lekanwang

As an investor in healthcare AI companies, I actually completely agree that there's a lot of bad implementations of AI in healthcare settings, and what practitioners call "alarm fatigue" as well as the feeling of loss of agency is a huge thing. I see a lot of healthcare orgs right now roll out some "AI" "solution" in isolation that raises one metric of interest, but fails to measure a bunch of other systemic measures.

Two thoughts: 1: I think the industry could take cues from aerospace and the human factors research that's drastically improved safety there -- autopilot and autoland systems in commercial airliners are treated as one part of a holistic system with the pilot and first officer and flight attendants in keeping the plane running smoothly. Too few healthcare AI systems are evaluated holistically.

2: Similarly, if you're going to roll out a system, either there's staff buy-in, or the equilibrium level of some kind of quality/outcomes/compliance measure should increase that justifies staff angst and loss of agency. Not all AI systems are bad. One "AI" company we invested in, Navina, is actually loved by physicians using them, but the team also spent a LOT of time doing UX research and feedback with actual users and the support team is always super responsive.

Pediatric allergy and atopic diseases in general is a mess an only slowly getting better. There's strong correlation between early eczema and food allergies, and now mild but convincing evidence it could even be causal. That eczema (and atopic disease in general) is strongly linked with microbiome and exposure to beneficial bacteria, especially in the first few days and weeks of life. This is also associated with malnutrition, diabetes, obesity, colic, and other symptoms we generally only treat symptomatically and in a silo. Yet for structural reasons, most pediatricians will at best tell you that early probiotics is a placebo. Top pediatricians in the know though will enthusiastically support targeted probiotics. Hell, the whole country of Bangledesh has a successful probiotic program -- https://www.science.org/doi/10.1126/scitranslmed.abk1107. It's amazing how much of common pediatric wisdom in the 80s/90s (clean newborn after birth, eat mushy prepared foods, enriched formula feeding, clean environments, avoid allergens, etc) are now seen as really harmful.

Freestyle rap should be the modern equivalent. Can judge on things like lyricism, flow/delivery, concept/cohesiveness, etc.

As a healthcare tech investor, I do see a lot of startups selling potentially dangerous AI systems into the healthcare system. That said, there are also a good number of companies that are implementing systems thoughtfully to address a number of issues that are very real in healthcare like staff burnout, continuing education, adherence to standard of care, managing complex value-based payment contracts and coordination of care, etc. The trouble I see is that clinician/hospital buyers of these systems can't always tell the difference. A basic initial filter that can be used is simply (a) does the team have an experienced medical professional with power on its executive team, and (b) does the team credibly know how to measure clinical quality impact of what they're building and do they have a plan to honestly measure it.

Metagenomic sequencing: The field exploded after technologies and techniques were developed for using next-gen sequencing to characterize entire populations/communities of living things, first with 16S rRNA sequences, then with full genomes. The cost to do this has also gone down many, many, many orders of magnitude in the last decade or two (just search "sequencing cost graph" on google).

Yep, you're right. Only meant to associate them in terms of why they're valued land, but yes, they're in separate riparian zones. Central valley is Sacramento/San Joaquin river delta.

The unfortunate thing is that large stretches of California desert (and much of the Central Valley to the north) used to be lake bottoms and have incredibly rich and fertile soils. Between that and a long growing season that's nearly cloudless, you get ideal growing conditions -- rich soil, lots of sun, low moisture (i.e. low disease load), and H2A labor -- provided you can control the irrigation. Ah the problems caused by mispriced externalities...

I like the Singaporean model of benchmarking high-ranking government salaries to a percent of the pay of top earners in the country (https://www.dollarsandsense.sg/heres-much-singapores-preside..., https://www.psd.gov.sg/docs/default-source/default-document-...), and pairing that with very active anticorruption enforcement.

MPs get paid over $1M USD under this model, but it incentives are much more aligned under this model.

Broadening out, this reminds me of one of the most eye-opening classes I've ever taken (and it was back in high-school!) -- Comparative Government/Politics: https://apcentral.collegeboard.org/media/pdf/ap-comparative-...

Was one of the few "see-the-matrix" moments I've had in my life.

Had a similar experience with GCP -- wanted to run an experiment in AWS, Azure, and GCP side-by-side. AWS and Azure were set up within a day. GCP required hitting up support to "turn on the feature," then a salesperson called to tried to upsell over a few calls before they'd turn it on. There was then confusing payment UI flow that meant my payment wasn't set up correctly. Overall took 3 weeks of back and forth to even start. This was 4 years ago, so maybe it's changed a bit now, but it's hard imagining depending on them as a business unless you're the scale of Snap and have leverage.

I'm finding with the teams I'm working with that the junior employees are the ones most impacted by not working from an office, but they often don't realize what they're missing -- the less formal forms of mentorship, stronger community, interacting with more people that's not on their team and in their role, overhearing context, the ability to have a 3-min quick chat with a senior person without a scheduled meeting, and developing that stronger sense of "what good looks like." I'm really concerned that we're going to have a two-tiered system where a bunch of people early in their careers are going to feel stuck in a few years and not even realize why.

You should -- protease inhibitors inhibit replication, but existing viruses should be more than enough to generate a durable response in most individuals.

Yeah, as parent mentioned in another reply below, ritonavir is an antiviral, but has no effect on SARS-CoV-2 (https://pubmed.ncbi.nlm.nih.gov/34048671/). Ritonavir is just used to inhibit CYP and therefore reducing clearance of the drug.

ACLU, Cato, EFF, and EPIC have all shown real success in legislative efforts and in court.

This is a decent state-by-state overview of the enforceability of non-competes: https://www.wonder.legal/us/guide/what-is-the-enforceability...

Reminds me of when a honeybee swarm occupied the parking lot behind our building and eventually mostly settled on a bike rack (this also happens to be the building of the downtown Palo Alto Philz). As a beekeeper, I was so excited that I emailed our head of ops and physical security in all caps asking if I could have them. After he called me to confirm that I wasn't trolling, he graciously offered to stand guard to keep folks away while I grabbed my bee suit to safely remove the bees and put them in a new home.

^ Yeah that's directionally accurate. The AMA is one of the two funders of LCME, the accrediting agency of medical schools for MDs (outside of the med school trade association itself). While they do not run medical schools, they have enormous power over training standards, who should be a physician, how medicine should be practiced, and who shouldn't practice. As an example, they lobby pretty regularly against the expansion of the role of RNs/NPs (https://www.ama-assn.org/press-center/ama-statements/ama-sta...). Much of this derives originally from the Flexner Report which created the current system of US medical education, which is still based on old sensibilities that physicians should be professional gentlemen and "proper" (and perhaps fueled by cocaine -- no seriously, google "halstead cocaine").

I am not a physician, but I have been in the guts of healthcare for quite a while, and the AMA continues to pop up as the man behind the curtain surprisingly often.

The AMA chooses who can be a doctor, who can train doctors, who can practice any kind of medicine, has extremely strong pricing power over the entire healthcare system via the RUC (https://en.wikipedia.org/wiki/Specialty_Society_Relative_Val...), advocates on behalf of doctors whether or not you agree, and much more, all under a guise of "non-profit advocacy for public health." They are a racket -- we would do well to have a rogue upstart competing licensing body.

I was fascinated with the history of the Sami when I visited Karasjok in the dead of winter, and spoke to a few who still tried to keep a somewhat traditional lifestyle. It did feel like it was a fight against time to keep those nomadic herding ways alive, and a really interesting philosophical debate about what really defines the Sami other than blood and (partly forgotten or purged) shared history if most are living lives indistinguishable from Norwegians and Swedes in their community.

You can't go faster than the speed of light in a vacuum. But, you can definitely go faster than light in a medium. Speed of light in water, for example, is around 0.75c.

I think you are perhaps conflating the public health definition of surveillance with what HN usually thinks of as "surveillance." (see https://en.wikipedia.org/wiki/Public_health_surveillance)

In the US, if you get certain reportable illnesses (COVID-19, E.coli infection, meningitis, etc), it's reported to the county/state health department, and via various channels, ends up anonymized with the CDC. Likewise, there are syndromic and lab-based public health surveillance networks that will monitor flu-like symptoms (ILINet) and will sequence samples (PulseNet) to help track cases in the US. Moreover, there's even more public health surveillance when it comes to drug safety -- if you tell your physician that you got any of a number of side effects from a drug you're taking, that is reportable to the FDA (FAERS). What Gottlieb et al are asking for is merely extending the best practices we have from flu, foodborne, and other diseases to COVID-19, in a way that has proven extremely effective, and also preserving privacy and liberty.

I think many of us here are rightfully concerned about an increasing surveillance state, but the term "surveillance" here is an unfortunate term collision, and we should resist the knee-jerk reaction in this case.

The WHO usually isn't in the business of sponsoring trials. The development of a single master protocol takes time, especially for one as large as this, and pulling one together within just a few weeks is already quite incredible, given the many, many partners that have to be involved, and how carefully you have to design it to balance a bunch of different factors.

I've worked on data related to healthcare and security (and sometimes both) for quite a while now, and I think there are a couple of general contextual themes, where, if present, means that you have to be extremely careful about applying "AI" (some kind of ML in most cases): (a) where there's a high cost for incorrect predictions (e.g. criminal recidivism, educational attainment, terrorist attacks, etc) (b) where causation is important (e.g. drug efficacy and safety, educational attainment, almost all of healthcare) (c) where you're in an adversarial domain (e.g. fraud, cybersecurity, security in general) (d) where high technical performance (precision/recall/F1/etc) isn't correlated with predictiveness of what you're actually looking for (much of healthcare)

In healthcare and security, there's starting to be an awareness of the snake-oil that's out there, but I still run into people regularly who ask for a magic algorithm that predicts patient outcomes or a security breach.

I think you may be referring to this publication: https://energyanalysis.lbl.gov/publications/co2-indoor-pollu...

Having worked closely with FDA, I've been very impressed by the people there and how thoughtful they are despite the pressure they're under and the almost complete lack of good software tooling due to the sensitivity, regulatory requirements, and complexity and specialization of the data. Most of the medical officers there could easily be a very successful (and much higher paid) private physician or a pharma researcher somewhere but choose to remain at FDA.

On one point people are making specifically -- unlinking safety with efficacy is really tough -- an immunooncology product that causes serious AEs/hospitalizations in a quarter of the population and costs $250k and only extends life by 6 months may be approved because the alternative is almost certain mortality; whereas another drug may not be approved simply because it increases bleeding risk in a small percentage of people because there are good alternatives already.

Yes, this is a very good point -- you're on the clock as soon as you file your patent, so sponsors have an incredibly strong incentive to take risks with trial design and use surrogate endpoints rather than direct endpoints for the primary to cut down on time. Then it's really an educated guess (albeit a highly educated guess) for your Phase III trial design to balance out your market vs chances of proving efficacy.

You'll need more patients, which is expensive.

But yes, you're actually right -- sponsors do run stratified phase 3 trials where subsets of patients are measured independently per pre-approved statistical analysis plan. The rub is that your patient population and your trial design needs to be sufficiently statistically powered to do that kind of analysis. The required number of patients quickly increases for the number of stratifications/subsets you're trying to prove an indication for, and patients are extremely expensive, so sponsors generally try to run the smallest trial possible while still being adequately statistically powered.

This is generally why in Phase IIs, sponsors will stratify by exploratory biomarkers, where there may not be sufficient statistical power per stratification, but can give them some idea for what to include in the IIIs.

I think there's false premise here that cancer has a single magic cure and we just need to focus on researching that instead of this expensive "personalized medicine."

Cancer is a complex beast, and what we term cancer is really a whole class of diseases with a plethora of underlying causes, some of which we have a basic understanding, and others which we are only starting to understand. Saying we are treating "cancer" as a whole makes as much sense as saying we're treating "fever" or "pain." So what's billed as personalized medicine in cancer at least is really an entirely correct attempt to dig a layer deeper to understand the actual genetic variations that lead the cells to overmultiply, then attempt to address those root causes, numerous as they may be.

We're currently still in the early stages of immunooncology, and the instruments are still blunt, but already more precise and informed than the chemo/radiation world.

As for the cost side, breakthrough R&D like this is not cheap (hundreds of millions to billions), and manufacturing of humanized antibodies (the drug itself) ain't cheap either. We somehow need mechanisms in place so biotech research continues to take these big risks. It is however a completely valid question to ask how we can better derisk this kind of research to the right degree so that certain gaps in the market (e.g. diseases with no cure) can be researched and filled. Big pharma fills that derisking role for the vast majority of smaller biotechs and research labs, often buying up IP and running the hyper-expensive Phase 3's with a 90% chance of failure. The FDA/EMA have the most power in derisking for big pharma, with their very difficult job of balancing the derisking and encouraging research vs safety.

Over time I'm sure the cost will decrease for immunooncology as we improve manufacturing and have get more data on their safety and efficacy profiles in the wild.

A lecturer at Stanford CS (Keith Schwarz) also has a good collection of various implementations of interesting algorithms: http://www.keithschwarz.com/interesting/

Cute idea. But as a beekeeper in the Bay Area, I have to agree with @tehchromic--there's some tricky stuff to consider to make this work:

- How will they prevent the bees from covering the internal mechanism with wax? Bees often cover plastic foundations with layer of wax as they build up the cells.

- More troubling, why won't the bees cover the mechanism with propolis (bee glue)? Bees tend to want to cover any crack with the stuff, which is really, really sticky stuff. Any time we've accidentally left an opening (that's not the entrance) somewhere in our hives, the bees have completely sealed it up with propolis, which is then extremely difficult to clean off.

- How will the honey flow freely? At bay area temperatures during Spring/early summer harvests, the temperature isn't high enough that the honey will flow freely when uncapped--it'd take on the order of at least 15 minutes, and more likely an hour or more if you want to do it cleanly, but leaving a jar of honey out for that long is going to attract all kinds of pests, and also bees, which kind of defeats the purpose of this.

- How do you clean this? As tehchromic mentions, this is going to have serious crystallization issues. During normal centrifuge or wax-scraping harvests, you end up having to wash all your equipment pretty thoroughly afterwards, or else the honey crystallizes and makes everything gunk up.

- Why would the bees uncap the harvested cells? I agree long-term that this would happen if you harvested a lot of honey, but I've rarely seen bees uncap the central stores of honey you're harvesting from, especially if you're careful about what you harvest to leave the bees enough for the winter.

- How will you deal with the additional pests and diseases? More nooks and crannies will encourage more pests to enter and roost. Having a spigot also encourages bad beekeeping, especially harvesting before the cells are capped. Honey is supersaturated sugars, which is why it doesn't go bad. But if you end up harvesting before it's capped, you're going to get nectar or sugar water that will actually go bad and additionally play host to a ton of pests and diseases.

Overall, I love that this is going viral--I honestly hope it attracts a bunch of new people to beekeeping, and I'm curious enough to support the Kickstarter and try one out, but I'm guessing those who stick with it will switch to normal Langstroth hives after a season or two.