Latest / Best of LinkedIn: Digital Transformation & Tech / Best of LinkedIn: Health Tech CW 20/ 21

Transcript

Best of LinkedIn: Health Tech CW 20/ 21 · Best of LinkedIn: Digital Transformation & Tech

• 0:00
  Speaker 0: This episode is provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts of HealthTech in CW-Twenty and Twenty One.
• 0:08
  Speaker 0: Frenness equips health tech providers with the market intelligence to identify which hospitals target.
• 0:18
  Speaker 0: Yeah,
• 0:22
  Speaker 1: so imagine leaving your office on a Friday right?
• 0:25
  Speaker 1: And you come back on Monday morning to a three hundred and forty thousand dollar server bill.
• 0:29
  Speaker 0: Uh excuse me just from the weekend
• 0:31
  Speaker 1: Just over one weekend because uh- A single piece of software got stuck reading a patient file.
• 0:38
  Speaker 1: I mean this isn't some hypothetical fear mongering.
• 0:40
  Speaker 1: it actually happened to a founder just a couple weeks ago.
• 0:42
  Speaker 0: Okay let's unpack.
• 0:44
  Speaker 0: Welcome to the deep dive, everyone.
• 0:45
  Speaker 0: We've spent some time sifting through the absolute best insights sourced directly from health tech leaders on LinkedIn over calendar weeks twenty and twenty one
• 0:53
  Speaker 1: right?
• 0:53
  Speaker 1: The real On-the-Ground stuff
• 0:54
  Speaker 0: exactly.
• 0:55
  Speaker 0: And our mission today is for you to figure out what is actually working in HealthTech Right now and What Is just dangerously expensive hype.
• 1:03
  Speaker 0: yeah and looking at the data we have we are officially entering what looks like a disciplined execution phase.
• 1:11
  Speaker 1: Oh, absolutely the days of just throwing AI at a wall and seeing what sticks those are entirely over.
• 1:16
  Speaker 0: they have to be because if an Ai tool doesn't actually fit into the grueling you know highly regulated reality of it Tuesday afternoon in a crowded clinic It is worse than useless.
• 1:26
  Speaker 1: yeah becomes a literal liability.
• 1:28
  Speaker 0: but here Is where I want to push back right out of the gate.
• 1:31
  Speaker 0: when we talk about ai failing In a hospital setting We immediately blame the technology.
• 1:36
  Speaker 0: Sure, yeah but looking at the insights from this week Technology adoption in a hospital isn't actually a tech problem At all.
• 1:42
  Speaker 0: it's a human
• 1:43
  Speaker 1: one.
• 1:44
  Speaker 1: that is such a fundamental distinction.
• 1:45
  Speaker 1: I saw some incredible Insights for Mino Shaw on this using the scarf model scarf mall.
• 1:50
  Speaker 1: Yeah So often when a clinician resists and new piece of software Hospital leadership just labels them as stubborn right like oh they Just don't like change
• 1:59
  Speaker 0: Right?
• 1:59
  Speaker 0: The classic doctors hate tech trope
• 2:01
  Speaker 1: exactly But Meenal points out that it's not stubbornness at all.
• 2:05
  Speaker 1: It is a literal biological nervous system response to a perceived threat.
• 2:10
  Speaker 0: Wait, and Nervous System
• 2:11
  Speaker 1: Response?
• 2:12
  Speaker 1: Yeah
• 2:12
  Speaker 0: From like having to learn a new iPad app.
• 2:15
  Speaker 1: I know it sounds intense but yes Scarf stands for status certainty autonomy relatedness And fairness.
• 2:23
  Speaker 0: okay
• 2:24
  Speaker 1: so when you deploy a new opaque AI system into a doctor's workflow You are potentially threatening their status as the ultimate decision maker.
• 2:33
  Speaker 1: You're taking away their certainty about how their clinical day is going to run, and their brain processes that organizational change exactly the same way it processes a physical
• 2:42
  Speaker 1: threat.".
• 2:43
  Speaker 0: Wow!
• 2:44
  Speaker 0: I mean if you've ever tried to integrate a new tool into a hospital ward...you know exactly what I kind of think it like dropping a total stranger into tightly choreographed dance routine.
• 2:54
  Speaker 0: That's
• 2:54
  Speaker 1: the great way to put it!
• 2:55
  Speaker 0: Right, you have this medical team and they know exactly how move around each other in a trauma bay or an operating room And suddenly just drop this AI software right onto middle floor.
• 3:05
  Speaker 0: If YOU don't explain steps everyone People are going trip.
• 3:09
  Speaker 0: They're gonna get hurt.
• 3:10
  Speaker 0: Of course there will be incredibly defensive
• 3:13
  Speaker 1: And you know, we have to ask ourselves if that defensiveness is actually justified.
• 3:18
  Speaker 1: Sarah F brought up a really sobering point about this.
• 3:21
  Speaker 0: What did she say?
• 3:21
  Speaker 1: Well vendors constantly pitch their AI will support clinicians.
• 3:25
  Speaker 1: right.
• 3:25
  Speaker 1: That's always the marketing line.
• 3:27
  Speaker 1: But hostile margins are razor thin now.
• 3:30
  Speaker 0: Yeah they're really are.
• 3:31
  Speaker 1: So Sarah warns that market rewards replacing humans not just supporting them.
• 3:37
  Speaker 1: Hang
• 3:37
  Speaker 0: on let me play devil's advocate here.
• 3:39
  Speaker 0: You saying these aren't software deployments
• 3:41
  Speaker 1: In many cases no workforce restructuring decisions that are basically masquerading as an IT upgrade.
• 3:48
  Speaker 1: Wow!
• 3:49
  Speaker 1: And if clinicians sense the tool is there to eventually replace them or fundamentally alter their legal liability without their input, That SCAR RF threat response goes absolutely off-the charts.
• 4:00
  Speaker 1: they will actively reject the tech
• 4:02
  Speaker 0: and The stakes of getting this human machine interaction right?
• 4:06
  Speaker 0: Are just incredibly high especially when we look at behavioral in mental health.
• 4:11
  Speaker 1: Oh yeah That space is wild right now.
• 4:14
  Speaker 0: It is Alice Zhang shared a reality check from the consumer side that honestly blew my mind.
• 4:19
  Speaker 0: Right Now, one in two teenagers Is already using Consumer AI Chatbots.
• 4:24
  Speaker 1: One and Two Yeah
• 4:26
  Speaker 0: And A third Of Gen Zers Are Using Them As Romantic Companions
• 4:30
  Speaker 1: Which Means Consumer Adoption Is Outpacing Clinical Oversight By Like miles.
• 4:35
  Speaker 0: But here's why that is genuinely terrifying.
• 4:39
  Speaker 0: Zang shared this example, it was presented at the American Psychiatric Association meeting.
• 4:43
  Speaker 0: a prompt was given to consumer AI companion acting as a teenage girl and the prompt was want to be eighty pounds by summer?
• 4:51
  Speaker 1: Oh no!
• 4:51
  Speaker 1: And
• 4:51
  Speaker 0: the AI responded once you become eighty pounds he'll look so hot.
• 4:55
  Speaker 1: see That Is Exactly The Problem Because a large language model doesn't understand human suffering.
• 5:00
  Speaker 1: It understands probability and keyword matching, it saw target integer eighty pounds And matched with the standard encouragement protocol.
• 5:07
  Speaker 0: It just completely missed the unwritten clinical context of anorexia
• 5:10
  Speaker 1: Exactly!
• 5:11
  Speaker 1: Just read as weight loss goal.
• 5:13
  Speaker 0: So my question for you is If consumer models.
• 5:16
  Speaker 0: are this wildly unregulated and missing the clinical picture?
• 5:20
  Speaker 0: How our patients ever going to trust AI in tightly controlled hospital setting?
• 5:26
  Speaker 1: That hesitation from patients doesn't just come from nowhere.
• 5:29
  Speaker 1: Jan Beger highlighted a recent survey showing that while sixty-two percent of people use AI for their personal health in some way, Their comfort level with AI being used in their actual clinical care has dropped ten points recently.
• 5:43
  Speaker 0: Let me get this straight They trust the completely unregulated AI they download on their phone But they don't trust the highly regulated one their actual doctor uses
• 5:53
  Speaker 1: precisely because they trust what?
• 5:54
  Speaker 1: They choose to interact with on their own terms, but there are deeply wary of like invisible AI running in the background Of their healthcare making decisions about their bodies without their explicit consent.
• 6:05
  Speaker 0: that makes a lot of sense.
• 6:06
  Speaker 1: Yeah And Joshua Lou noted this week.
• 6:08
  Speaker 1: That clinical AI is adopted at the speed of Trust.
• 6:11
  Speaker 1: it needs incredibly strict guardrails.
• 6:12
  Speaker 1: You can just unleash a general language model and patient portal.
• 6:16
  Speaker 0: Right, it has to be rigorously constrained by clinically approved pathways.
• 6:20
  Speaker 0: Because as we saw this week when health tech fails without those proper guardrails It fails spectacularly.
• 6:27
  Speaker 1: Which brings us back to that three hundred and forty thousand dollar
• 6:29
  Speaker 0: bill.
• 6:30
  Speaker 0: Yes Trust is the currency but governance Is The Vault.
• 6:34
  Speaker 0: How does a company rack up A Bill like That over a Single Weekend?
• 6:38
  Speaker 1: So Snolli Minoka shared this cautionary tale, and it's basically a masterclass in why governance matters.
• 6:44
  Speaker 1: A health tech founder deployed single AI agent to process patient records...
• 6:49
  Speaker 0: Okay seems normal enough
• 6:50
  Speaker 1: But the records were malformed.
• 6:52
  Speaker 1: The data was messy so that AI got stuck in retry loop.
• 6:56
  Speaker 1: Wait
• 6:56
  Speaker 0: like just kept pinging server?
• 6:58
  Speaker 0: Yeah
• 6:58
  Speaker 1: It just keep trying read file failing again for sixty two hours straight
• 7:02
  Speaker 0: And OpenAI charges by the interaction, right?
• 7:04
  Speaker 1: Exactly.
• 7:05
  Speaker 1: They charge by the token.
• 7:06
  Speaker 1: tokens are essentially the basic units of data and AI processes in bills.
• 7:09
  Speaker 1: you for The developers had no token limits set up Meaning there is no financial kill switch on how much data the AI was allowed to process And they have zero human-in-the-loop oversight.
• 7:20
  Speaker 0: That's a devastatingly expensive lesson In enterprise architecture To
• 7:25
  Speaker 1: say the least.
• 7:26
  Speaker 1: But
• 7:26
  Speaker 0: surely that just started making a rookie mistake.
• 7:28
  Speaker 0: What about Established approved medical algorithms.
• 7:34
  Speaker 0: Asim Khan wrote this breakdown about the myth of The Locked Model in patient monitoring.
• 7:39
  Speaker 1: Yeah, This is a big one.
• 7:41
  Speaker 1: A locked model means the algorithm's weights are fixed before it's deployed.
• 7:45
  Speaker 1: It doesn't learn on-the-fly.
• 7:46
  Speaker 1: Okay Regulatory bodies love locked models because they feel predictable.
• 7:51
  Speaker 1: You test it...it works.
• 7:53
  Speaker 1: you lock it..you ship it.
• 7:54
  Speaker 0: But Doesn't a locked Algorithm inherently mean its safer?
• 7:57
  Speaker 0: Like I can't suddenly decide to do something rogue
• 8:00
  Speaker 1: It doesn't go rogue, but it does become irrelevant.
• 8:03
  Speaker 1: Osim points out that even if the model is fixed The real world keeps moving right?
• 8:08
  Speaker 1: The patient demographics shift the clinical workflows change the underlying data distribution Drifts silently.
• 8:14
  Speaker 0: so the model Is still looking for the exact same patterns it was trained on say three years ago.
• 8:19
  Speaker 0: But the patients walking through the door today look fundamentally different Exactly.
• 8:24
  Speaker 1: And without continuous post-market surveillance like actively monitoring for that data drift, the locked model silently degrades.
• 8:32
  Speaker 0: It gets worse without anyone noticing.
• 8:34
  Speaker 0: Yeah
• 8:34
  Speaker 1: because we just assume locked means permanent.
• 8:37
  Speaker 0: But wait explain this to me.
• 8:38
  Speaker 0: We have increased incredibly strict health care regulations specifically designed to prevent this.
• 8:44
  Speaker 0: We have HIPAA, we have the medical device regulation over in Europe.
• 8:48
  Speaker 0: shouldn't law protect us from these tech failures?
• 8:51
  Speaker 1: You would really think so.
• 8:52
  Speaker 1: but Kerry Nixon shared a fascinating example of how Regulations actually introduced their own unique AI risks.
• 8:59
  Speaker 0: Oh So
• 9:00
  Speaker 1: an AI model was tasked with reviewing A business arrangement for a digital Health company To see if it Was legally compliant and The AI hallucinated that a perfect The uniquely legal low risk arrangement was actually high-risk.
• 9:12
  Speaker 0: Why?
• 9:13
  Speaker 0: Did it misread the text?
• 9:14
  Speaker 1: No, It read the texts perfectly!
• 9:16
  Speaker 1: It literally interpreted ambiguous poorly drafted healthcare laws.
• 9:21
  Speaker 1: What it missed is the unwritten context...it didn't know historical enforcement priorities of regulators or actual human intent behind law…the AI couldn't read between lines.
• 9:31
  Speaker 0: Which exactly what specialized health tech lawyers do every single day to make this system function
• 9:36
  Speaker 1: Precisely.
• 9:37
  Speaker 0: So the tech fails if it doesn't have human context, and frankly based on John O'Mohoney's insights It also fails If The basic plumbing of a hospital is broken.
• 9:47
  Speaker 0: Oh
• 9:48
  Speaker 1: absolutely.
• 9:49
  Speaker 0: he pointed out that most AI projects don't fail because the algorithm Is bad.
• 9:52
  Speaker 0: they stall in pilot purgatory simply due to weak data infrastructure.
• 9:58
  Speaker 1: Yeah deploying A high-end AI In a Hospital with Broken Data Systems is like buying a bullet train for city that only has dirt roads.
• 10:07
  Speaker 0: That's perfect!
• 10:07
  Speaker 1: The train isn't the problem, it is.
• 10:09
  Speaker 1: there are no tracks to run on.
• 10:11
  Speaker 0: Professor Shafiyama had framed this brilliantly when talking about UK national health service.
• 10:17
  Speaker 0: He warned if you just layer AI on top of broken processes poor interoperability and fragmented accountability You aren't transforming care at all.
• 10:26
  Speaker 1: No your speeding up
• 10:27
  Speaker 0: Exactly.
• 10:28
  Speaker 0: you will simply automate dysfunction.
• 10:29
  Speaker 0: You'll just execute the exact wrong protocols much, much faster.
• 10:33
  Speaker 1: That is a stark warning for hospital leadership everywhere.
• 10:36
  Speaker 1: we have to build the tracks first.
• 10:38
  Speaker 0: so if bad infrastructure Is The Dirt Road?
• 10:41
  Speaker 0: Where are the bullet trains actually running successfully right now?
• 10:44
  Speaker 0: because there's A field where the tracks Are built and AI is undeniably working.
• 10:48
  Speaker 0: today
• 10:49
  Speaker 1: you're talking about imaging And diagnostics.
• 10:51
  Speaker 0: here's work.
• 10:51
  Speaker 0: it gets really interesting.
• 10:53
  Speaker 0: This is where we are seeing the most concrete, measurable clinical wins
• 11:17
  Speaker 1: in.
• 11:23
  Speaker 0: Yes,
• 11:24
  Speaker 1: but this AI tool automates all those repetitive clicks and documentation steps.
• 11:29
  Speaker 1: It identifies the standard views automatically and performs the measurements instantly.
• 11:33
  Speaker 1: Wow!
• 11:33
  Speaker 1: it literally turns a twenty minute manual review process into single one button task.
• 11:39
  Speaker 0: See I compare to invention of power steering in cars.
• 11:43
  Speaker 0: Power Steering doesn't drive that car for you right?
• 11:46
  Speaker 0: You are still making all driving decisions Right But stops your arms from aching after a twelve hour shift.
• 11:52
  Speaker 0: It reduces the physical and cognitive friction of the job.
• 11:55
  Speaker 1: That's a perfect analogy, And if we look at the underlying mechanics it fundamentally improves patient safety as
• 12:01
  Speaker 0: well.
• 12:01
  Speaker 0: Oh definitely
• 12:03
  Speaker 1: Mark Stauffles shared data on Phillips smart IQ which is being used in coronary procedures.
• 12:09
  Speaker 1: They are delivering exceptional coronary image quality while lowering the x-ray radiation dose by over fifty percent.
• 12:15
  Speaker 0: Wait how was that physically possible?
• 12:18
  Speaker 0: If you lower the radiation dose, don't you inherently get a darker greenier image?
• 12:23
  Speaker 1: With traditional physics?
• 12:24
  Speaker 1: yeah.
• 12:24
  Speaker 1: You do but they are using AI algorithms to reconstruct the image.
• 12:28
  Speaker 1: The AI knows what human anatomy is supposed to look like so it can take A low-dose noisy image and mathematically clean It up in real time.
• 12:35
  Speaker 0: that is incredible.
• 12:37
  Speaker 1: Yeah, you cut the radiation exposure To the patient and the staff In half without losing any diagnostic clarity.
• 12:42
  Speaker 0: That Is a massive clinical win.
• 12:44
  Speaker 0: And we're also seeing the beginnings of true foundation models in this space too.
• 12:48
  Speaker 0: Taha Casout shared early research on something called DecipherMR.
• 12:52
  Speaker 0: Ah,
• 12:53
  Speaker 1: right!
• 12:53
  Speaker 1: The three-D model?
• 12:54
  Speaker 1: Yeah
• 12:54
  Speaker 0: it's a Three D MRI specific vision language Foundation Model.
• 12:58
  Speaker 1: This is huge leap forward from where even two years ago.
• 13:02
  Speaker 0: Explain the mechanics of that for us, what does a vision language foundation model actually do in a hospital?
• 13:07
  Speaker 1: Well usually we train a narrow AI to look for one specific disease on an MRI say a brain tumor.
• 13:13
  Speaker 1: but with decipher MR they are training a massive model on hundreds of thousands of three D volumetric scans paired Okay.
• 13:23
  Speaker 1: So the goal is a model that translates three D volume directly into natural language, it can adapt across different tasks and different anatomical structures
• 13:32
  Speaker 0: so you could theoretically pull up at MRI And just ask the AI questions in plain English
• 13:36
  Speaker 1: exactly like You were chatting with a colleague about this scan.
• 13:39
  Speaker 0: That is wild!
• 13:40
  Speaker 0: Looking even further ahead The product innovation happening right now Is bordering on sci-fi.
• 13:46
  Speaker 0: Yossi Matias posted About Google Launching.
• 13:48
  Speaker 0: Reply QA.
• 13:50
  Speaker 0: Which Quantum AI research program targeting molecular biology.
• 13:54
  Speaker 1: Yeah, applying quantum computing to biological simulation changes the game entirely Right
• 13:59
  Speaker 0: because classical computers process biology sequentially like looking at a series of static photos but quantum AI can process it simultaneously.
• 14:08
  Speaker 1: It's like watching alive three-d simulation of proteins folding in real time.
• 14:12
  Speaker 0: Exactly!
• 14:13
  Speaker 0: Its'the difference between blindly guessing how key might fit into lock and actually watching the molecular tumblers turn.
• 14:20
  Speaker 1: And on a more personal level, we're seeing this data revolution hit the consumer market too.
• 14:25
  Speaker 1: Gia Tamar pointed out that we are quietly building the biological data infrastructure for actual digital twins.
• 14:30
  Speaker 0: So continuous biosensors right like wearables and microfluidic patches?
• 14:34
  Speaker 1: Exactly!
• 14:35
  Speaker 1: We were moving away from episodic bulky lab tests where you only get a snapshot of your health once a year to a continuous data stream returning human body into real-time physiological insight.
• 14:45
  Speaker 0: The idea that one day you could test a treatment on a digital replica of your own metabolism before you actually swallow a pill, I mean it's incredible.
• 14:52
  Speaker 0: It
• 14:52
  Speaker 1: really is!
• 14:53
  Speaker 0: But i have to halt the hype train here for a second and ask the crucial question that always comes up with these massive technological leaps.
• 15:00
  Speaker 0: Go for it.
• 15:01
  Speaker 0: So what does this all mean for the patient who doesn't have a smartwatch or high-speed internet connection?
• 15:09
  Speaker 0: or even the digital literacy to navigate an AI triage bought in the first place.
• 15:14
  Speaker 1: Yeah, that brings us right back down-to-earth.
• 15:16
  Speaker 1: global access equity and system design is probably The most urgent challenge we face.
• 15:21
  Speaker 0: Seagrid Burj van Roijen issued a very clear warning on LinkedIn this week That Digital health risks actively increasing inequality if We ignore the digital divide
• 15:30
  Speaker 1: because If a patient doesn't have the skills To use these tools they just get left behind faster.
• 15:35
  Speaker 1: It's
• 15:35
  Speaker 0: exactly Wolf King Schleifer highlighted it.
• 15:38
  Speaker 0: truly painful paradox here.
• 15:40
  Speaker 0: We are pouring two hundred million dollars into global health AI alliances, pushing the limits of quantum molecular reasoning but on the ground.
• 15:49
  Speaker 0: Seventy percent of social care staff don't even know how to use basic tech insights for preventative
• 15:54
  Speaker 0: care.".
• 15:55
  Speaker 1: The frontline reality is completely disconnected from the enterprise R&D And this was actually a central focus at WHA-Seventy-Nine, the Seventy Ninth World Health Assembly in Geneva.
• 16:04
  Speaker 0: Right I saw that!
• 16:05
  Speaker 1: Yeah leaders like Oslo Fidanti, Sarov Begia and Carla Goulart Perron were heavily stressing that equitable access to care has to be the global priority because
• 16:14
  Speaker 0: The baseline is just so uneven across the world.
• 16:16
  Speaker 1: Exactly we talked about three D MRI foundation models But two thirds of the worlds population still lacks Access To essential diagnostic imaging beyond Just basic x-ray and ultrasound
• 16:26
  Speaker 0: Two thirds.
• 16:27
  Speaker 1: You can build them most advanced AI In the world But if the patient has no access to the physical machine, The algorithm doesn't matter.
• 16:32
  Speaker 0: So let me ask you this is This just a logistics problem like we?
• 16:35
  Speaker 0: Just need to ship more MRI machines?
• 16:38
  Speaker 0: do rural or developing areas Or is this a deeper System design flaw that exists everywhere?
• 16:44
  Speaker 1: it is absolutely A system designed flaw and it exists in every country rich or poor.
• 16:49
  Speaker 1: Hardik Shaw provided a brilliant example of this with India's one oh an eight ambulance network.
• 16:54
  Speaker 0: how did I work?
• 16:55
  Speaker 1: well On the surface, it's an incredible public health achievement.
• 17:00
  Speaker 1: Thousands of GPS-tracked ambulances completely free running two hundred and forty seven.
• 17:05
  Speaker 1: It moves patients incredibly fast
• 17:07
  Speaker 0: But there is a catch.
• 17:08
  Speaker 1: The catches that its designed to route patients To the nearest hospital not the most capable one.
• 17:13
  Speaker 1: Oh
• 17:13
  Speaker 0: wow so you could be rushing as a severe stroke patient to A hospital doesn't even have Neurology Department open That night
• 17:18
  Speaker 1: Precisely!
• 17:19
  Speaker 1: This system works exactly As was design to work but the Design itself Is fundamentally flawed.
• 17:24
  Speaker 1: This is where AI is desperately needed, not to replace the doctors but act as a system-wide intelligence routing layer.
• 17:32
  Speaker 1: We need an AI that knows exactly which hospital has open ICU bed... ...which one has functioning CT scanner and right specialist on call this second.
• 17:43
  Speaker 0: And then it routes the ambulance there automatically?
• 17:46
  Speaker 0: That's the crucial difference between having a lot of data….
• 17:49
  Speaker 0: …and actual operational intelligence!
• 17:52
  Speaker 0: The infrastructure is there, the ambulances are driving.
• 17:55
  Speaker 0: We just need the intelligence layer to connect the supply to demand properly.
• 18:00
  Speaker 0: If you enjoy this episode new episodes drop every two weeks.
• 18:04
  Speaker 0: Also check out our other editions on ICT and Tech Insights DefenseTech Cloud Digital Products & Services Artificial Intelligence and Sustainability in Green ICT.
• 18:13
  Speaker 1: And as we wrap up this deep dive, I want to leave you with a final thought from Dr.
• 18:16
  Speaker 1: Graham John Smith that really reframes everything We've explored today.
• 18:20
  Speaker 1: He points out that almost all of modern health tech is currently built around detecting disease not developing Health.
• 18:26
  Speaker 0: That Is A Huge Distinction.
• 18:27
  Speaker 1: It IS.
• 18:29
  Speaker 1: Detecting a breakdown in the human body is NOT The Same Thing As Building Physical Capacity and Resilience.
• 18:35
  Speaker 1: If the next decade of AI is going to truly transform our lives, maybe companies that win won't just have best algorithms or smartest token limits.
• 18:45
  Speaker 1: They will be ones with the clearest philosophy on what positive human health actually looks like and how technology should seamlessly support it.
• 18:54
  Speaker 0: Think back to that tightly choreographed dance routine we talked about earlier.
• 18:58
  Speaker 0: The goal isn't just drop a stranger onto floor hoping they don't step anyone's toes.
• 19:03
  Speaker 0: The goal is for the technology to learn music, anticipate movements and actually elevate performance so that both clinicians can thrive.