Matteo Lai

Biomedical engineer, PhD in AI for medical imaging. Building AI that holds up on real clinical data. Ex technical co-founder in digital health.

[Matteo Lai @Matteo_Lai3]

For 28 years I lived my life in stages. School, degree, PhD: each one was preparation for the next.

Then I started putting all that preparation to work, on something of my own.

Here I'll share what it's teaching me, about AI in medicine and about building.

[Matteo Lai @Matteo_Lai3]

@Berci The problem is that there are not enough checks about that, physicians are free to use patients’ data as they prefer, generally without consequences - interviewing physicians for my venture I found several…

[Matteo Lai @Matteo_Lai3]

@HealthcareAIGuy 40 years to build, and now every AI company wants what's inside😅

[Matteo Lai @Matteo_Lai3]

Source: https://t.co/0SNQI3W37T

[Matteo Lai @Matteo_Lai3]

A spreadsheet with 3,182 patients' data went into an unapproved AI tool.

AI is already in clinics: one doctor I interviewed pays for ChatGPT himself, two others I know use it daily.

Doctors, I'm curious: does your hospital give you an approved tool, or are you on your own?

[Matteo Lai @Matteo_Lai3]

@rickyhorwitz Probably we’ll be able to do all these stuffs also with ChatGPT, the point is just that we’d be giving for free our data to companies that does not respect users’ rights (at least very far from EU GDPR)

[Matteo Lai @Matteo_Lai3]

@w_milczynska We reached the same direction starting from the opposite sides :)

[Matteo Lai @Matteo_Lai3]

@joshfgray_ The problem arises when providers start thinking only about money without caring about patient data

[Matteo Lai @Matteo_Lai3]

@anuruddh_m Sadly, I think we’ll end up in a similar situation in Italy, where healthcare is public

[Matteo Lai @Matteo_Lai3]

@chrissyfarr Never too late :)

[Matteo Lai @Matteo_Lai3]

@anuruddh_m I think the point is that nowadays everyone can code, so it’s full of teenagers with projects/startups in tech, but thankfully for the moment healthtech projects are mainly limited to people with experience (at least for the moment…)

[Matteo Lai @Matteo_Lai3]

@EricTopol @Nature "Comparable to a single clinic visit" undersells it. The watch doesn't compete with the visit, it reaches people who never have one.
66% sensitivity vs 0% for the undiagnosed who never get measured

[Matteo Lai @Matteo_Lai3]

@HealthcareAIGuy Great story, and I believe it. But we only hear about the saves.
For every flag that caught a real risk, how many were wrong or ignored? Without that number, it's an anecdote

[Matteo Lai @Matteo_Lai3]

@_K_Stiles You cannot […], yet

Obviously a CEO in the SV speaks the language of SV inflating everything he says, but we need people like that to progress

[Matteo Lai @Matteo_Lai3]

@dieaud91 AI is a dangerous tool which could bring to a better society in the right hands - but we have to admit that not everyone has good intentions

[Matteo Lai @Matteo_Lai3]

@kevinace AI would bring to a better world thanks to people like you, but we have to admit that AI in the wrong person can be dangerous and actually hurt society

[Matteo Lai @Matteo_Lai3]

@manojpandey66 You’re partially right, but progress is not made by realistic people, and he’s only speaking the language of the Silicon Valley

[Matteo Lai @Matteo_Lai3]

@rintaro_ito Thanks for sharing!

[Matteo Lai @Matteo_Lai3]

@a_garciahermoso Doctors should recommend physical activity before prescribing drugs

[Matteo Lai @Matteo_Lai3]

@AI_4_Healthcare @bryan_johnson Every compute problem is a data problem first. Medicine is no exception.
The issue isn't compute vs biology. It's computing on records that miss part of the patient, and calling it prediction of health.

[Matteo Lai @Matteo_Lai3]

@w_milczynska @daisydwolf Garmin is very nice for that!