David M. Bennett MD (FAAOS, FPOSNA, DABOS)

Closeup of horse face

Closeup of Zebra face

Whether hoofbeats make you think of horses or zebras depends on what side of the Atlantic you sit.

#culturalhumility #MedEdSky #GlobalSurgery

"What? $3000 to put a PDF online?"

Dancing Boston Dynamics Robot Knows Its Revenge For This Will Be Sweet

Dancing Boston Dynamics Robot Knows Its Revenge For This Will Be Sweet theonion.com/dancing...

Things that make bones focally disappear in children:

Osteomyelitis
Infected fractures
Tumors
Syndromes NF-1 (related to tumors)
Hemophilia
Gorham-Stout
Scurvy
Severe malnutrition
Auto-immune diseases (RA/PA)
Allergic reactions (metal allergy)

Pre-op planning ↑ surgeon efficiency & outcomes: 94%-100% surgeons agree it’s vital, yet only ~50% practice it routinely. Analysis of fracture, tactics, logistics ↓ intra-op errors. Tools like templating, 3D imaging enhance precision. Lack of prep risks poor outcomes.

If this perspective finds purchase for you; take a look at one application of these ideas I am researching. We used AI to code a calculator which predicts pediatric remodeling 92% accurately, and then extracted the equation to validate it. pediatricorthopaedics.unicornplatform.page/graphremodel...

With that being a probabilistic likely outcome that materializes: I have chosen to learn the math, science, reasoning & figure out how those parts can make me better when I don’t have access to AI; and expect for it to be removed. Nobody has a “right to AI access” and it will surely be taken away.

More than world domination and extermination of all humans: the more predictable immediate AI danger is mass addiction then eventual withdrawal from public availability behind a massive paywall that only corporations, billionaires and governments can access.

The high probability of having AI removed means that we shouldn’t become reliant on it.

But coming to that conclusion for many involves a type of probabilistic thinking that they may not have developed without having AI in their life for a bit.

I think we need this anti-AI perspective. especially with the odds being good that we won’t always have this unprecedented access to it given corporation and authoritarian trends. Most of my research taps AI to extract nuggets that can be used even after AI goes away (if you learn the math).

Many of the most brilliant minds I know are fervently against AI and I love understanding their perspectives as well and consider them closely. This platform has such voices who I love @abraham.bsky.social @aptshadow.bsky.social are two and the Sci-fi writing community at large.

I also love reading the counter and anti-AI viewpoints - they make absolutely great points against it. One argument against AI that I don’t see often enough is that it can be taken away at any time. Most of my life didn’t have AI- excel drag and paste and Age of Empires battles was the best we had.

I wasn’t interested in probability theory or linear algebra at all. After all even now we know humans can extrapolate better from a more limited data set so taking time to learn, perform and apply complex math to my decisions seemed inefficient. Still is in some ways.

The difference between having AI be a huge learning opportunity boost & AI being an insular threat (for individuals) I think is mostly curiosity.

AI itself made me think harder because I am curious about everything, and it’s wildly interesting how math can probabilistically predict the future.

Making my post op instructions more appealing…

Congratulations to Phoenix Children’s Hospital Pediatric Orthopaedic Fellow of the year!

Formula for screw density pedicle selection in the absence of apical dystrophism or curves about 65 (I do higher density in the apical concavity in those cases)

In medicine, just as in life; doing the right thing isn’t enough. You must also avoid doing the wrong thing. One bad call can void years of good care. Every case is a chance to lose everything. Past success offers no protection. Every decision counts. Integrity is nonnegotiable. Just ask them:

14.7% of AIS pts w/ curves ≥70° had type-3 SC cords → ↑IONM alerts vs type-1/2 (40.0% vs 12.6%, OR 4.60, 95% CI 1.34–15.53). ∃↑risk w/ traction (p=0.002). Type-3 cords ⇒ 12.2% of cohort but ≈50% of alerts. SC morphology + traction = ↑risk ∴ preop SC type + traction plans must ↧IONM risk.

I extend to you the finest of morning greetings, with every expectation that the endeavors you face shall be executed with both finesse and exactitude. May the day ahead offer both the clarity needed for insight and the satisfaction of seeing your carefully considered objectives come to fruition.

Math→decisions. Regress→mean, outlier→avg; exp val ∑p·outcome guides risk; rare p<0.01 events matter; prosecutor fallacy P(E|H)≠P(H|E); corr≠caus; WWII missing bullets warn sel bias; Bayesian Δbelief key. Surgeons use exp val for implant risk, track unseen denom.

More on Repealing the Laws of Economics Check out this great listen on Audible.com. In his latest memo, Howard Marks discusses the implications of governmental intervention in economies. Considering examples such as rent control, fire insur...

www.audible.com/pd/B0FDKB9Y7...

Great paper this month in jBJS on how tenting probably shouldn’t be considered an absolute indication for clavicle fixation in kids.

Model obsolescence by user self-sufficiency is the goal here, not reliance on AI. Think of this as simplifying highly complex systems with more variables than we can normally navigate down to their most readily usable and independently verifiable format.

Or you could then design a controller and analyse how the entire system behaves

in many practical systems the process variable is modelled by some differential equation that you would derive based on what ever laws are relevant and from these equations you can use tools like the Laplace transform first to just analyse how the system behaves.