Cheating Insomnia. The New Science of Sleep

Richard Bradbury: If you found last week’s episode of mattsplained a little bit too manic, don’t worry. This week his delivery will be a lot sleepier – because that’s what we’re talking about: sleep, insomnia and the breakthroughs that may allow us to finally get a handle on the science of sleep.

Matt Armitage:

• Genuinely I’m not sure if this week will be sleepier.

• Insomnia is all about not sleeping. And a lot of the time that mania comes from not sleeping.

• Especially with me.

• This is a topic that we’ve covered on msp pretty regularly over the years.

• That’s partly because I am one of those people that has trouble sleeping.

• And have had ever since I was a kid. Even when I was in primary school, the sun would regularly be up before I slept.

• Which isn’t as bad as it sounds – in the summer in the UK, the sun gets up pretty early.

• But even still – for an 8-year-old to be sleeping at 4 or 5am.

• It’s a miracle that I turned out to be so well-balanced.

• [pause]

• Richard was supposed to say something there. I guess he just has nothing to say about how normal I am.

• So. I’ll move on.

• But before I do – a quick diversion into the world

Richard Bradbury: Did you say text to image? Sorry, I was taking a nap…

Matt Armitage:

• That’s insulting on two fronts:

• Firstly, sleeping during the show – rude.

• Secondly, taking micro naps is just rubbing the inability to sleep in my face.

• But I will soldier on because I’m a professional.

• Pause

• No comment from Richard again.

• Text to image – we’ve talked about this is a few times this year.

• Not in detail. We’ve mentioned models like Open AI’s DALL-E2 and and Google’s Imagen.

• Which can put together really cool concept images from text inputs.

• And more recently I’ve been playing around with MidJourney which is a Discord based public beta.

• And to quote a friend of mine from a conference stage this week:

• I really hate Discord.

• I echo that sentiment. Discord is a bit like a Jack Kerouac novel.

• Dense, impossible to navigate. Fun the first time and something you never want to go back to

• Funnily enough, I don’t think Kerouac did a lot of sleeping, either.

• Though I’m not sure if that was driven by chemicals as much as it was insomnia.

Richard Bradbury: The babblings of a sleep-deprived madman?

Matt Armitage:

• Just for the record, Richard isn’t aiming that comment at Jack Kerouac.

• As reported in Futurism, Google has just released a concept paper and sample images for a new 3D image generator called DreamFusion.

• What does a text to image generator do?

• Exactly what it says. Type Hiker meeting a tiger in the jungle.

• That’s a real term that I used to generate an image with MidJourney.

• The AI takes your commands, including prompts for lighting, backgrounds and visual effects.

• And turns your sentence into art.

• Like anything AI related it can be hit and miss – you may have to tweak your terms a bit,

• or simply repeat the original prompt until the machine’s vision more closely resembles the idea you have in your head.

• Due to the computing and processing grunt it requires, creating 3D images is much more processor intensive than creating 2D images.

Richard Bradbury: You did say during last week’s show that Ethereum’s Merge is creating a glut of GPUs on the market. Couldn’t that provide some of that 3D imaging power?

Matt Armitage:

• I’m touched that you remember what we talk about during the shows.

• I don’t think it relates to this instance so much because it’s not about local processing power.

• These images are created by neural networks in the cloud.

• At least they are for now.

• It’s possible that we might see hybrid models in the future.

• A lot of our mobile devices already make use of AI embedded on them locally,

• while reaching up to the cloud for more intensive tasks.

• For general consumers – tools by OpenAI and especially Google feel more natural to use.

• Imagine if Google embeds text to image in its search page options?

• That’s a flat learning curve compared to what MidJourney is doing.

• Which takes me back to I hate Discord. You don’t want to have to learn the platform in order to use the tool.

• And availability and accessibility can – it’s not a given – but they can bring more success than having a superior product that has a much higher cost or adoption curve.

• And Google has that enormous edge in that millions of us use its marquee search product numerous times a day.

• Back to DreamFusion – surprisingly, Google’s previous 3D modelling text to image system, DreamField, was trained on an OpenAI model.

• But for DreamFusion they’ve moved back in-house and trained the model using their own proprietary Imagen system.

• So the images are actually rendered in 2D first and then upscaled, which helps to reduce that processor load.

Richard Bradbury: Is there a beta we can try?

Matt Armitage:

• There are links to the paper on github in the show notes and lots of pretty images.

• They’re not 100% realistic, which shows the stage that the project is currently at.

• But anyone who has been playing with Midjourney over the past few weeks can see how rapidly these services improve.

• Google isn’t committing to an specific time frames to allow public access to the model.

• Let alone when we might see it as one of their products.

• But if they do get something that works in the near future: imagine what it could do for game design, CGI for movies.

• As well as allowing us to create all kinds of flights of fancy, perhaps from our browser?

• One of the things that is lacking in the current model of Midjourney is the ability to return to characters or creations that the machine has come up with.

• We talked about ReSpeecher last week – the company that has created a realistic Darth Vader text to speech model for Disney.

• If a commercial version of DreamFusion includes the ability to create individual entities and develop them.

• That changes the game.

• As I said to a friend this week – using tools like Midjourney or Dreamfusion means he’s not a designer anymore.

• He’s a writer.

Richard Bradbury: Are we not going to talk about Elon Musk’s Optimus robot reveal last weekend?

Matt Armitage:

• Not today – although we will be back on the bots soon.

• There’s actually a piece about the launch and the convergence between EV technology and robotics on the Kulturpop substack.

• I talk a little but about the launch, how the development of battery technologies and AI control systems for their cars is fuelling the company’s rapid advances with robot building.

• You can find that at Kulturpop.substack.com.

• I imagine there will also be something on Musk’s latest flip flop on Twitter.

• Or maybe there won’t. Because at his point who cares?

• And at the point of recording this – no one seems to have request a pause in the trial based around Musk reneging on the deal.

• we may get to see Musk’s lawyers fighting to get out of a deal their client now wants to complete.

• While in the background, the consortium of banks and debt providers behind Musk have their own issues with the deal as current market condition could force them to move ahead knowing they will make a loss.

• So, that will be a suitably Kafkaesque continuation of a huge story about a platform that only the people who write huge stories cares about.

Richard Bradbury: That’s virtually the whole first half and we haven’t started on sleep.

Matt Armitage:

• Yes, so we’ll get into that after the break. First a quick note, also about sleep.

• A lot of people use those white noise generating apps to help get babies off to sleep.

• And they have been shown to help people with ADHD.

• But a lot of people don’t like the sound of white noise, which, especially at high volumes has a lot of high frequencies.

• Enter brown noise. Which actually mimics the movement of pollen particles in water.

• Users claim that it’s a lot more pleasing than white noise, because it has fewer of those higher level frequencies.

• So what you’re left with is a more muted mid-range kind of sound.

• It’s a bit of a Tik Tok sensation at the moment, with people documenting their responses to hearing it for the first time.

• And claiming that it helps to calm them, drift off to sleep and allows them to concentrate more.

• Have you messed around with this kind of stuff?

Richard Bradbury: replies

Matt Armitage:

• I’ve always hated the white noise stuff. It doesn’t make me calmer.

• Quite the opposite. And while white noise has shown clinical benefits for people with ADHD

• I do find brown noise – some of it at least – quite relaxing.

• I’m not sure if it’s enough to send me to sleep. But scientifically the evidence for brown noise isn’t there yet.

• It’s unlikely that brown noise has any magical properties that white noise doesn’t have.

• It may simply be that people find the sound more appealing.

• That’s not a criticism – if you like it, you’re more likely to use it.

• But as to what it’s doing to the brain – some scientists have pointed out that white noise mostly has therapeutic effects at high decibel levels.

• So what you might be experiencing listening to it on heapdphones or your computer speakers is really sound masking.

• The frequencies of the sound are simply blocking out background noise and allowing you to concentrate with fewer distractions.

• So, yeah, if you think it works for you, keep going. But as with any sound – check the levels and keep your ears safe.

• Because I am not a doctor.

Richard Bradbury: Don’t nod off. We’ll be back with better ways to sleep after the break.

BREAK

Richard Bradbury: We’re talking about insomnia today, specifically about advances in the methods of treating it. The first question guess is:

Is this a solution in search of a problem?

Matt Armitage:

• It’s thought that up to a third of people have trouble falling asleep and / or staying asleep.

• I’m going to race through stuff in this half – for more depth you can read an NS article by Dave Robson titled

• We are finally waking up to the causes of insomnia and how to treat it

• While up to around 10% of people may meet the clinical definition of insomnia which is that it occurs at least three times over a week over a 3-week period.

• And leads to cognitive impairments during the day.

• And that it’s devoid of external factors – like a crying baby.

• Crucially, women are more likely to suffer from insomnia than men, the relevance of which we’ll get to in a minute.

• I’ve always been in the sleeping problem category – I don’t know about you?

• Which camp are you in?

Richard Bradbury: replies

Matt Armitage:

• Currently I’m in a pattern of falling asleep really late, waking up at 4am whether I’ve been asleep for 3 hours or half an hour.

• Then being awake for 1.5 to 2 hours before dropping off and waking up at 9am feeling exhausted.

• However, when you speak to someone who doesn’t have trouble sleeping – like my wife.

• They look at you with bafflement and say things like:

• Why don’t you go to bed earlier? Why do you have music or a radio on?

• Because for them, sleeping is lying down and closing their eyes.

• Whereas for millions of us, sleeping is about performing some kind of magic routine that tricks our bodies to sleep.

• And that’s where the gender gap may play a part as well.

• Women are 60% more likely to experience insomnia than men.

• And as we see across the board in medicine – health conditions that primarily affect women are more likely to be trivialised

• And are more likely to be under researched.

• Of course, this show contains no gender diversity so I can only ask you:

• Do you have a sleep ritual or methodology?

Richard Bradbury: replies

Matt Armitage:

• For me, it’s listening to audiobooks. But it has to be a single voice, not a multi cast recording.

• More than one voice and it’s stimulating my brain too much.

• The same with music - I start analysing it and separating all the parts.

• Radio normally has more than one voice, there are ad breaks, music, news reports.

• My brain has all of those things going on at once, anyway.

Richard Bradbury: So audiobooks are your brown noise?

Matt Armitage:

• I guess that a really good way to frame it.

• With the monotony of that voice, with a tempo to the way they speak that you can predict,

• My mind can slip away. Whereas, if I try and listen to the same thing night after night,

• I drift back into thoughts because I already know what’s coming next, so I don’t focus on it.

Richard Bradbury: But none of this is science…

Matt Armitage:

• No. This is a problem in search of a solution.

• Lack of sleep or poor sleep comes with costs.

• Depression, heart disease, Alzheimer’s, diabetes. All the fun stuff.

• And there are also business costs. Sleep deprived workers reduce productivity.

• They may take more time off because of associated health conditions.

• But until now, our approach to insomnia has been of the one size fits all.

• Get a better mattress. Improve sound proofing in the room. Avoid light bleed and electronic device screens.

• Cut down on caffeine and alcohol.

• Take sleeping pills or anti-anxiety medication.

• Which ignores the fact that the roots of those sleeping problems are not uniform.

• Take me – I’m an alcohol and caffeine free zone. I can take a sleeping pill and lie awake all night.

• Thankfully, as we’re seeing in so many medical fields now, we’re moving away from this monolithic approach.

• We have the technology and the resources to look at tailored and individual responses that addresses the problem from those root causes.

• And helps to combat them throughout that therapeutic chain.

Richard Bradbury: Are we talking about things like battling the nocebo effect?

Matt Armitage:

• Yeah. It’s interesting you mention that. Increasing amounts of research suggest that medications that regulate things like melatonin levels are largely ineffective.

• They may have a short-term effect but even that is uncertain.

• The same with benzodiazepines – the medications typically prescribed for sleep disorders.

• You may experience benefits if you use them very occasionally but less so if you depend on them regularly.

• I’m not going to get into the chemistry of that – there’s not really time.

• But that nocebo effect – yes. A lot of people with insomnia go into what’s called a ruminative state at bedtime.

• Essentially your brain goes into overdrive when it should be trying to power down.

• It’s not catastrophizing – but it’s easy to get locked into cycles of negative thoughts.

• So during the daytime, you start to dread sleeping. The nocebo effect is the opposite of the placebo effect.

• You start planning for not sleeping later that night. Expecting it. So you fuel and fulfil your prophecy.

• In fact, one of the interesting outcomes of the nocebo effect is a group of people that the NS article refers to as complaining good sleepers.

• These are people who sleep fine but convince themselves they don’t.

Richard Bradbury: Can we backtrack to the ruminative brain…

Matt Armitage:

• Yes, so you might think – oh he just means people are having thoughts at bedtime.

• We all have thoughts at bedtime.

• Research from the Netherlands Institute for Neuroscience suggests that there is an inertia in the ruminative brain.

• It’s fixed. And there is often a corresponding anxiety arousal in the amygdala.

his chimes with research from: 2019

• Which showed that people with insomnia have reduced connectivity between the front lobe – that control module of the brain…

• And other regions like the insula, where a lot of that emotional information is processed.

• Which can lead to greater problems moderating stressful or emotional stimuli.

Richard Bradbury: So how do we fix it?

Matt Armitage:

• Behavioural therapies are becoming increasingly important to sleep reform.

• And I use that word reform deliberately.

• These are structured, tailored strategies that work on the root causes of one person’s sleep issues.

• So it is about changing the way that person approaches the task of sleeping.

Richard Bradbury: Isn’t calling sleep a task part of the problem?

Matt Armitage:

• Which explains why I’m desperately in need of one of these interventions.

• One approach is CBTI - CBT for insomnia.

• These are structured sessions where a therapist helps to devise strategies for the sufferer to identify and then break out of those ruminative patterns.

• One way may be to focus on staying awake rather than falling asleep.

• Which is a little counter-intuitive but is designed to overcome the anxiety that comes from failing to fall asleep.

• It’s called paradoxical intention.

• Studies have shown the CBTI approach to improve sleep in around 70% of cases, with up to 40% reversing their insomnia.

• Although, it’s important to note that this isn’t a cure in the sense that you will never experience sleeplessness again – which is why researchers refer to it as going into remission.

• The biggest holdup with this methodology at the moment is the lack of trained sleep specialists.

• So a section of the sleep industry is looking at whether digital tools – AI powered apps – can deliver some or all of the benefits that CBTI has.

Richard Bradbury: Does the app approach work? Isn’t there too much of that risk of download, try and forget?

Matt Armitage:

• Yes – I agree. Also, because there are lots of apps which are designed to take your money and have very little interest in your experience or results.

• The NS uses the example of Sleepio – an app that is being championed by health authorities in a number of countries.

• It consists of a tailored, AI guided 6 week therapy course.

• First, it’s not tying you into any pay forever type structure, and is approved, and usually paid for, by those health authorities or medical providers.

• The AI works to tailor the course to the data you put in.

• This may be in the form of a sleep diary or smartwatch or other wearable.

• So far, trials have shown the app to be comparable to in-person cognitive therapy sessions.

• Other apps that are coming out include SleepSpace which does something similar but has some neat additions

• – we’re back to brown noise again – SleepSpace uses your phone’s speakers to play bursts of sounds at you throughout the night that are designed to induce the frequencies of brainwaves needed for restful sleep.

Richard Bradbury: Do you see yourself sleeping any time soon?

Matt Armitage:

• Like a lot of people with sleep issues, I spend an awful lot of time thinking about sleep.

• Wishing I was asleep.

• So, yes, I see myself sleeping tonight. How well and for how long – who knows?

• But I hope that soon I’ll have the opportunity to try out some of these apps and approaches.

• Because people shouldn’t be spending a great deal of their waking hours thinking of sleep.

• And because I would love to experience what seems so easy to other people.

• To lie down and go to sleep.