Latest / Uncomfortable Conversations with Josh Szeps / PREMIUM: "A.I. for Humans" with Prof. Joel Pearson

Transcript

PREMIUM: "A.I. for Humans" with Prof. Joel Pearson · Uncomfortable Conversations with Josh Szeps

• 0:00
  Get out humans, welcome to the safe space for dangerous ideas, here's a dangerous idea
• 0:07
  for you.
• 0:08
  What if amid all the hoo-ha about artificial intelligence and the coming revolution and
• 0:13
  whether or not we have to be worried about artificial systems taking over the world and
• 0:18
  launching nukes at us all or turning us all into paper clips, what if the real concern
• 0:23
  about artificial intelligence is right under our nose or rather inside your head.
• 0:30
  The question of how we and our societies and our workplaces and our schools and our
• 0:35
  governments and our homes will adapt to artificial intelligence over the next year or three or
• 0:40
  five or ten is one that sometimes gets ignored or overlooked in all of our hyperactive
• 0:47
  babble about the promise of artificial intelligence.
• 0:49
  Big, big changes are coming and they're going to take place mostly for most people on
• 0:55
  the ground in the real world of your life when you walk in the front door of your home,
• 1:00
  when you sign up for a university class, when you try to get a new job or when you get
• 1:05
  laid off as a result of AI systems being incorporated into the company that you work at.
• 1:11
  Professor Joel Pearson is just the person to plan for such an eventuality, he thinks that
• 1:16
  governments and institutions all over the world should be treating this like the coming
• 1:20
  of a pandemic or any other major natural disaster which is foreseeable because it is foreseeable,
• 1:27
  it's coming and we don't want to deal with it the same way that we dealt with, for example,
• 1:31
  social media use among juvenile children where eight years after the fact we realized that
• 1:37
  it really wasn't a good idea and we have to figure out how to manage it.
• 1:41
  As a result, Joel wants to collect together a bunch of brainy wants to sort of wrestle through
• 1:47
  some of these ideas and I thought it would be interesting to preview some of the things that such
• 1:51
  wants would consider in Joel's ideal universe. Joel runs the Future Minds lab, he's a psychologist
• 1:57
  and a neuroscientist, the Future Minds lab is part of the University of New South Wales and he
• 2:01
  also has a consultancy called MindX which works with big companies like Sachi and Google and Pixar
• 2:09
  and Samsung to help tailor products to how people's brains actually work so he spends a lot of time
• 2:15
  at the interface between the human and the artificial. We talk about the state of artificial
• 2:20
  intelligence as it has progressed in 2024 what we can expect in 2025 and how you should arm
• 2:26
  yourself for the changes that are coming. I hope you enjoy it as much as I did, the one I only
• 2:31
  Professor Joel Pearson. If you think back to the beginning of 2024, we currently further ahead or
• 2:40
  further behind than you would have expected us to be on artificial intelligence. Yeah, it's a
• 2:45
  funny blend of both. So in some areas it's moving ahead really quickly. I think the robotics is
• 2:49
  shooting ahead in interesting ways but most people aren't really aware of it yet and then some of the
• 2:55
  other stuff open AI I think it's you know that chat gpt's are creeping forward bit by bit. I think
• 3:01
  that's slower than a lot of people were hoping. I thought that you know be chat gpt 5 6 7 by now.
• 3:08
  So in some areas slower and it's kind of frustrating. How do they measure? Is there any
• 3:13
  objective measurement by which a 4 becomes a 5 becomes a 6? That's just their name.
• 3:18
  That's just their based on their learning and so they're working on the next iteration and
• 3:23
  probably the iteration after that. If they do a major training thing. If you think of like an
• 3:28
  earthquake scale. You know that a 4 is like 10x water 3 years or something because it's a
• 3:34
  logarithmic scale. They add letters and numbers and decimal points and things so they could be
• 3:42
  up to gpt 9 if they wanted to name it. They can do whatever they want. But I think the numbers
• 3:47
  are meant to signify the major training sets where they've had this thing training in the
• 3:51
  background on the however many billions dollars worth of computers and they finish that training
• 3:56
  and then they do the reinforcement things and they do the testing and it takes an extra 7 8 9 to a
• 4:02
  year many months to a year for them that goes public for the 50 checks. When you say that robotics
• 4:08
  is further ahead than you would have expected but maybe chat gpt is not. Is that because the performance
• 4:13
  of the large language models at the moment is more similar to what they were 12 months ago than
• 4:17
  you might have expected them to be. Yeah I think they're it's more similar but I think they're
• 4:23
  being more careful than they were originally. I think they're trying to slow down a little bit
• 4:27
  so I think where they're training at the moment is you know a fair bit in front of what's publicly
• 4:32
  available. Right so they're trying to be a bit more careful than lots of other startups are just
• 4:36
  trying to rush into things as quickly as possible. Careful by what measurement.
• 4:41
  Like like doing the red team the beat like trying to get people humans to test it to try and
• 4:49
  all kinds of things whether it be to make deep fakes whether it be to make you know biological
• 4:54
  weapons or like instruct me how to do this to not instruct teenagers to self harm or
• 5:01
  takes your own lives so all every kind of possible imposing danger rails on the intelligence.
• 5:06
  Basically yeah it's interesting isn't it I mean how is is that a race that one can keep up with
• 5:13
  or is it a game of whack-a-mole. I think it's they're trying to keep up with it but it's yeah it's
• 5:17
  more whack-a-mole. It's like hacking there's always going to be people ahead of the curve they're
• 5:21
  going to figure out how to break the machine or twist this or get under the hood of the thing
• 5:27
  yeah and make it work in different ways. And I've heard some murmuring slightly from AI
• 5:32
  skeptics that we're sort of beginning to see the edges of how useful
• 5:38
  like these training models can be in the sense that they've essentially read the entire internet
• 5:44
  now and they're bumping up against a natural ceiling where the exponential growth of the past
• 5:49
  a couple of years. Yeah so there's this talk and it's hard to know whether that's some of that
• 5:55
  might be true but for my understanding it's there's no real evidence for that so people are saying
• 5:59
  oh we're reaching a plateau now and everything's going to slow down but I don't see there's any
• 6:05
  real so there's different yeah there's the data sources and there's real data so there's two
• 6:09
  things that are happening people they're changing the way the models are learning so the the
• 6:16
  the new version of chat GPT which thinks if you use this one so you ask it to do something and
• 6:23
  I want to just give you the answer straight away so it thinks and it checks its own answer
• 6:26
  and iterates through it a few times right so they've added on reinforcement learning with human
• 6:31
  interaction so that is not just the brute force black box training that is getting humans to then
• 6:38
  shape it a bit more to try and get more logical reasoning in there so it can rethink
• 6:43
  in longer and deeper way by humans you mean the user yeah users right so you're saying that I
• 6:48
  would have an experience where I would ask chat GPT question and it would then ask me to
• 6:54
  contribute to its answer so they'll do a thing where yeah so they do it the general public I think
• 6:58
  as well I'll say here's two different answers which one you've seen that yeah that's like help
• 7:03
  getting us to help tune their model a bit yeah but they also have internal teams of doing that
• 7:08
  in more sophisticated ways right interesting so there's that kind of learning on top of the other
• 7:12
  kind of learning then there's synthetic datasets they're creating so they have other teams
• 7:16
  creating synthetic datasets defer the train on what's a synthetic dataset so it's rather than pulling
• 7:21
  you know all the videos from internet or all the copyrighted material which is all the legal
• 7:26
  all the the lawsuits going on at the moment it's creating a dataset using using different
• 7:32
  AI's to create datasets than any other AI's to train on that and dataset being just a bunch of
• 7:37
  material written stuff videos photos it could be anything and does the new AI it sounds incestuous
• 7:44
  that sounds like yeah it sounds like how disease where you're hitting the account back to the
• 7:47
  back to the back to the meat and it kind of it is because if yeah how do you know it really works
• 7:52
  really so yeah and an AI can just generate a whole bunch of a different AI that's independent
• 7:58
  that's the other one yeah and then you tell the other one read that learn from that yeah that's
• 8:03
  interesting and that buys you something where is the newness being injected into that closed system
• 8:10
  I think there's a lot of I'm not 100% sure I think because the the training sets I guess from
• 8:17
  one set will be different there's noise injected into that yeah I'm not sure where yeah the
• 8:26
  newness comes from then you have obviously you know robots that are now walking not only amongst
• 8:32
  officers and factories they're walking out in the wild you can see the Tesla robot now trying to
• 8:37
  walk up hills and fall over down grassy slopes and things and so that's learning so you think of
• 8:42
  that's all happening in the background and there's what say 15 of these humanoid robotic companies
• 8:47
  working on these things and all those senses of pulling in data you already go to some places I
• 8:51
  was at a restaurant in Singapore and like a robot robot they're just whizzing around everywhere they
• 8:56
  take the take your order yeah yeah yeah yeah deliver things to you well they have that clean
• 9:00
  I've seen that in in Santa Monica yeah well you walk down the street and there's little robots
• 9:05
  scooter by and I think it's going on they're like that's like Uber eats kind of delivery thing
• 9:09
  the future yeah I was like yeah yeah Australia bizarre okay so anyway the point is that all like
• 9:14
  just like all those sensors to have many cameras on a Tesla are learning all the time recording all
• 9:18
  the time or so that's a massive brand new set of data robotics have brand new it's creating tons
• 9:25
  of tons of data all the time so you have all that fresh data coming in so the more robots doing
• 9:31
  things the more fresh data sets hmm so I yes I know I there must still be a huge amount of visual
• 9:38
  data that the systems don't yet have and haven't ingested because I've heard you know conversations
• 9:45
  between AI people about you might ask it to you might give it a photo and say can you change the
• 9:51
  person in this photo into a dragon and have like a cat flying by and the AI will do a really good
• 9:58
  job but there'll be some common sense things that it doesn't know like that a flying cat I don't know
• 10:03
  can't be inside a closet yeah or something like that so the laws of physics don't yeah in other
• 10:11
  words but maybe if but the idea is if we just train them if we just now that they've learned all
• 10:15
  the words we just have to train them on all the photos and all the videos you've ever been taken
• 10:20
  and then they'll gain an intuition about how the natural world works and why cats can't be flying
• 10:26
  inside cupboards yeah and I think at the moment all the video models are trained separately to the
• 10:30
  word models and the other or the the large language models and so yeah if you combine those you
• 10:35
  might get something interesting my understanding looking at all the output of all the video platforms
• 10:41
  is that they don't learn the rules of physics very well from training just on videos and some like
• 10:46
  the gravity won't be quite right you know horses running will kind of float a little bit like gravity
• 10:50
  is not quite working sort of bounce legs will make it just weird there's a bit strange gravity physics
• 10:56
  things that get a bit wonky interesting but I'm sure they'll get they'll get they'll get proved
• 11:02
  yeah so I mean where I'm at with this is really the big picture so let's think five 10 years I do
• 11:09
  think that everything that uses electricity will be influenced in by aliens some way so everything
• 11:16
  in your house your car you obviously already tech and then if you think about that for a second that
• 11:21
  will filter down to change basically every aspect of society and our lives so it's going to change
• 11:28
  obviously education jobs and there's you know every consultancy firms trying to figure out how many
• 11:34
  jobs going to be lost from AI it's pretty hard to figure out accurately estimates to range from
• 11:40
  20% up to 60% in 10 years and so if you think about that just by itself right you think of take
• 11:49
  any country and you take away 60% of the jobs even over a five-year period 60 even over a decade
• 11:56
  that's pretty disruptive in itself then well I mean we've seen what happens when you when like
• 12:02
  coal mining becomes not a thing in Appalachia and you know everybody gets addicted to oxy
• 12:09
  content and votes for an ultram but not where I was going Josh but I mean guys like the kind
• 12:15
  social disruptions that happen when they're large scale yeah it's not a simple thing and it has
• 12:20
  yeah but I in their education and so I you know I work at a university and the university is still
• 12:26
  spend the last two years arguing over whether students should be allowed to use chat GPT for their
• 12:30
  essays right and my thing is not only you know sure we'll have AI's teaching and there already
• 12:35
  is Khan Academy and these interesting products and services already but it's more like what is
• 12:40
  even the role or purpose of a school or university in 10 years if you all have our personal AI built
• 12:46
  into our wearables our clothes has our history it's all sealed off with privacy we don't have to
• 12:52
  worry about that it has it's tapped into my physiology there's also changes it knows when I'm stressed
• 12:57
  it knows all my habits it can pick me up on you know cognitive biases it can help me make
• 13:01
  decisions it has access to all the information in the world and let's say it has more compute than I do
• 13:07
  more intelligence um what is the role of me going to school my kids going to school what do
• 13:13
  we need to learn at university do we still even need those two things mmm and I think the answers
• 13:18
  straight out no or yes some probably somewhere in between but the point is we don't we don't
• 13:24
  have an answer for that right so just pause on the academic question and the and the sort of
• 13:28
  governance question and just dig a little deeper in this world that you're articulating about
• 13:33
  you know wearables and so on because I think AI has become such a buzzword that you know
• 13:38
  a a home electronics company will just say that it's got an AI and the light bulb and the light bulb
• 13:44
  just means that my bananas AI and they're just means that the light comes on when you walk in the room
• 13:50
  or something yeah you've all know a harrari you know tries to define it by saying it's AI if it can
• 13:57
  make in decisions about what it is that you want without having to ask you about what you want
• 14:04
  especially if it's making those calculations on the basis of your of your state of your physiological
• 14:10
  status you were just saying yeah so you know a coffee machine that turns itself on at seven in the
• 14:14
  morning and makes you a cup of coffee is not an AI no a coffee machine that notices that you had
• 14:20
  a bad night's sleep and therefore you select the extra strong blend of coffee and bruise it at
• 14:26
  precisely the moment that you're going to want to drink it it's AI yeah it adds extra things to it
• 14:31
  adds extra whatever your sense of the guana and the garana nicotine the area all the good
• 14:35
  the nootropics to your car it changes the lighting it changes puts certain music on it does this
• 14:40
  it have what your schedule for the days is adjusted it changes your exercise routine and whatever
• 14:45
  hundred other settings and how does that bump up against my desire to feel like I am the agent
• 14:51
  of my life and not being bossed around by devices yeah I don't think we know the answer that
• 14:57
  that research hasn't been done and it will probably get done by engineering firms rather than
• 15:02
  scientists because everything's moving so quickly I think that the agency in the because it's
• 15:06
  very important that we feel like we have the freedom of free will even if we don't I don't think we
• 15:11
  do it but that's for another day we've had long conversations on the radio about free will
• 15:17
  taking calls from listeners yeah most of them are universally believe that we do have free will
• 15:22
  and we have to be disabuse of the naive notion but we do we know from mental health work that it
• 15:28
  it's very important that we feel like we do independent of whether we do or not yeah and just
• 15:32
  that we we feel like our tech is is servant is a servant of ours right yes not the other way around
• 15:37
  yeah um but I mean the counter argument to that is that we just push out choices and freedoms to
• 15:43
  other things that we want to do and then so we will quickly get into sort of universal basic
• 15:47
  income territory there so you know there is is they are going to bring on the next renaissance
• 15:50
  so we're going to have this sitting around you know playing guitars or doing poetry painting
• 15:55
  right that kind of we get to just play around doing our hobbies and enjoy life
• 16:02
  kind of like what people think of retirement should be even a retirement doesn't often
• 16:06
  doesn't work that way it's actually quite bad for you and bad for your health
• 16:09
  well yeah wouldn't that whole thing be bad for your health I mean when you see I don't think
• 16:14
  universal basic when you see spoiled spoiled rich kids or people who spend all their time playing video
• 16:20
  games and drinking slurpees like you see the way most people spend leisure time and it's not
• 16:28
  it's fulfilling no it's not and yeah it's a return to bad for your health most people who go and
• 16:34
  holiday I think they want it they want it they're going to be feel great if they do nothing
• 16:38
  after a week they can't stand it I mean not everyone but most people I'm like that yeah and often
• 16:42
  we get that wrong it's one of that we think all we want to do is have you know no job do nothing
• 16:47
  and as soon as we're there we fall apart and mental health problems arise and physical health
• 16:55
  anyway so going back to and to agency I think the idea would be we can push agency to other parts
• 16:59
  of our life and we would use or I guess the AI is to take care of things we don't want to do yeah
• 17:04
  and this will become an engineering problem basically and a customer experience user interface problem
• 17:10
  one out of how much can you hide what the AI is doing so that the person is unaware of the way
• 17:16
  in which they're being manipulated by the AI perhaps so we still feel like we're in control so if
• 17:21
  I've got a in your analogy there where you've got wearables and the AI is one step ahead of you
• 17:26
  because it understands your physiology let's say I'm cranky yeah I'm wearing a watch that is
• 17:30
  measuring my physiology maybe my glasses are as well and maybe the maybe my living room is as
• 17:36
  well there's some integrated system right that's monitoring me my bed probably knows that I had a
• 17:42
  bad sleep and what there are ways for artificial intelligence in five or 10 years to restructure my day
• 17:51
  or provide me with suggestions sure be less than that yeah it'll be a couple of years I think for
• 17:56
  that right but I think yeah I mean think of it as like your personal assistant that's with you 24
• 18:01
  seven that you will be able to trust in other words the way the hackable you want to worry about
• 18:07
  privacy issues and when you say you'll be able to trust it and it won't be hackable and we
• 18:12
  want to worry about privacy issues one thing that looms large for me is here is that there are a
• 18:20
  couple of economic models for technology one is think of a video game for example or a smartphone
• 18:26
  one is you just pay a lot of money upfront and you buy the thing and then you own it and it's
• 18:30
  private another is a company gives it to for free and you become the product and advertisers pay
• 18:36
  the company for access to your attention this might be the difference between apples selling point
• 18:43
  and you know the Google ecosystem where Google will sell everything about you and make things cheaper
• 18:47
  and Apple reassures you that what happens on an iPhone stays on my phone or a video game that
• 18:52
  tries to coerce your children into getting to it's free but it tries to coerce your children to
• 18:57
  getting to have a higher levels where they can buy in game benefits and things like that so it
• 19:02
  keeps them on a treadmill and shows them ads rather than something you buy outright.
• 19:07
  Is there a risk that we end up with a two tiered AI world where I can afford you and I can afford
• 19:12
  the expensive one that's private but the the ordinary one that people use is using their physiology
• 19:20
  in their state of mind to prey upon them and suggest to them that the correct solution might be this
• 19:26
  new new or tropic that has you know just become available and and deliver it to you for just 2499
• 19:31
  with a month the three months trial just by yeah I mean we're already there right there are free
• 19:40
  versions and there are paid versions already now the leading open AI has said that they're not
• 19:47
  going to do advertising or ads there was something I think they you know in one of the hundred
• 19:53
  interviews that Sam Altman's recently done I think he said that well the door is now open
• 19:57
  to potential advertising so there's there's as that whole business model of theirs and the structure
• 20:01
  of their company changes from this original not for profit to this for profit they're moving rapidly
• 20:06
  towards they will probably introduce advertising of some sort in some way I'm sure a low tier
• 20:13
  and the other thing that is going to reshape is like it basically looks like the US and China
• 20:18
  are leading the way in AI fairly substantially now and then Australia and Europe
• 20:25
  UK a little bit with some of Google's work there but it's really going to be the US and China
• 20:31
  and so there may be this sort of divide between the people or nations that use products from one
• 20:36
  area or use products from the other area whether China opens those up to everyone or not
• 20:41
  is still a bit unclear so there'll be that kind of what they call a silicon curtain sort of idea
• 20:49
  that the world will be sort of split up based on which AI products from which companies that's
• 20:54
  interesting that's part of the global arms race for lack of a better word to try and you know
• 21:00
  country supporting companies and companies racing as fast as they can to dominate the AI space
• 21:07
  and just to put a flag also in the in the near future here which is that there's this sort of
• 21:13
  holy grail which people call AGI or unofficial general intelligence which we don't yet have
• 21:19
  what do you understand that to be because the people who think that we'll get there so and think
• 21:23
  it'll be as revolutionary as anything so there's all I mean there's different definitions of that
• 21:29
  and there's not I don't think it's that interesting to argue about all those definitions but
• 21:32
  the idea that that an AI system can do everything a human could do and more or above that capacity
• 21:40
  so it's the idea so at the moment you can have you know even our phones can do things we can't do
• 21:45
  so it's not like being able to do math that we can't do or something like that it's being able to
• 21:49
  do things generally from controller car through to do something in the kitchen through to do videos
• 21:55
  through understand the world or it so we can move through all these different spaces and do
• 21:59
  different things then you get something towards this general this generalizability or super generalizability
• 22:05
  all these different phrases people use for that right so and to come up with innovative solutions
• 22:10
  to things the way that a human that can be part of it I don't know if that's generally included I
• 22:15
  think it's more the fact that it once you can generalize across all the different things that a
• 22:19
  human can do rather than have narrow different AIs one for driving a car one for doing writing one
• 22:24
  for doing film one for doing this photos one for doing something else once one think one can do
• 22:30
  all of those things then it's generalizable right and I mean I sometimes think in try to put myself in
• 22:37
  the mindset of the future looking back on the moment that we're currently sitting in in the same
• 22:42
  way that we might look back on like I've said before for example that when 9-11 happened which
• 22:50
  is within the living memory of people of our generation it's hard to remember that not only
• 22:56
  were there no iPhones and YouTube and you know let alone Facebook and Tiktok and Twitter there were
• 23:04
  no iPods remember the original white iPod the big wheel that you scrolled your thumb around that would
• 23:12
  like wait a brick like the same way as a brick and kept three songs on it that was released just
• 23:19
  after 9-11 and then you sort of think back and you go wow so that was a world where and obviously
• 23:25
  when we were teenagers you know you made a plan with your friends to meet at the movie theater and
• 23:32
  you would go there at that time and there was no way that you could contact them once you went
• 23:37
  to the house you just stood there and if they didn't show up you just stood there looking at people
• 23:43
  passing by and gazing at the clouds until eventually you just wandered off because you got
• 23:48
  blah blah waiting for them right there was no way to even send a text message to them and that
• 23:52
  seems like such ridiculously ancient history to a Tiktok generation now and so I tried to sort of
• 23:58
  use that as a way of noticing the pace of technological change that we're living through and think all
• 24:05
  if you extrapolate out from that into the future and assume that we've got like exponential rates
• 24:09
  of change then there will soon come a time when we look back on this moment that we're sitting
• 24:15
  here together and we'll go I can't believe that was a world in which we didn't what exactly like
• 24:22
  what's the answer that what is the most what will be the most striking thing about the technological
• 24:30
  world we live in now when we look back on it in five or 10 years my guess is that our everyday
• 24:36
  lives will feel like we're surrounded by creatures that are artificial I don't mean physiological creatures
• 24:42
  but I mean yeah that we will be in constant conversation with creatures that with entities that
• 24:50
  fields that seem to be sentient to us and seem to be as intelligent as we are and that those that
• 24:56
  will just always be interacting in a series of conversations with those things or with a thing
• 25:01
  yeah that is our personal systems that is our travel arranger that is how lawyer that is how
• 25:07
  our accountant will basically all have a failings of really smart useful creatures yeah so everything
• 25:16
  like 24 years from now I think it will be like that we'll have artificial humans we'll have
• 25:22
  artificial intimacy we're I mean we're already having the next big revolution taking over from
• 25:26
  TikTok is obviously artificial intimacy we can swing back to that in a second but yeah I think
• 25:31
  you're right we'll have we'll have all kinds of deep relationships with things that we'll call
• 25:35
  now called artificial but by then I don't think we'll care whether they are or not I don't think
• 25:43
  universal basic income utopia thing is going to work the way people keep talking about it that
• 25:48
  we're just going to have them to do I don't think I think humans are too competitive to greedy to
• 25:52
  human nature doesn't work with universal basic income right people want to outdo the other
• 25:57
  and we've seen that every time we have a new you know I can do my email on the phone fantastic it
• 26:01
  doesn't mean that extra time it buys I go off and write poetry I mean I just work hard I
• 26:07
  write peat with my neighbor or whatever although just in defensive universal basic income I don't
• 26:10
  think you need to be one of the polyanna idea lot of who believes that universal basic income is
• 26:15
  going to lead to all of us just doing hobbies all the time to nonetheless defend on the economic
• 26:20
  grounds a universal basic income to smooth out the social disruption of a huge you know the
• 26:27
  redundancy of massive sectors of the workforce you could just use it the same way that we use the
• 26:32
  idea of social security or yeah for all or something we're like you know let's just make sure that
• 26:38
  everybody has thirty thousand dollars a year and then they can keep working so that kind of thing
• 26:42
  could work but yeah I think we just have to be careful with the psychological side of that because
• 26:45
  a lot of the talk tends to evolve to hear the rest of this conversation go to uncomfortable
• 26:50
  conversations.substac.com slash listen and you will get your own personal premium podcast feed
• 26:57
  with at least three extra episodes of the podcast every month and heaps of extra stuff including
• 27:03
  the remainder right now of the fabulous conversation you've just been hearing if it was worth listening
• 27:08
  to this much of don't rob yourself of the rest pull out your phone right now and search for
• 27:13
  uncomfortable conversations with Substac.