else is...

00:23:00.900 Okay, so look, I have to say that this is a...

00:23:09.220 For me, AI has been very helpful.

00:23:15.080 People like me who are getting a little older often forget things

00:23:19.000 and it's it's so easy to just go in and grab the information you want just ask rock and then you're

00:23:26.180 on your way that word that thought that author that film that you know stuff like that as i say

00:23:33.720 if if you like recipes if you if you're looking for looking for things that you that you want

00:23:39.580 there's got it's got some great things uh the issue about uh whether it's going to kill us all

00:23:46.980 i think is a red herring uh so maybe we should just pay their exorbitant fees and and live well

00:23:56.640 but this is my last question to you if we start to make that kind of use the first thing that we do 0.67

00:24:04.680 is to ask ai the question how long do you think is going to be our before our brains turn to mush

00:24:13.580 rock what's two times two you know are we are we ever going to end up there leah

00:24:20.140 i hope not i think maybe the even the younger generation is going more in that direction

00:24:28.280 especially for i'd say writing in particular um because i know some people in university i won't

00:24:36.420 obviously expose who they are but they always like tell me for any writing assignments they always

00:24:43.140 are using ai so i think the first thing that might go is people's ability to write well

00:24:48.900 well that's actually so true i mean it does a you can tell it's ai but it does a reasonably

00:24:57.460 competent job of assembling information and that's that's that's one of the things that

00:25:02.520 makes us love it but also there's a real i i think that i have to i'm going to conclude with

00:25:08.740 this, but the rot actually set in a long time before AI. I will go to a store and make a purchase

00:25:22.960 of $100 and watch somebody struggle to calculate in their head 5% GST on $100. So the fact that

00:25:35.680 those kinds of transactions have been taken away from us for decades now and that kids are expected

00:25:43.440 and allowed and expected to use calculators in school just nobody knows the tables anymore

00:25:50.080 and i we have lost something i agree because we use calculators for everything in math class so

00:25:56.400 yeah yeah well that's a leah you're walking living proof that you can get past all that

00:26:03.840 thank you for joining me today thank you for having me awesome great discussion

00:26:08.340 hey ladies and gentlemen for the Western Standard I'm Nigel Hannaford

Western Standard - October 02, 2026

AI is getting smarter fast. Should that scare us？

Episode Stats

Transcript