Transcript: Let Them Eat Tokens: AI Efficiency vs. Global Good
0:00 Brad If you're two guys in a garage or two entrepreneurs or one entrepreneur or a small, you know, resource strapped nonprofit,
0:06 Brad then maybe that's the way we should look at this productivity gain and say there is something good in that.
0:27 Scott Hello,
0:27 Scott Brad.
0:28 Scott Welcome to this week's edition of How Many CTOs Does It Take to Screw in a Lightbulb?
0:35 Scott My name is Scott Pohrad, and this is my trusty co-host.
0:41 Brad Brad Hefta-Gaub.
0:42 Scott And I don't think it's really how many CTOs does it take to screw in a light bulb.
0:47 Scott I think it's how many CTOs does it take to screw in the AI.
0:52 Brad The AI.
0:53 Brad It's all about the AI these days.
0:55 Brad So many AIs.
0:56 Scott All AI all the time.
0:59 Scott I just got off the phone talking to a startup about their AI.
1:05 Scott You told me you've seen some things on the Internet about AI this week, and I saw some things on the Internet this week about AI,
1:13 Scott and I posted something on LinkedIn that went viral about AI.
1:18 Brad There
1:19 Scott So
1:19 Brad you
1:19 Scott it
1:19 Brad go.
1:19 Scott looks like it is all AI all the time.
1:24 Brad Yep.
1:25 Brad Where should we start?
1:26 Scott We should start with Apple Siri.
1:30 Brad Okay.
1:30 Scott I asked Apple Siri something simple the other day,
1:34 Scott something that chat GPT voice back and forth could answer in one with its little pinky finger,
1:41 Scott the easiest thing.
1:42 Scott It was something along the lines of, hey Siri,
1:46 Scott what time is the kickoff of Monday Night Football?
1:48 Scott Something like that, you know, something
1:50 Brad It's
1:50 Scott so simple,
1:53 Scott right?
1:54 Scott and it couldn't do it.
1:56 Scott It just showed me the link. It just gave me a list of links from a search.
1:59 Scott And I thought to myself,
2:00 Scott how is Apple so far behind on this?
2:04 Brad Yeah, it's super weird.
2:06 Scott How
2:06 Brad It's
2:06 Scott is it
2:06 Brad super
2:07 Scott they
2:07 Brad weird.
2:07 Scott haven't just done a deal with OpenAI?
2:09 Scott They did a deal with Google and Yahoo and all those things to be the search provider.
2:14 Scott You think they could be doing a deal with those guys?
2:17 Brad I have heard nothing but the same kind of complaints that you just had about how poorly Siri looks in the context of, you know, and by the way, Siri just decided to that I wanted to talk to it. I'm not talking to you, Siri, that how how far behind it appears.
2:37 Brad Based on kind of everything else that's happening in voice based agents.
2:43 Brad And so you're not alone in the perception that it makes no sense that Apple's so far behind on this.
2:51 Scott Makes no sense.
2:54 Brad One of the theories that I have heard is that.
2:58 Brad they are being more conservative they're waiting for the trends to continue they believe that price performance is going to come way way way down and that they sort of they've been doing like they announced at their last dev days a couple months ago a bunch of essentially on device um small model stuff and you know it could be that they're
3:22 Brad You know, they don't really have a cloud business.
3:25 Brad They're one of the few, they're one of the mag sevens that don't have a cloud business really.
3:30 Brad And they're all about their devices. And so the theory that I heard was,
3:36 Brad hey, they're just focused on let's wait,
3:39 Brad just drive this, drive these models onto the devices themselves and not play in the cloud game.
3:48 Brad Now,
3:48 Brad that strategic decision on their part,
3:51 Brad I don't know if that's a good one.
3:52 Scott It's a strategic decision that I think will lose them device users.
3:57 Brad Well,
3:58 Brad to your point, like as far as you're concerned,
4:00 Brad you don't care whether Siri is actually running on the device or is on some cloud thing,
4:05 Brad something like that.
4:06 Brad So you're just like.
4:07 Brad Why? I just want this thing to work correctly and it's not working correctly.
4:12 Brad So it does feel like the device is failing you.
4:13 Scott Yeah.
4:14 Scott And if Google's putting that in their device,
4:16 Scott people are going to be like, this is better.
4:18 Scott I can just talk to this thing and it just tells me when Monday Night Football is.
4:22 Brad Yeah.
4:23 Brad Fair enough.
4:24 Scott This might be another hypothesis.
4:26 Scott Apple has always taken the position of privacy.
4:30 Scott They've tried to stake out that ground.
4:33 Scott And maybe they're not able to have a solution through Siri that meets whatever standards they have.
4:40 Brad Maybe.
4:41 Scott Okay,
4:42 Scott well, what have you seen in AI this week?
4:44 Brad Well,
4:44 Brad I saw a couple of things that I think were super interesting.
4:47 Brad One of the one of the things that came out,
4:49 Brad this was not this week, but pretty recently was this Axios, this kind of political newsletter,
4:55 Brad did a survey and this was kind of like right before the New York mayoral election.
5:02 Brad And so they basically found that.
5:08 Brad A plurality of voters in each of the eight states that are the most hotly contested swing states.
5:16 Brad So a plurality of voters in each swing states have an unfavorable impression of the AI industry.
5:22 Brad The AI will raise their energy costs. This is certainly one of the things that we're seeing and that they believe increased use of AI will make their lives worse.
5:31 Brad The higher the income,
5:32 Brad the more favorable swing voters are.
5:34 Brad are were towards ai and the more likely they were to believe that we make their lives better um and in some states republicans were essentially split on ai where independents and democrats all held unfavorable views so
5:47 Scott Hmm.
5:48 Brad this is kind of a sort of a i mean we don't usually talk about politics on on this podcast and um you know because there's so much to talk about with tech anyway but this is an interesting place where it's like
5:59 Brad like you know public opinion is going to read on on tech and and colorado just passed a new state law and california has a really restrictive state law i heard the colorado law has like
6:13 Brad individual rights to sue if you believe your in any of your like individual liberties have been uh restricted or infringed by ai so it's like some really super uh like broad wording like what what does it even mean right like my civil liberties have been have been somehow violated by ai like what does that even mean and so now what are you going to do are you just not
6:38 Brad not like you can't ship an ai-based product to colorado for fear that you're the individuals will be able to sue your company so that's kind of that's also a little bit like i mean you know that that sounds like that's going to put some kind of a some kind of a damper on ai innovation
6:55 Scott It's going to make work for AI lawyers.
6:58 Brad It could work for the lawyers.
7:00 Brad We don't know, maybe.
7:01 Scott I think the energy,
7:04 Scott I'm glad,
7:04 Scott I shouldn't say I think, I'm glad.
7:07 Scott That the energy concern is percolating because I have energy concern.
7:12 Brad Well, the cost of energy has been skyrocketing around all the locations that the big players are building,
7:20 Brad guess what,
7:21 Brad ding, ding, ding,
7:21 Brad data centers.
7:23 Brad And that tips into my next little piece of data that I learned.
7:26 Brad I think in a recent episode,
7:28 Brad you and I were talking about this and I kind of talked about this whole like challenge around like.
7:32 Brad like this circular revenue and a lot of these big players are like basically you know kind of doing these deals that that no cash changes hands but
7:40 Scott Mm
7:40 Brad they-hmm. all are both basically both booking revenue well another another piece of data is that the mag 7 um so the apple you know facebook google etc those guys uh they're apparently in the last quarter they had 60 more than 60 percent of their free cash flow was spent on
7:57 Brad on building new data centers.
7:59 Brad So they're building data centers like crazy,
8:02 Brad and those data centers are all going to consume and drive up local energy prices.
8:08 Brad So even if you ignore the potential climate impact of all this construction and energy usage,
8:18 Brad At the consumer level,
8:19 Brad like if a data center goes up near your house, your energy costs or in your state,
8:24 Brad your energy costs are going up.
8:26 Brad And so, yeah,
8:28 Brad I mean, like I said, man,
8:30 Brad the rent's too damn high.
8:32 Scott That's too damn high.
8:34 Scott I don't want to get too deep into politics, but I will add one puzzle piece to this, which I heard or read lately,
8:43 Scott which was.
8:45 Scott You know, some states, they tend to be redder states, have more, you know, liberal,
8:53 Scott easier to work with sort of zoning and land use and development laws.
8:57 Scott And then as a result, things like data centers and wind farms and solar farms and things are getting built there.
9:03 Scott And that was causing a little bit of a backlash. So it's interesting how all these things are intertwined.
9:11 Scott twined with each other.
9:13 Brad Yeah.
9:13 Scott And I wonder also,
9:15 Scott so I think I've said this before on the podcast,
9:18 Scott and if not, I'm going to say it now,
9:20 Scott which is I worry about the climate impacts of this stuff.
9:22 Scott Like is the value we're getting from the fact that I can ask the AI when Monday night football is,
9:28 Scott when I could easily just go look at the TV program and is that worth it?
9:33 Brad TV guide, still have those paper copies in the TV
9:35 Scott Yeah,
9:35 Brad guide.
9:35 Scott every week I get a little TV guide at my house.
9:38 Scott It's not, you know, you can easily go look.
9:39 Scott to look up you know just ask google what time does you could type it in what time is monday night football um and so i do worry about the the roi so to speak on the climate and then there's the other thing is are we are we seeing over investment here right i mean like is this an arms race are we going to have a a bunch of zombie debt you know like shopping malls like a bunch of zombie data centers out there
10:05 Brad Well,
10:05 Brad yeah,
10:05 Brad and I think that's kind of what I was saying before is I'm starting to hear analysts in the space who are bullish,
10:13 Brad whatever techno optimists who are saying like,
10:16 Brad oh, yeah,
10:17 Brad like, I mean, I just heard this interview with where Sam Altman was like,
10:21 Brad oh,
10:21 Brad for sure, we know that there's going to be an overbuild. We just can't predict.
10:26 Brad We have such demand right now.
10:28 Brad We're still building as fast as we can.
10:30 Brad But surely there'll be some kind of infrastructure overbuild. So like the idea that like the the the you know, the head of the parade of like, you know, AI optimism is saying, oh,
10:42 Brad I am certain that we are overbuilding infrastructure.
10:45 Brad Right.
10:46 Brad Like.
10:47 Brad That seems like a pretty big deal to me.
10:50 Scott Yeah, that's right.
10:51 Scott Well,
10:52 Scott I guess there'll be a cool thing that comes out of that,
10:54 Scott which is that we'll see the price of tokens hit the floor when basically everybody's got these capital investments that they just want to get anybody to use them.
11:03 Scott So they're going to bring their.
11:05 Scott margins basically done it's going to be the ultimate commodity unless i guess some models can really do well data centers will be the commodity because it's just power it's just electricity i guess uh the models unless they can really differentiate themselves um will be commodities too and it seems like i'm starting to get the vibe not not no pun intended that of the of the models i pay attention to it's claude that's sort of starting to differentiate itself a little bit
11:32 Brad Yeah,
11:33 Brad there's a lot to...
11:34 Brad a lot to unpack there i mean i think there's no doubt that you know another another uh interesting podcast i heard recently which there there was some there was some sort of classic uh elon problems with the statement that he made here but he was basically making the case that like of course like token prices are going to fall of course uh because everybody's working to try to make these the the more compute happen with smaller effort right like that's
12:01 Brad that's the big thing and he like compared like the amount of electricity that the human brain uses which is like ridiculously small you know it's measured in you know it's measured in watts not you know gigawatts right and how much you know he made some statement about like you know look we built all of humanity you know on this this small power consumption now
12:23 Scott Wow.
12:24 Brad that's that's ignoring that's ignoring
12:26 Brad you know the network effect of a lot of humans and and history and that's the you know language and the major thing that we do but but again i think this kind of speaks to the the whole trend of conversation right the trend of conversation in the in the leaders of the parade is all around yeah we're pretty sure we're overbuilding yeah we're all trying to basically lower the compute cost yeah we're
12:52 Brad we're all kind of like you know building like crazy right now but we think this is all gonna um the prices are all coming down and so you know that's like
13:04 Brad This is very, very similar to the the turn of the century,
13:08 Scott Mm
13:08 Brad the turn
13:09 Scott -hmm.
13:09 Brad of the millennium,
13:09 Scott Mm
13:09 Brad when
13:10 Scott -hmm.
13:10 Brad it was all about building out fiber. And
13:12 Scott All about
13:13 Brad so
13:13 Scott fiber.
13:13 Brad and,
13:14 Brad you know, like the same exact like the same exact type of just on the edge of, you know, kind of dicey financial transactions that that Enron went over the edge and,
13:28 Brad you know.
13:29 Brad Broke the rules on that kind of stuff, that circular revenue,
13:32 Brad that stuff is happening right now with these big players.
13:37 Scott But hey,
13:37 Scott the stock market's going up because of it, so we're all benefiting from that, right?
13:42 Brad unless
13:42 Scott And so,
13:42 Brad your electricity
13:43 Scott yeah,
13:43 Brad prices are going up so
13:44 Scott that's right.
13:45 Brad it's interesting i just think i just think it's an interesting thing like what what's going to happen here uh you know if again like we put our cto hat on right like there's no doubt in my mind that like my engineers are benefiting incredibly from this like they're so much faster than they were even a year ago and you know to your you mentioned that you know you kind of put
14:09 Brad kind of posted that thing on linkedin this week about like you know uh what vibe but verify right like it's for real it's happening what but is that all of a sudden is there some backlash coming even bigger backlash that's coming that's and what's going to happen then
14:25 Scott If history is a lesson to be learned from,
14:27 Scott there is a backlash coming.
14:29 Scott I don't know what it is or when it's coming,
14:32 Scott but I think it'll be coming.
14:34 Scott I see a lot of uses of AI that aren't very interesting.
14:37 Scott So for instance, on my stationary bicycle at home,
14:42 Scott there's now an AI summary of the ride that I'm about to do. It's like, really?
14:48 Scott Is that necessary?
14:50 Scott So I actually, you mentioned LinkedIn,
14:52 Scott I actually had a post on LinkedIn this week that went viral.
14:56 Scott Before I tell you about my viral poster,
14:58 Scott can I tell you about another post I saw that has nothing to do with AI,
15:02 Scott but
15:02 Brad Yeah,
15:02 Scott simply
15:02 Brad sure.
15:03 Scott has to do with sage wisdom?
15:07 Scott Kent Beck,
15:08 Scott you know the name Kent Beck?
15:09 Brad Yeah.
15:10 Scott I think Kent Beck might be a little bit of a curmudgeon, but I also think he's very wise.
15:15 Scott And he told me on LinkedIn,
15:18 Scott he wrote this message to me personally and basically to everybody else that follows him. So it wasn't really personal.
15:24 Scott He said that his washing machine told him that two undersized loads is more effective than one oversized load.
15:38 Brad Oh,
15:38 Scott And
15:39 Brad I
15:39 Scott he said, even my washing machine knows agile principles.
15:48 Brad love it.
15:48 Scott I loved it too because it's true.
15:50 Scott It's really true.
15:52 Scott Okay,
15:52 Scott I wrote a post that went viral this week.
15:55 Scott Every once in a while you just put something out there and thousands of people.
16:02 Scott glom onto it. And I started my post with these words,
16:06 Scott a rant on AI.
16:09 Scott That's probably what got everybody's attention.
16:12 Brad Yeah, I'm sure that's what pushed the algorithm.
16:14 Scott I had read something, we may have discussed it here,
16:17 Scott according to an MIT study,
16:19 Scott AI is being used most effectively for back office efficiency so businesses can cut costs and make more money.
16:28 Scott Hooray.
16:29 Scott And then I ask the question,
16:31 Scott what about doing good for the world? What about making people's lives better?
16:37 Scott I'm still looking for these kinds of AI uses.
16:40 Scott What is that's what's most interesting to me about AI.
16:43 Scott I would be grateful if you left examples in the comments and I got many comments with some that I think are good examples and some that I don't think are good examples. To me, that's the most that's that's what I'm interested in with AI.
16:58 Scott That's what I've been thinking about.
16:59 Scott You know, I think listeners of this podcast know certainly, you know.
17:03 Scott No, I was never like an AI naysayer, but I've been on the back end.
17:07 Scott I haven't been on the front edge of AI like you've been on. I've been on the back edge of AI.
17:10 Scott And where I'm at right now is like,
17:12 Scott okay,
17:12 Scott this is cool.
17:13 Scott There's lots of cool stuff happening here.
17:15 Scott Like, but I'm not really interested in like saving a minute of time for some people in some department somewhere.
17:20 Scott I'm interested in doing something that.
17:23 Scott is interesting and cool and
17:25 Brad What
17:25 Scott cheap.
17:25 Brad was the best example people gave you?
17:27 Scott Gosh, the best example,
17:30 Scott I'll just read you a few.
17:32 Scott First of all,
17:32 Scott people,
17:33 Scott technical people can sometimes become sticklers. Have you ever noticed that?
17:43 Scott Yep, that chuckle means you've noticed it, you've seen that happen before.
17:46 Scott So
17:46 Brad I've seen
17:46 Scott the
17:46 Brad it.
17:47 Scott notion of AI and ML get a little mixed up sometimes.
17:51 Brad Oh, yeah, sure.
17:53 Brad Did some of your ML buddies respond?
17:55 Scott One person said,
17:58 Scott I read about a fork, like an eating utensil that canceled out hand tremors to make it easier for people to eat.
18:08 Scott And then another person chimed in and said, well, that's really ML.
18:11 Scott That's not AI.
18:12 Brad Ha ha ha ha ha ha ha ha ha ha.
18:15 Scott Another person said,
18:17 Scott and this was interesting.
18:20 Scott And this is a person I really look up to.
18:22 Scott I know actually all these people are people I know just about another person I really look up to said to me, well,
18:31 Scott I work with a lot of resource strapped not for profits and for them saving a little back office money is really meaningful.
18:41 Brad Yeah.
18:41 Scott So that's interesting.
18:43 Scott That's a positive way to look at him.
18:44 Scott Another person actually has listed a number of examples.
18:48 Scott I actually have not gone through all of the examples that he listened to.
18:55 Scott But he just pointed out places,
18:57 Scott something called Alpha Fold, which is helping drug design.
19:02 Scott Duolingo, which,
19:03 Scott by the way, I don't know if you know this,
19:05 Scott I am a Duolingo user rocking a seven.
19:08 Scott 750 day streak on Duolingo talked about you mentioned pointed to some examples of how anthropic and open AI are trying to get into AI for science they don't really know exactly what they're doing open door which is trying to I guess solve some problems in real estate I'm not really sure that wouldn't be possible without sort of
19:33 Scott deep research ai so that's just some things that got mentioned i'm just scrolling down here um this person pointed out a company that managed to save i don't know if it was a person actually i don't know exactly who but someone saved a bunch of money in hospital bills um
19:48 Brad Yeah.
19:49 Scott here's another list of things that i'll just stop after this one another actually this person i don't think i know who they are um now see this is where i think it gets interesting right okay
19:59 Brad This person said, look, just last week he saved five hours in lawyers doing stuff,
20:06 Brad $100,000 a year in SAS costs,
20:09 Brad which allowed him to hire another person.
20:12 Brad He automated accounts payable so he could get.
20:16 Brad his vendors paid faster and so i don't know if those are actually really that interesting and then it goes on there's lots more comments here for any of for the 39 people who listen to our podcast just go to linkedin find my profile find my posts you'll find well this one
20:30 Scott your your pro your linkedin profile is always in the show notes so they just click on
20:33 Brad there
20:33 Scott that and
20:34 Brad you go there you go so i don't they know but that's the stuff i'm looking for that i think is interesting um to try and
20:40 Scott i don't want i don't want to be in that stickler department like
20:42 Scott like ai versus ml but you know carbon robotics uh their local seattle company that's built um a laser weed uh laser weeding tractor and they have like gathering all this photographic data and they're like boom for sure it's all it's all ml it's all it's all ai in that sense and
21:00 Brad What is a laser weeding tractor?
21:03 Scott It's basically like, I think somebody even said this on your post about pest control.
21:07 Scott Basically, it's all for farmers to basically get higher yields out of their fields because they don't have to put pesticides,
21:14 Scott they don't have to put herbicides. They can basically like use this tractor and it's like it's automated. It's like drives up and down the field and like zap weed.
21:25 Scott Yeah.
21:26 Scott Yeah.
21:26 Brad That's cool.
21:28 Scott I know the founder,
21:29 Scott Paul Mike sell of that company.
21:30 Scott He's Paul's Paul's a super smart guy.
21:32 Scott Worked with me back in the day at Real Networks. One of those 10X engineers for sure.
21:36 Brad Super smart people,
21:37 Brad man. They got to be super smart these days.
21:40 Scott Yeah.
21:41 Brad We need AI for dummies. That's what I'm going to make. I'm going to make AI for dummies.
21:46 Brad And it's where you can just ask stupid questions and it gives you answers at an appropriate level.
21:54 Scott I don't want to say we already have that in Siri because Siri will just like light up again and think I'm talking to it.
22:00 Scott How do we boil this down into like actionable things for CTOs and for executive engineering people or how do we be better partners with our other stakeholders?
22:14 Brad I'll tell you about a frustration I have actually in a hot take I gave to the CEO I'm working with this week.
22:21 Brad And I hope I'm not going to get myself in trouble for this,
22:24 Brad but what's the worst case scenario? I get fired.
22:28 Brad what's the worst case sir i get fired and then i can't pay my mortgage and then i'm homeless and then i die and you know there's worth ways to go so um and there's a stop in there before homeless there's a stop at my mother-in-law's house where my mother-in-law and my wife one of them end up killing the other and going to jail and so you know it's
22:48 Scott Wow, that's a dark turn.
22:50 Scott So
22:50 Brad a dark it's a dark turn
22:51 Scott let's hope that doesn't happen.
22:53 Brad It's easy.
22:54 Brad I see this happening in our company.
22:56 Brad It's easy to find opportunities to make people more efficient.
23:00 Brad And when I said to our CEO this week, I was like, can we just stop doing everything,
23:04 Brad all the make more efficient stuff?
23:06 Brad Just forget about it.
23:07 Brad I mean, people want to do it, do it.
23:09 Brad But let's stop putting emphasis on that.
23:10 Brad Could we just figure out how to use AI to actually achieve our mission,
23:15 Brad which is.
23:17 Brad all our mission is to change lives one task at a time has actually been our mission but like how can we actually like provide our service better with ai like that is what is going to grow our revenue what's going to grow this business and ultimately which is going to serve that mission like because people naturally go to i can do this thing to write this document faster we can do this thing to automate this thing in accounting because it's just very tangible right like but it's harder to get the momentum going on
23:43 Brad on the product slash business side so i guess if there's a lesson i want to say to ctos out there is you know try to channel that ai energy towards actually making whatever you're working on better and that can be difficult and i'll tell you something we're doing we're about to have an ai focused hackathon coming up pretty soon actually we're not calling it a hackathon we're calling it a
24:08 Brad let's just call it a hackathon we have a different name for it because we want people to be a little bit more expansive than that but like create some space for people to just go play and figure out what could we do with this like i have an idea that i want to do with this hackathon of like what we could do
24:25 Brad Um,
24:26 Brad but in my day-to-day, what we're focused on, what our business priorities right now is not taking a week to screw around to do Scott's silly idea that maybe it'll work,
24:35 Brad but we're going to create some space for people to do that.
24:37 Brad So, um,
24:38 Brad I think that's something that is CTOs that we can do to create the room for people on our teams to find ways to use AI to have.
24:50 Brad meaningful impact on the world as opposed to just meaningful impact on the bottom line of the company I don't totally mean to sound like a Marxist here but like and I'm not by the way but like you know I just feel like so much of what we do
25:10 Brad is something about this technology revolution compared to what have I been through I've been through the first dot-com boom so the e-commerce boom well actually let's go back the first dot-com boom had a few parts to it it had online media and it had e-commerce and then
25:25 Scott Mm
25:25 Brad I went
25:25 Scott -hmm.
25:25 Brad through the social media boom and then I went to the gig economy boom and I went through the cryptocurrency blockchain boom and now I'm like in the AI boom and I feel like for some reason this one
25:36 Brad more than any others naturally not through any malicious intent just naturally is lending itself to benefiting the capitalists as opposed to the proletariat oh we have the bourgeoisie as opposed to the proletariat right and so i guess as as leaders we have to use whatever
25:55 Brad Well, I guess if you're a Marxist like I am now, apparently,
25:57 Brad we have to use whatever influence we have to say how can we actually use this technology for better for the consumer as opposed to just better for the bourgeoisie.
26:10 Brad And that's my take.
26:13 Scott Well, and...
26:14 Scott Yeah,
26:15 Scott so this is definitely getting into the politics and philosophy, and it'll be interesting as I ponder how I want to respond to that.
26:23 Scott There's a version of there's I sort of feel like I kind of want to play devil's advocate with this.
26:29 Scott You
26:29 Brad Do it
26:29 Scott know, there's a fair amount of what you just said that I 100 percent agree with.
26:32 Scott And then the but the argument that I've heard to counter that is,
26:37 Scott but what about the solo entrepreneurs? What about the two guys?
26:41 Scott guys in the garage they now they the one person that said something like hey i work with a lot of like small cash strapped non-profits and they spend quite a bit of time you know trying to figure out how to be more um optimal in their usage so so to me that story this idea that like well there's not anything that's it's not the case that the only people that benefit from efficiency gains and productivity gains
27:09 Scott are big corporations with lots of people where you end up doing risks because of that productivity gain they're not the only places that productivity gains result in good things i mean that's that's the that's your point about the capitalist you know
27:22 Scott basically get more done, get the same amount done with less people.
27:26 Brad I
27:26 Scott But if you're two guys in a garage or two entrepreneurs or one entrepreneur or a small resource strapped nonprofit,
27:33 Scott then maybe that's the way we should look at this productivity gain and say there's actually some there is something good in that.
27:39 Brad agree with that entirely.
27:41 Brad And I saw that comment there before,
27:44 Brad and I'm glad we went through those again because it makes me think,
27:46 Brad okay.
27:47 Brad To me,
27:48 Brad that's interesting.
27:49 Brad Somebody saying,
27:49 Brad I'm going to make this up. I don't know. I'm making some, I'm making a flux capacitor such that, you know, community health organizations can manage their case load more efficiently and then they can take on more patients and so on. I don't know,
28:03 Brad whatever would work.
28:04 Brad Like, well, there you're saying, hey, I'm going to use AI to actually do some good for the world.
28:07 Brad And so you're right.
28:08 Brad It just seems like maybe it's just the space I'm in is.
28:14 Brad I just see people tend to lean toward just making employees more efficient.
28:20 Brad But I guess what you're saying is depending on the context in which you're in, whether you're in a big company or a small company,
28:28 Brad actually making employees more efficient has different benefits.
28:32 Scott Yeah.
28:33 Brad Yeah, that's an interesting way to think about it. I guess, okay, maybe I'm just a jerk actually.
28:37 Scott No, I mean, I mean, I mean, you know,
28:40 Scott there's this concept that you hear in like in venture capitalists, they often get around and they talk about like,
28:45 Scott oh, or, you know, sell side analysts or whatever they get around,
28:48 Scott they get it together and they're like, oh,
28:49 Scott that company needs to get more fit,
28:51 Scott get fit, get fit.
28:53 Scott Right.
28:53 Scott And that's like that's like large corporations speak for increased productivity and cut staffing,
28:59 Scott right,
28:59 Scott like lower, lower costs as as part of.
29:03 Scott productivity gains and that basically means lay off people or if people right and you know i i'm sure you've been there before like i know i've been there where um
29:16 Scott when it was you know roaring roaring hey hey hey um we were you know higher like crazy because you're just trying to move as fast as you can and you've got to and you know everybody's super busy so you're you don't have time to think about efficiency you're just busy cranking stuff out as fast as possible and it's it's a true statement that like companies in that in that particular part of the cycle they will hire people that they don't
29:42 Scott necessarily have something for all of them to do uh long term right and
29:48 Brad I've never
29:48 Scott i
29:48 Brad had that
29:49 Scott mean
29:49 Brad problem.
29:49 Scott i i mean i i've definitely had the problem of growing too fast and growing beyond our means and then later kind of coming back and going ah we probably didn't need to hire all those people
30:03 Brad I was responding to not having something to do. Yes, absolutely.
30:07 Brad That's a pattern I see.
30:09 Brad very consistently at companies I've been at, companies just get out over their skis.
30:14 Scott And so in those environments,
30:16 Scott in a world where, I mean, this is an interesting like,
30:19 Scott you know, if AI can help me.
30:22 Scott take my current employees and make them more productive,
30:25 Scott then maybe I won't accidentally hire people that aren't sustainable for me.
30:30 Scott So that's another version of the productivity has a different lens.
30:35 Scott And if you're the solo entrepreneur who's trying to bootstrap, then that's definitely where you're at,
30:39 Scott right?
30:39 Scott You're just trying to get as much done as you can with the least amount of resources possible.
30:44 Brad Okay. And now the solopreneur, the small business person,
30:47 Brad that's interesting.
30:48 Brad I got a question.
30:48 Brad Question for you.
30:49 Brad In our world, in the technology world,
30:51 Brad like there's kind of like frameworks you can get, you know, it's like it's a whole package.
30:54 Brad I'm wondering if we could make like AI,
30:57 Brad this is kind of like a business idea,
30:59 Brad AI solo entrepreneur in a box.
31:03 Brad Like you just you got everything here. You've got your people operations,
31:08 Brad you've got your finance,
31:10 Brad you've got your
31:12 Brad what marketing website I don't know what else even you need you need you got your legal AI stuff it's all right there maybe it's not maybe it's just a layer on top of some best of breed stuff but like really it's almost like kind of like a workday or a Zoho kind of thing
31:30 Scott Yeah, Zoho is an example of that,
31:32 Scott and there's a handful out there that like, you know, kind of a.
31:36 Scott I mean, I don't know all the names of them,
31:38 Scott but, you know, definitely there's the when the gig economy thing was was taking off like in every and all the states started cracking down on like self-employment, then there was a boom in there was a boom in companies that basically were like,
31:50 Scott oh,
31:50 Scott we'll help you form your LLC. And also while we're at it, here's the solo entrepreneur in a box,
31:56 Brad Yeah.
31:56 Scott you know, so here's your built in here's your built in, you know, boilerplate legal contracts. Here's your built in.
32:02 Scott will sell you business insurance and so yeah
32:04 Brad oh all
32:05 Scott i mean i well
32:06 Brad the good ideas are gone but
32:08 Scott i mean probably
32:09 Brad what did that guy in the patent office say like in 1921 i don't know everything that could be we don't need a patent it was like the head of the patent office it was like we don't need a patent office anymore everything that's been invented has been we've invented it all yeah
32:23 Scott anything that can be invented has been invented yeah
32:25 Brad who could imagine anything else
32:28 Scott Yeah, exactly.
32:29 Scott All right. So so so there is this tension around like productivity gains and the implications of that for, you know, everyday workers.
32:41 Scott And then there's the tension of like,
32:43 Scott yeah, but, you know, hey, this gives people an opportunity to be solo entrepreneurs,
32:47 Scott et cetera.
32:50 Scott What about your other posts that you had about LLMs don't eliminate the need for experts?
32:56 Scott Tell me more about that.
32:58 Brad Okay, that wasn't really my post.
33:00 Brad It was just reposting a guy named Chris Sod.
33:03 Brad I think that's how he pronounces his last name.
33:05 Brad I'm acquainted with Chris. We met a long time ago in the mid aughts.
33:10 Brad I think he was working for a company that we,
33:13 Brad that my company was.
33:15 Brad um buying their software and then we got to hang out a bunch of south by southwest and then i think he went on to be like early at uber and
33:22 Scott Yep.
33:22 Brad if i recall and he's a pro he was an engineer and a product guy and he's he's really bright on the startup stuff and i love his attitude and he had this post about um
33:34 Brad how he was talking about how how you know people are saying oh you know chat GPT is killing the need to hire domain experts and somewhere in there he said you know LLMs don't
33:47 Brad eliminate the need for experts they amplify the need for experts and it actually goes right back with that thing you quoted earlier which is another post which was vibe then verify and it's really the verify part which is and this is what we did the same thing we've talked about with software development and having the agent go write you some software it's the the job a lot of the job is turning to the verify part where you actually need the expertise
34:12 Brad And then we've talked about this, too,
34:14 Brad and I don't know that anybody has the answer to it, which is, well.
34:18 Brad How are people going to get the expertise?
34:20 Scott Yeah.
34:22 Brad And which actually leads to another post. I didn't write this
34:26 Scott Wait,
34:26 Brad post.
34:26 Scott wait, wait, before we go off the vibe, but verify and the need of the expert, I'll tell you a funny thing. Literally, this just happened today.
34:32 Scott One of my very,
34:34 Scott very senior developer DevOps senior guys,
34:36 Scott he was like, he basically was trying to say that we we've we were all Docker Docker compose in most.
34:44 Scott in most of our our stack especially our local development stack and then we started using kubernetes now in the cloud and one of the ideas that was floated by one of our new devops guy was that there's this there's this miniature version of kubernetes instead of k8 about called k3s which is like a mini kubernetes and and my my senior devops guy was like
35:09 Scott Oh,
35:09 Scott I'm going to try to I told Claude to convert all our Docker stuff into K3s and it went crazy.
35:15 Scott It started going all over.
35:17 Scott It was just running all over the place.
35:19 Scott And I don't know exactly what it did.
35:21 Scott I still need to go in there and unbury it from all the rubble it created
35:25 Brad Yeah.
35:25 Scott and figure out if it made any progress or not.
35:29 Brad Okay.
35:29 Brad So my friend, Sean Nolan,
35:31 Brad my friend,
35:31 Brad I consider him a mentor really.
35:33 Brad He was, I worked for him.
35:34 Brad I mean, this will date us. I worked for him in the 90s.
35:39 Brad He was the CTO I worked for and he was a great CTO to work for. He's super bright.
35:45 Brad He's a he's a bright guy and he's just a high energy guy and positive and he's curious and he's retired now from actually having a day job,
35:54 Brad but he's always he's a tinkerer and he's never going to stop being tinkering. And so the other day he was he posted about how he was curious about some question.
36:03 Brad Oh, I'll just give a little bit. Apparently,
36:08 Brad the United States government subsidizes regional airports and regional airlines so that everybody can have access to transportation.
36:17 Brad And so he had some curiosity about that and he went and dug up some data about some different things and then he did some analysis and he fed this all to ChatGVT.
36:26 Brad But when he asked ChatGVT or Cloud Code, I don't know what he was using, he fed it to the model.
36:30 Brad I'll just call it the model,
36:31 Brad the board.
36:32 Brad He fed it to the board, but he said, I need you to write me.
36:35 Brad A program,
36:37 Brad I want it to be a node with JavaScript that takes these three data sets and does these things to them and outputs a CSV because he knew once he got what he wanted into a CSV, then he could analyze it in the way he wanted.
36:50 Brad And I asked him, I asked online,
36:52 Brad and then he ended up writing,
36:53 Brad this was a blog post that he posted on LinkedIn and I left a comment there and then he wrote another blog post in response. And what I said was.
37:00 Brad Sean,
37:01 Brad why did you ask it to write you code?
37:06 Brad Like you're an engineer and it allowed you to double check that it was doing it right and that was doing it your way, but you could have asked it to do the analysis all the way through to get you whatever answer you wanted to know about regional airport subsidies because you're curious and couldn't help yourself.
37:24 Brad But you wouldn't have been able to verify if it was right then in that case.
37:30 Brad And most people don't have the skills to actually have it write you code that you can then run locally that you can then get your CSV that you can see if the results are right and so on and so forth.
37:42 Brad And I just pose that as in the vibe but verified dilemma. That is.
37:47 Brad a problem,
37:48 Brad which is if we ask, he could simply ask the question
37:52 Scott Yeah.
37:52 Brad to the model and the model might have come up with the right answer or might have not come up with the right answer.
37:57 Brad We know the model would have answered confidently.
38:01 Scott Yeah.
38:01 Brad And so
38:03 Scott Well, it's really still bad. It's still bad at math.
38:06 Scott It's really still bad at simple math. Like I,
38:08 Scott you know, I use it all the time to do things like nutrition tracking and stuff like that. And it'll tell me like,
38:15 Scott oh,
38:15 Scott that meal that your breakfast had,
38:17 Scott you know, seven grams of fiber in it and your lunch had 15 grams of fiber. So you've got 11 grams of fiber so far for the day. And it's like seven and 15 does not equal 11.
38:29 Scott What are you talking about?
38:30 Scott how
38:30 Brad How can
38:31 Scott yeah
38:31 Brad you rely on it for anything then?
38:33 Scott so that's why i think to your point though that's the reason like you've got to verify what it's doing like if you just take it at face value it's like for sure gonna just yeah it's a mess still that part is still a mess there
38:46 Brad It's a mess.
38:49 Scott you go
38:49 Brad Here's what I'm taking from this conversation before we wrap.
38:52 Brad I'm going to go try to find.
38:55 Brad I'm going to have to find I'm going to go talk to the not-for-profits that I support in our local area and I'm not just going to go do some product management, some customer development and try to just see like are there ways that we could build some small products.
39:12 Scott Yeah.
39:12 Brad Maybe some small SaaS services or something that would be inexpensive or free to not-for-profits that would help them be more efficient or effective.
39:22 Brad How could our food bank,
39:23 Brad that's the one, you know, we support the local food bank.
39:26 Brad How could the food bank benefit from this stuff?
39:30 Brad I
39:30 Scott let
39:30 Brad don't know.
39:30 Scott them eat tokens
39:32 Brad Let them eat tokens.
39:33 Brad That needs to be the title of this episode.
39:36 Brad I know you're the guy that comes up with the titles.
39:38 Brad So let them eat tokens.
39:39 Brad Can we finish on that?
39:41 Scott finish on that wow that was quite a run yes
39:47 Brad Brad,
39:47 Brad it's been fun this week as always.
39:50 Scott good talking
39:51 Brad I look forward to talking to you next week.
39:52 Scott what do we say
39:53 Brad Smash rate and subscribe.
39:55 Brad Review,
39:56 Brad rates, subscribe,
39:56 Brad all of it.
39:57 Scott Do it all.
39:58 Brad Thanks, everyone.
40:00 Brad This episode of How Many CTOs Does It Take Podcast was edited by Saffron Epigal. Our theme song was AI Generated by Brad.
40:08 Brad Please follow us on all the social media platforms at at HowManyCTOsPod.
40:13 Brad You can find us on YouTube,
40:15 Brad LinkedIn,
40:16 Brad Instagram,
40:16 Brad Facebook,
40:17 Brad Threads, X,
40:18 Brad TikTok,
40:18 Brad and Patreon,
40:19 Brad or at our website at HowManyCTOsPod.com.
40:24 Scott We
40:24 Brad And remember to rate,
40:26 Brad review,
40:26 Brad and subscribe on the podcast platform of your choice.