back to indexIs Overstimulation Ruining Your Life? - How Your Phone Affects Intelligence, Focus & Productivity

Chapters
0:0 Are We Getting Dumber?
22:19 Is it possible to build a better Twitter?
27:23 Should I ditch my higher paying job to avoid stress?
32:19 Can I remain relevant with younger, tech savvy colleagues?
35:57 How should I navigate my time during while receiving severance?
38:41 What should I do next after being fired from my federal job?
42:27 Practicing focus on the weekends
46:6 An engineering Kansan system
57:26 The AI Blind Spot
00:00:00.000 |
So several people recently sent me the same article. 00:00:04.300 |
It was written by John Byrne Murdoch, and it had a provocative headline, 00:00:11.020 |
So I'm going to take a closer look at this claim. 00:00:15.180 |
First, I want to develop a better understanding of why the data seems to show that we are getting dumber. 00:00:23.840 |
Use that understanding to help find practical ways that you as an individual can push back on that trend. 00:00:30.340 |
And not only not get dumber, but make sure that you continue to get smarter. 00:00:34.220 |
So this article that I was just citing was inspired by some recent analysis that was released by the Organization for Economic Cooperation and Development. 00:00:42.820 |
They do this regular test called the PISA, which benchmarks teenagers around the world and their knowledge of math, reading, and science. 00:00:53.600 |
It's a useful test to kind of understand what's going on. 00:00:55.880 |
So they looked at—there's a recent analysis of this. 00:00:59.200 |
The article looked at that recent analysis, plus some other tests that have been given worldwide recently. 00:01:03.460 |
And the author of the article made the following conclusion, and I'm quoting here, 00:01:08.200 |
Across a range of tests, the average person's ability to reason and solve novel problems appears to have peaked in the early 2010s and have been declining ever since. 00:01:20.220 |
So I have a graph to show here, Jesse, bring up this graph for those who are watching instead of just listening. 00:01:26.600 |
So here's one of the key graphs that indicates this point. 00:01:29.560 |
It shows performance and reasoning and problem solving tests over time. 00:01:33.820 |
On the left-hand graph, what you see is a line for science, reading, and math. 00:01:37.780 |
And you can attest, Jesse, that right around 2012, all those lines go downward. 00:01:47.620 |
Literacy takes a big spill right around 2012 and goes down dramatically ever since. 00:01:55.920 |
So that indicates the thesis of the article that, hey, starting in the early 2010s, at least according to the test, we're doing worse. 00:02:10.140 |
Well, the article points to an obvious culprit based just on the forensic evidence of the timing. 00:02:14.160 |
These trends seem to occur right around that 2012 to 2014 period is where we see these downward shifts. 00:02:23.940 |
There have been other things that worry us that got more prominently worrisome starting around that time. 00:02:29.020 |
For example, teenage mental health deterioration is another one. 00:02:37.300 |
This is when we got worldwide ubiquity of smartphones became a reality. 00:02:41.840 |
So the article points out correctly, we seem to be seeing a negative turn on these tests of various reasoning and intelligence abilities around the time smartphones come. 00:02:53.960 |
But I don't think it's useful to just leave it there. 00:02:56.640 |
So if we just say, yes, smartphones seem to have led to us getting dumber. 00:03:04.740 |
We're probably not going to get rid of phones. 00:03:07.500 |
Most people need various aspects of the phone and app ecosystem to operate. 00:03:12.180 |
So it sort of leaves us without much to do except for to shrug our shoulders and say, well, I guess phones may as dumber, but what are we going to do? 00:03:17.940 |
It's sort of like cars came along and traffic deaths, you know, got higher. 00:03:21.880 |
Here's another here's a source of 20,000 new deaths a year that didn't exist before cars. 00:03:27.260 |
And it was like, this is just something we're going to have to live with. 00:03:29.040 |
It feels that way sometimes when we're dealing with these cognitive impacts of smartphones. 00:03:33.980 |
But I think we can do better and I want to do better today. 00:03:38.400 |
So I'm going to look closer and I'm going to try to develop a hypothesis that explains at least partially, 00:03:43.860 |
specifically why, what mechanisms of smartphones are making us perform worse on these tests, making us dumber. 00:03:48.960 |
Because if we know more specifically what about these things is making us dumber, 00:03:53.800 |
then maybe we have a chance of reversing that even without having to get rid of our phones. 00:03:58.740 |
So to look closer, what I'm going to do is pull up another graph. 00:04:03.280 |
Here's another graph from this article that gets at what specifically is changing in the smartphone era. 00:04:10.540 |
So we see here on the left, a graph over time, measuring percentage of respondents that are saying they have difficulty thinking or concentrating. 00:04:19.380 |
This comes from another survey called Monitoring the Future, which John Byrne Murdoch sort of pulled up. 00:04:23.640 |
What you notice here is it's relatively stable, difficulty thinking or concentrating until that same inflection point of around 2012. 00:04:32.200 |
And then it shoots up and we see a aggressive upward trend. 00:04:37.720 |
On the right, we have another graph, percentage of people saying they have trouble learning new things. 00:04:46.840 |
Same time that difficulty thinking or concentrating shoots up. 00:04:50.080 |
So, of course, right at the smartphone inflection point, we see mechanistically that people suddenly reported at much higher rates, 00:04:57.780 |
having difficulty thinking or concentrating and having trouble learning new things. 00:05:04.300 |
That is where I think we're seeing the effect of smartphones. 00:05:08.640 |
And if we look at a little bit closer, why is smartphones now causing us to have difficulty thinking or concentrating or trouble learning new things? 00:05:16.940 |
I believe we can identify what I think of as a cognitive death spiral here. 00:05:25.440 |
The phone itself is not the problem, of course. 00:05:27.720 |
It's the ecosystem of attention economy that arose around the smartphone. 00:05:31.700 |
Pre-smartphone, if you were building a sort of information platform, your Facebook pre-smartphone, 00:05:37.100 |
you were building a product that was trying to be maximally useful to users. 00:05:41.820 |
I want to make Facebook so useful that people will think to log in and want to be a member of it. 00:05:51.960 |
You can find out what your friends are up to. 00:05:55.880 |
Post-smartphone, we had a shift towards an attention paradigm where the idea now is not being useful, 00:06:03.220 |
They realized users were, you wanted a large user count if you were trying to raise money, 00:06:07.700 |
but once you had a company running, you wanted to monetize those users. 00:06:13.760 |
And this is when the goal of platforms changed to not being as useful as possible, 00:06:18.640 |
So, we get ubiquitous smartphone use pickup around this time. 00:06:21.860 |
Why does that cause a cognitive death spiral? 00:06:25.180 |
You have this rhythm in your life of constantly being distracted because 00:06:29.060 |
the apps on your phone are designed to grab your attention. 00:06:33.260 |
It has faster, more desirable stimuli than other things in your life. 00:06:36.600 |
So, now you're rewiring these circuits in your brain 00:06:40.460 |
so that the reward circuits are very much tuned towards if a phone is nearby, 00:06:46.960 |
Let's have our dopamine cascade focus on the action of looking at that phone 00:06:52.100 |
because that we have learned, these circuits have ingrained, 00:06:54.320 |
that's going to give us the quicker hit to whatever else we are doing. 00:06:57.340 |
And because the phone is ubiquitous, we constantly have those reward circuits firing 00:07:05.520 |
Look, if you put a donut in front of me, I'm going to build up a reward. 00:07:09.360 |
You know, every day at four, you put out donuts at the office, 00:07:11.580 |
I will build up a reward circuit where, like, I'm really looking forward to that donut. 00:07:14.760 |
If you now follow me everywhere I go with a cart of donuts, there's going to be a problem. 00:07:19.480 |
So, that's what started happening with the phone. 00:07:20.660 |
So, now our mind gets rewired to craving this more faster pace of stimuli. 00:07:26.340 |
That can directly impact our ability to concentrate because that's distracting us. 00:07:35.420 |
It's harder to sort of apply our existing intelligence. 00:07:38.400 |
But the reason why I think it creates a cognitive death spiral 00:07:41.120 |
is that it also means we spend less time on the type of activities that could make us smarter. 00:07:48.700 |
So, if two things going on at the same time is our mind gets rewired for faster stimuli, 00:07:53.460 |
we have a harder time applying our existing intelligence, 00:07:56.340 |
but we also have a harder time engaging in activities that would make us smarter. 00:08:06.520 |
What we have on the screen here is a chart showing the decline of reading. 00:08:13.340 |
So, this is a percentage of U.S. teenagers who read in their leisure time. 00:08:18.540 |
One of these plots on here shows who says they hardly ever read, 00:08:23.160 |
and the other plot shows who reads almost every day. 00:08:29.520 |
It's like moving mildly down through the 80s and 90s. 00:08:34.260 |
Right around 2012, that goes down real sharply, 00:08:37.540 |
and the people reporting that they hardly ever read goes up real sharply. 00:08:46.260 |
Reading is an example of an activity that makes you smarter. 00:08:48.740 |
The brain circuits involved in reading makes you smarter. 00:08:55.520 |
You can better sustain your attention on abstract targets. 00:08:57.720 |
You can better manipulate information and build and construct worlds in your mind. 00:09:03.460 |
It is just straight-up exercise for your mind. 00:09:05.200 |
It's why it's been at the core of sort of every academic curriculum 00:09:09.520 |
So, it's one among other activities that we do less of 00:09:14.640 |
and when we rewire our mind for faster stimuli, 00:09:17.120 |
we're less likely to actually, as we see in that graph, 00:09:20.080 |
we're less likely to actually spend time doing that. 00:09:23.160 |
We have a hard time applying whatever intelligence we have, 00:09:25.860 |
and we slow down or completely stop the increase of our intelligence 00:09:31.280 |
as we do activities that would naturally get us there. 00:09:40.120 |
and we're having a harder time applying the intelligence we have. 00:09:43.140 |
Okay, so really now what we're talking about, 00:09:52.500 |
of having our brain rewired for faster stimuli, 00:09:57.920 |
spending less time with activities that foster intelligence. 00:10:02.840 |
I wanted to interrupt briefly to say that if you're enjoying this video, 00:10:11.060 |
The Lost Art of Accomplishment Without Burnout. 00:10:14.440 |
This is like the Bible for most of the ideas we talk about here in these videos. 00:10:20.020 |
You can get a free excerpt at calnewport.com slash slow. 00:10:35.920 |
that don't involve us having to go back in time. 00:10:38.240 |
Now, before I talk about what that could particularly be, 00:10:40.520 |
here's the analogy that came to my mind from 60 years ago, right? 00:10:48.220 |
the economy shifted from being primarily industrial agricultural 00:10:51.900 |
to having this very strong sort of office-centric knowledge work sector. 00:11:04.560 |
because before you were probably working on a farm 00:11:12.280 |
And now suddenly you're sedentary because you're in an office. 00:11:27.180 |
You just have a heart attack and die in your 60s. 00:11:32.100 |
Well, we didn't say we need to shut down the offices 00:11:37.200 |
We said, what was the thing we're missing now from the farms 00:11:47.060 |
In 1920, you didn't have to think about exercising. 00:12:14.580 |
and maintaining our ability to hold attention 00:12:36.500 |
about how one might have a cognitive exercise routine 00:12:43.960 |
Reading is pull-ups and push-ups for your brain. 00:13:18.140 |
in the constant companion model of your phone, 00:13:24.400 |
Go there if you need to check your text messages. 00:13:29.720 |
Again, you want to sort of break out of this pattern 00:13:40.040 |
but I just learned this term stimuli stacking. 00:13:50.020 |
multiple streams of stimulus at the same time. 00:13:56.100 |
and maybe you even have like a different device 00:13:57.980 |
on which you're like following something else. 00:14:00.540 |
And supposedly some of the streamers like Netflix 00:14:13.240 |
because you can't actually look at your phone 00:14:17.420 |
We want your mind to be used to like doing one thing 00:14:30.880 |
It's going to be squirrel, squirrel, squirrel. 00:14:36.640 |
Get used to just being alone with your own thoughts 00:14:48.040 |
Playing the guitar requires a lot of concentration 00:14:56.040 |
A particular sport requires a lot of work and focus 00:15:03.100 |
and give you obvious rewards as you get better. 00:15:09.300 |
So anyways, I thought that was a cool article. 00:15:12.360 |
It's not just the phone itself makes us dumber. 00:15:17.720 |
It's particularly the way that it's rewired our brain, 00:15:22.780 |
of we have a harder time applying our intelligence 00:15:26.040 |
Look, man, when you're in the office building 00:15:46.360 |
Because we were just doing this all the time. 00:15:49.460 |
and we didn't have like constant distractions 00:16:00.480 |
but we couldn't because it was buried in snow. 00:16:12.800 |
and you would trudge through this to a library 00:16:16.600 |
and have to like read your books for a while. 00:16:22.900 |
So you got to like force yourself to read books. 00:43:03.600 |
And part of this rotation means that every six 00:43:16.000 |
pseudo productivity is almost the name of the 00:43:19.860 |
There's a lot of context switching throughout the 00:43:22.660 |
So my question is really, given that this is a 00:43:27.340 |
temporary role and given that at the moment I feel 00:43:31.940 |
that my ability to concentrate and work deep has 00:43:34.440 |
diminished, what are some practices I can do maybe over 00:43:40.500 |
weekends or in my free time to improve that ability to 00:43:44.620 |
concentrate and once I roll off this rotation and go into my 00:43:50.120 |
new role, what are some things I can do early on to get back into 00:43:54.220 |
the deep work routines and habits to ensure that that skill maintains 00:44:05.960 |
I think, by the way, Jesse, because Alfie mentioned the word 00:44:09.960 |
pseudo productivity, which comes from my book, Slow 00:44:13.680 |
Productivity, that we could play the theme music. 00:44:17.900 |
Someone brought up slow productivity organically. 00:44:22.040 |
So we're going to play the slow productivity theme music. 00:44:25.860 |
I mean, first of all, this emphasizes there's some roles that were deep 00:44:31.040 |
Like deep work is just a type of effort, right? 00:44:34.320 |
It is a good way of maximizing cognitive abilities for a lot of 00:44:39.680 |
Alfie's in a role right now where it's not important. 00:44:41.780 |
So he's not spending a lot of time in the state of deep work. 00:44:49.040 |
But I think it's a cool question of like, how do I make sure I don't 00:44:52.700 |
lose that ability, that ability to concentrate that, you know, will be 00:45:01.180 |
We talked about this in the beginning of this episode, that if you're 00:45:05.280 |
constantly in like a high stimuli type of situation, you get less comfortable 00:45:10.380 |
We see that in the data that you spend less time doing activities that make you 00:45:13.780 |
And then you get even less comfortable, less smart and less comfortable 00:45:18.680 |
So do the type of things we talked about earlier in this episode. 00:45:21.880 |
Force yourself to read books outside of work. 00:45:23.800 |
Practice the non-constant companion model of the phone is plugged in, not with me when 00:45:30.560 |
So you're just used to doing one thing at a time. 00:45:33.760 |
You're going to work on professional and personal projects just in my head while I'm 00:45:37.320 |
walking with no source of distraction with me. 00:45:40.400 |
These are all things that are going to help your mind be comfortable with sustained 00:45:44.700 |
concentration and actually strengthen your mind's ability to do things. 00:45:49.040 |
You might add into there the hard hobby, you know, learn a new skill, learn how to computer 00:45:53.600 |
program, learn how to do like microelectronics or woodworking. 00:45:56.360 |
Be really careful about cognitive calisthenics during this period in which you were basically 00:46:00.880 |
doing the cognitive equivalent of smoking during the workday. 00:46:03.320 |
You've got to sort of offset the damage that's happening with active improving activities. 00:46:07.520 |
So yeah, you'll be out of this rotation soon. 00:46:12.900 |
And then when it comes time to schedule deep work again in your next rotation, you're not 00:46:16.540 |
going to struggle with maintaining your concentration. 00:46:18.840 |
You're going to feel like your instrument is still well-practiced. 00:46:31.960 |
Ian says, great discussion on Kanban boards and systems. 00:46:36.040 |
I love the space and has been a great reminder to me of how simplicity here is what works. 00:46:43.060 |
Attached are images of my engineering Kanban board, which I created around 2014 or so, that 00:46:51.140 |
Still my most powerful organizing system yet. 00:46:54.600 |
I'm still yet able to, I've still yet been able to replicate this to do or do better in very 00:47:03.020 |
This is around the third version revision of the system and nothing I've done yet has replicated 00:47:07.780 |
the visual organizational power, simplicity, and political leverage with internal customers. 00:47:12.540 |
So I'm going to read about his system that he uses. 00:47:15.960 |
And then for those who are watching, he sent some pictures of it that I'll bring on the screen 00:47:23.720 |
So he's talking about like the, the optimal way his, his Kanban based system, like how 00:47:30.380 |
We did a standup meeting in front of the board, Monday, Wednesday, Fridays. 00:47:33.540 |
It was the center of the engineering office nestled between two filing cupboards and hard 00:47:38.080 |
We had custom cards with blue tack on the side of the cupboard and whiteboard. 00:47:42.080 |
We had constant visual reminder of what's going on. 00:47:48.980 |
And we had queues, parking lots, done piles, et cetera. 00:47:51.960 |
It worked really well for parking and queuing items rather than solve the last job that came 00:47:56.660 |
If there was a real fire, we just made the call to drop everything and solve the fire. 00:48:00.960 |
I think we gave up on the percentage done bars on the cars as that just didn't add up value 00:48:06.900 |
Great for visual prompt to others of what we were working on and where things were at. 00:48:10.600 |
I've tried digital systems and they just don't work nearly as well or as powerful 00:48:14.920 |
All right, I'm going to bring up some of these images here for those who are watching. 00:48:21.980 |
It is on the wall between two filing cabinets. 00:48:24.220 |
For those who are listening, it's on a whiteboard. 00:48:28.420 |
So there's blue tacked paper cards on a whiteboard that's up on the wall. 00:48:36.180 |
One of these columns is to do next unassigned. 00:48:40.420 |
So we talked about this in slow productivity. 00:48:42.740 |
You have a place to keep track of work that needs to be done in your group that doesn't 00:48:48.800 |
It's not being forgotten, but it's not just in someone's inbox or in someone's head. 00:48:56.600 |
Now, look, this column is a cool way of doing it. 00:49:06.980 |
The person in the second row is working on three things. 00:49:10.140 |
The person in the fourth row has four things. 00:49:11.640 |
You can see specifically the four things they're working on. 00:49:14.040 |
You'll notice, Jesse, two of these things are sort of on the border between two rows. 00:49:20.400 |
So I think that means those two people are working on that together, maybe? 00:49:24.400 |
We can't actually read the cards, but they have the information about the task. 00:49:28.660 |
It's got to be because there's another one at the end, too. 00:49:35.680 |
So it's like what people are working on at the moment. 00:49:43.800 |
So if we look down here, we see there's rows for lots of people. 00:49:50.680 |
I talk about this in Slow Productivity, doing this for yourself, having a queue of like what 00:49:57.380 |
So I can look at Adam's row here and see one, two, three, four, five, six, seven things, 00:50:01.660 |
one with a red dot, like things he's waiting to work on. 00:50:04.400 |
And then next to it would be like, here's the four things he's working on now, knowing 00:50:07.620 |
that as he finishes one of these things, he'll pull something else over there. 00:50:10.340 |
So none of this is being kept track of in their head. 00:50:12.640 |
And none of this is being kept track of just their email inboxes. 00:50:19.520 |
Here's work we've assigned, but it's not being worked on. 00:50:23.400 |
And we can just see it visually who's working on what and its status. 00:50:26.300 |
Notice over here on the left, this is called parking lots. 00:50:30.080 |
They have on the side of the cupboard parking lot. 00:50:33.740 |
This is a Kanban idea where it's things you don't know. 00:50:42.480 |
Like these are things we're thinking about, but we're not looking to get these done right 00:50:47.800 |
So anyways, what I love about this, and I'm really big on this idea in my book, Slow 00:50:51.080 |
Productivity, is they're keeping track of what needs to be done, its status, and who's working 00:50:57.040 |
on what in a centralized, transparent way, as opposed to allowing workloads to exist non-transparently, 00:51:06.080 |
It makes a huge difference in the experience of your day. 00:51:08.860 |
Because now, if you're one of these people represented by a row on this board, you're 00:51:12.260 |
only working on the things in your row in the in-progress column. 00:51:17.720 |
And you don't have to waste cognitive cycles on all these other things. 00:51:20.380 |
But Jesse, look at how many things are in to-do next, unassigned, and parking lot. 00:51:28.300 |
Without a system like that, all of those things would just exist on people's plates and be 00:51:33.740 |
generating potential, hey, what's going on with those emails? 00:51:35.760 |
Or can we just have a quick meeting to check in on those emails? 00:51:38.420 |
Or just in the back of your head is something you're supposed to be working on that you're 00:51:42.240 |
So I see all of those cards that are in all of those other places as cognitive overhead 00:51:49.480 |
So they're going to finish stuff much quicker, and people are going to be much happier. 00:51:57.260 |
That's a great demonstration of how these type of task force systems can work for teams. 00:52:12.980 |
We'll talk about our friends at the Defender line of vehicles. 00:52:19.020 |
We're talking about the Defender 90, the Defender 110, and the Defender 130. 00:52:26.220 |
They're designed in a way that has like the modern features or conveniences you would want 00:52:31.860 |
But they also have that tough, rigid body design, durability, that lightweight monocoque architecture 00:52:40.040 |
for extra strength that you can take this thing on adventures. 00:52:43.300 |
Good-looking car that can be smooth and comfortable, and yet also take you where adventure might hold. 00:52:54.960 |
I am going to get you from our sponsor, Defender 110, that you can drive. 00:53:00.280 |
The bad news is it's wrapped with a picture of me. 00:53:03.520 |
So it's going to make sure of like an advertisement for the podcast. 00:53:07.880 |
I'm going to have like a big thumbs up on both sides of the car. 00:53:10.700 |
And there's a speaker on top that's going to chastise people for being on their phone as 00:53:17.320 |
It's a cool enough looking car that you probably still get away with it. 00:53:23.940 |
I haven't seen the one yet, again, that was parked outside our office after the last Defender 00:53:29.340 |
But I see them everywhere now that I know about them, and it's cool. 00:53:35.800 |
Anyways, if you want to see what these cars look like with or without the Cal Newport wrap, 00:53:39.960 |
I don't know if that's a standard feature yet. 00:53:48.560 |
So you can visit LandRoverUSA.com to learn more about the Defender. 00:53:53.320 |
I also want to talk about our friends at ExpressVPN. 00:53:58.340 |
How do you choose which internet service provider to use? 00:54:02.620 |
The sad thing is most of us don't have very many options, right? 00:54:06.540 |
ISPs are often operating like monopolies in your region. 00:54:08.980 |
This is the internet service provider that's nearby. 00:54:10.960 |
They use this monopoly to take advantage of their customers because you don't have any other 00:54:17.140 |
you don't have any other recourse if you don't like what they're doing. 00:54:19.080 |
And you get things like data caps and bandwidth throttling, et cetera. 00:54:22.380 |
But here's one thing that these ISPs are doing that you can push back on using ExpressVPN. 00:54:27.440 |
They're trying to keep track of every website you visit, right? 00:54:35.500 |
Do you know what, Jess, I'm going to put you on the spot here. 00:54:38.220 |
I never know like what network terminology is well-known or not. 00:54:41.580 |
If I say packet, what do you think of when I say like an internet packet? 00:54:51.960 |
So when you're communicating on a network, like you're communicating to the deeplife.com over 00:54:56.480 |
the internet, your communication gets broken up into these little messages called packets. 00:55:01.340 |
And the front of the packet's like a address on an envelope. 00:55:10.060 |
And then that packet gets sent through the internet. 00:55:11.840 |
It gets bounced from router to router until it gets to the website. 00:55:14.700 |
So every router has to look at the address along the way until it gets to where it's going. 00:55:20.120 |
And then the destination can open up that envelope and, oh, here's what you're sending me, a request 00:55:26.640 |
And a lot of websites these days use a secure protocol so that the stuff inside the envelope 00:55:33.780 |
So if I'm your internet service provider and you hand me this envelope, hey, get this to 00:55:39.980 |
I don't know what you're sending to the deeplife.com, but I can see that's who you're talking 00:55:46.860 |
So just like in the mail, you can have a thick envelope, so I really don't know what 00:55:50.260 |
you're mailing, but who you're mailing it to, my postman knows. 00:55:54.120 |
So ISPs, just look at the address on these envelopes and they know, hey, here's who you're 00:55:59.960 |
I don't know what you're saying, but I know who you're talking to and they sell that data 00:56:02.780 |
With a VPN like ExpressVPN, you get around that. 00:56:06.280 |
And the way you get around that is you take the envelope you really want to send. 00:56:09.340 |
I really want to talk to the deeplife.com, but I don't want people to know. 00:56:12.480 |
I'm going to put that inside another envelope and I'm going to send that envelope 00:56:17.840 |
So now all your ISP knows is, you know, Jesse's talking to a VPN server. 00:56:23.760 |
And then the VPN server can open that out and take out your real envelope. 00:56:26.900 |
Oh, you really want to talk to the deeplife.com. 00:56:28.740 |
It will talk to the site on your behalf and then it will put the response back in a big 00:56:34.160 |
And now all your ISP learns is that you're talking to a VPN server. 00:56:45.400 |
It makes it so among other advantages, you can communicate with other sites and services 00:56:51.560 |
without like your service provider knowing who you're talking to. 00:56:54.660 |
If you're going to use a VPN, use the one I recommend, which is Express VPN. 00:57:03.980 |
You fire up the app and click one button and now all of your network communication on that 00:57:07.240 |
device is going to go through the VPN and is protected. 00:57:09.860 |
It works on phones, laptops, tablets, and more. 00:57:12.780 |
So you can, no matter what you're checking through the internet with, you can have VPN privacy with 00:57:17.100 |
It's been rated number one by tech reviewers at CNET and The Verge among others. 00:57:22.840 |
So you should use a VPN and Express VPN is a great option. 00:57:26.460 |
So protect your online privacy today by visiting expressvpn.com slash deep. 00:57:36.280 |
You can get an extra four months free, but only when you go to expressvpn.com slash deep. 00:57:45.220 |
All right, that VPN ad got me into a technology mindset. 00:57:51.420 |
I think like my goal here, Jesse, is to give people like at least one interesting thing to 00:57:56.240 |
throw into like a dinner party conversation about technology. 00:57:59.660 |
You should come out of the tech corner with like a little tidbit you can pull out to be 00:58:03.900 |
So I want to elaborate briefly an idea that I've been playing with. 00:58:08.500 |
It came up in a panel discussion I was in recently. 00:58:10.800 |
I spoke at a board of directors meeting this morning that came up again. 00:58:14.320 |
So I'm sort of playing with this idea that there's a potential blind spot in the world 00:58:21.000 |
of AI and in particular, a blind spot about where big impacts are going to come next. 00:58:26.760 |
So when we think about generative AI tools like chat GPT and economic impacts, which is really 00:58:32.180 |
the topic that's at the heart of a lot of my reporting on AI, a lot of the focus when you 00:58:37.780 |
see people talking about products or you see the products that are being produced, 00:58:40.740 |
by the big players, particularly at like OpenAI or Microsoft or Google, is a focus on the 00:58:46.760 |
ability of these generative AI tools to generate text. 00:58:50.660 |
We're thinking about the advantages of these tools as the text they can produce. 00:58:58.620 |
So we're thinking often when we're thinking about like sort of non-tech applications of generative AI. 00:59:13.720 |
So not like in programming or data analysis or these type of things. 00:59:20.380 |
But I'm becoming increasingly convinced that the first sort of ubiquitous productivity gains 00:59:30.280 |
from generative AI, so gains that are going to be cross-industry, are going to come from 00:59:35.200 |
the symmetric ability of these models, which is to interpret text that's input. 00:59:48.040 |
And based on the understanding, it can produce text very well. 00:59:51.100 |
I think the big next ubiquitous productivity gains are going to be based on the interpretation 00:59:56.220 |
And in particular, the ability of these models to be natural language interfaces to other 01:00:01.720 |
So like the example I like to give is maybe I have a piece of software where I don't know 01:00:10.780 |
I don't really know how to do advanced analysis or data cleaning in the spreadsheet. 01:00:24.120 |
You know, I want it to like take this data and like do a regression, but I don't know like 01:00:28.300 |
what to click or where to pull or how to do this. 01:00:32.800 |
This is a place where generative AI can help because you could just say in natural language, 01:00:38.340 |
And what these models are very good at is translating between languages. 01:00:42.780 |
So it can translate what you want to do from natural English language to some sort of highly 01:00:48.220 |
structured macro machine language that the application understands. 01:00:53.300 |
I want to take out all the column, all of the rows from column B that have a dollar amount 01:00:58.020 |
And with the rows that remain, I want to build a pie chart that buckets them in intervals of 01:01:08.420 |
And the language model takes that and then spits out on the other end, a bunch of sort 01:01:12.380 |
of like very well formatted macro commands, which you can then feed to the spreadsheet. 01:01:17.860 |
That's where I think the low hanging fruit is that's going to be plucked next, or at least 01:01:22.720 |
That more importantly to me than you writing the email on my behalf is you helping me take 01:01:30.040 |
advantage of the power of software tools that already exist. 01:01:33.480 |
That's low hanging fruit productivity gains for both the individuals and for organizations, 01:01:38.800 |
because now you need less experts and you need less people. 01:01:42.840 |
So that is what I'm keeping an eye on right now. 01:01:45.560 |
One of the reasons why I think this is not being emphasized is that interpretation of text 01:01:50.840 |
and translation of it into machine language doesn't require massive models. 01:01:54.380 |
And if you're Anthropic, if you're OpenAI Microsoft, right? 01:01:58.940 |
If you're Google, you want massive models to be the thing that people care about. 01:02:03.460 |
Because you're among the only companies that can afford to create these massive models. 01:02:06.940 |
You really see the might and power of the massive models in production. 01:02:14.940 |
Look at the code it created and like how the code compiles right away. 01:02:20.100 |
Like that's where you really get into the power of it. 01:02:22.300 |
But you don't need a 600 billion parameter model to take natural language commands for Excel 01:02:28.140 |
and turn them into like spreadsheet commands. 01:02:29.920 |
You could probably train a much smaller model to do that. 01:02:31.880 |
And like a lot of companies could probably do that. 01:02:33.620 |
So they want the focus to be on text outputted because that requires the fanciest models. 01:02:39.460 |
But I really think this is the low hanging fruit. 01:02:41.960 |
And the reason why I'm pretty sure it'll be plucked is that it, again, 01:02:44.460 |
it doesn't require a 30,000 GPU, you know, nuclear power plant power data center in order to train. 01:02:52.440 |
Like much more modest models can be natural language translators. 01:02:56.160 |
And so a smaller company can build their own or multiple companies can have their own version of these models for their particular tools. 01:03:03.280 |
So that's the idea I want to throw out here now. 01:03:05.320 |
Don't just focus on the ability of language models to generate text. 01:03:11.100 |
Small agile models that unlock the power of existing software tools, I think is going to be a big deal. 01:03:17.900 |
So along these lines, like I think it's a misnomer. 01:03:22.580 |
Again, people want to think about the language model, for example, doing all the work. 01:03:30.420 |
They want to think about, like, I have a bunch of data. 01:03:32.360 |
I want to just input that data to my language model. 01:03:37.480 |
Like it's going to move through a language model and a result will come out the other side. 01:03:42.140 |
We don't want a language model that's so big and it's been trained on so many things. 01:03:47.460 |
You can give it like a bunch of data and the model itself can actually do some statistical analysis. 01:03:52.060 |
No, what you want is a really predictable, dependable, high quality statistical analysis software. 01:04:00.000 |
And the language model, you tell it what you want to do with the data and then it tells the statistical software, here's the analysis we want to run. 01:04:07.680 |
And then the statistical software does the analysis, right? 01:04:11.500 |
We want, and I think this, again, is where the low-hanging fruit is going to be. 01:04:14.320 |
Not massive models that just is like does everything you want it to do, but they unlock the things that other software does for the average user. 01:04:21.180 |
All right, so that's my big idea I'm throwing out there, trademark, trademark 2025. 01:04:27.640 |
I don't have a catchy name for it, but keep that in mind. 01:04:29.760 |
Look for natural language processing as like the next, maybe the first big killer app of this, not this oracular idea of just like I talk to this oracle and I give it data. 01:04:43.020 |
I just don't think that's the most efficient way to get the near future value out of generating that. 01:04:48.640 |
All right, so there's my tech corner idea of the week. 01:04:51.620 |
And with that, I think we'll wrap up this episode. 01:04:54.280 |
We'll be back next week with some more deep questions. 01:05:00.140 |
If you like today's discussion about how we're getting dumber and what to do about it, you might also like episode 336, which was titled On Screens and Solitude. 01:05:09.540 |
It has a lot of good ideas there as well for getting back in control of your own brain. 01:05:16.620 |
So the writer Derek Thompson, who I know and I like, has a big new feature article in The Atlantic right now. 01:05:21.780 |
Many of you sent it to me, so you probably have heard of it. 01:05:28.580 |
Americans are now spending more time alone than ever. 01:05:32.440 |
It's changing our personalities, our politics, and even our relationship to reality.