Patrick Gladding
144 posts
Patrick Gladding
@patrickglad
Cardiologist with interest in DNA
New Zealand Katılım Ocak 2010
220 Takip Edilen53 Takipçiler
Trump is responsible for the rise of "frankly"?!
I had no idea.
Etymology Nerd@etymology_nerd
Thank you NYT, very cool
English
#ImageOfTheWeek by Anish Adhikari and Caleb Wutawunashe from Lenox Hill Hospital/Northwell Health, New York, New York, USA.
What is the preferred drug for this ventricular tachycardia? ow.ly/xntH50WF4Jp
#cardioed
English
English
Snap decision: if you only have $20 to spend:
English
@patrickc @GabriCorso Get life insurance companies involved. Here's an actuarial table for PCSK9i heart.bmj.com/content/104/20…
English
Something I've been thinking about: gene editing drugs (like Casgevy, Luxturna, Zolgensma) are a new paradigm in therapeutics, where it may be possible to cure many diseases with single administrations rather than managing them continuously via drugs that require ongoing use. Despite that promise, though, the CRISPR-focused biotech firms (CRSP, NTLA, EDIT, BEAM) have not been doing well.
There are a lot of problems with imperfect financial incentives in healthcare. Is this an important new one?
In particular: why would insurance companies pay enough for the benefits here? An average enrollee lasts only a few years in the US, so any given insurance company internalizes only a small portion of the benefit from a permanent cure. But if insurance companies (rationally) underfund, pharma companies will then (rationally) underinvest.
In theory, single-payer systems should pay more because of better incentive alignment, but in practice they don't have a tradition of paying enough for pharma innovation. (For the most part, they tend to ride on US coattails.)
So what do we do?
English
@bindureddy The Abacus version shows no sign of censoring, as suggested by other social media posts.
English
The R1 in the Deepseek app appears to be a distilled or an abbreviated version of some sort
The COT and the output tokens are much smaller compared to the ones from the full model
Maybe someone can verify this. If not we will run an objective eval in the morning and publish results
English
“But why would a superintelligence bother collaborating with a mere human?”
Ah yes: like why would God ever bother asking mortal man for help? Indeed. Well, I can’t refute those who believe in the supernatural except to say: magic isn’t real. Get back to me when we see miracles.
English
They think if jobs are made obsolete by AI that therefore people are made obsolete. But people are not their jobs. People change jobs all the time.
English
Also adding this because I think it’s relevant
Ethan J. Weiss@ethanjweiss
@DrDanMO I’ve been saying for a while that some drugs make doctors happy, but not patients (statins), some make patients happy, but not doctors (viagra), few make both doctors and patients happy, and even fewer of those do so despite costing $1000/mo and having real tolerability issues
English
The unique thing about GLP1RAs is that these drugs are not going away despite whatever hysteria is drummed up by skeptics . Unlike most drugs (eg, statins), people desperately want them, not because they will reduce the risk of serious chronic disease (they will)
English
Does it make sense to still use P<5e-8 as significance threshold in today's GWASs that include more and more rare variants (as the imputation quality continuously improve with large data and better reference panels)?
An interesting preprint uses real world data to investigate this question and finds that the traditional P<5e-8 yields 20-30% false positive rate. The authors recommend a stricter threshold of P<5e-9 instead.
Cheruiyot, Yang & McRae. bioRxiv
biorxiv.org/content/10.110…
English
@doctorveera This is more impressive pubmed.ncbi.nlm.nih.gov/34426707/
English
Controversial take: this is interesting, but still I am failing to see what is groundbreaking about it as many are praising it to be.
The inherent stability of blood counts of an individual within a range much narrower than the physiological range (determined by "healthy" population) is interesting. But beyond that the phenotypic correlations of blood counts with mortality and disease risks within the physiological range is something already known (I think).
The extra thing here is the blood counts are measured in terms of "setpoints" (a fancy term for the mean of multiple measurements per individual, excluding outliers) instead of the simple arithmetic mean of all measurements per individuals (which is what we typically do in the UK Biobank).
Likewise, the GWAS of blood count setpoints is almost the same as past GWASs of blood counts, just the phenotypes are a little more precise leading to slightly increased genetic signals.
Foy et al. Nature
nature.com/articles/s4158…
English
English
Insanely great. Protein folding AI helped @Quantum_Si develop the world’s first Next-Gen Protein sequencing $QSI.
And enabled @ProteinEvo to make the world’s fastest enzyme to turn old plastic into new & produce the first biologically recycled cloths with @StellaMcCartney.
elvis@omarsar0
We live in incredible times.
English
@KexinHuang5 Outstanding! Vinpocetine, a NP, ranks with MI and athero, linked by pollution and PM2.5 Can the evidence for this inference be interrogated further? Epinephrine ranks with MI. Causality vs association? Directionality of effect? Epi induces AMI, and used for Rx cardiogenic shock
English
We provided this visualization tool for TxGNN predictions and explanations across 17,080 diseases to everyone at txgnn.org! Try it out!
Also check out the supplementary for tons of robustness and ablation studies!
This work is the result of a multi-year collaboration with an incredible team: @payal_chandak @WangQianwenToo @_toolazyto_
@AkhilVaidMD @jure @girish_nadkarni @BenGlicksberg Nils Gehlenborg, and @marinkazitnik!
📄 Read more:
• Paper: nature.com/articles/s4159…
• GitHub: github.com/mims-harvard/T…
10/10
English
📢 Super excited to share our new study @NatureMedicine on developing and validating an explainable graph-based foundation model for drug repurposing, designed specially for rare diseases, which collectively affect 300 million patients globally!
🧵1/10
English
@ylecun Why not put data centres into space? It solves the thermal issue, and could be coupled with QC, needing near zero Kelvin. Cosmic rays etc, cost to launch and maintenance are issues. cnbc.com/2024/06/27/eur…
English
AI datacenters will be built next to energy production sites that can produce gigawatt-scale, low-cost, low-emission electricity continuously.
Basically, next to nuclear power plants.
The advantage is that there is no need for expensive and wasteful long-distance distribution infrastructure.
Note: yes, solar and wind are nice and all, but they require lots of land and massive-scale energy storage systems for when there is too little sun and/or wind. Neither simp,e nor cheap.
Danielle Fong 🔆@DanielleFong
i've been going on and on about this, but basically, AI flips electricity on its head. previously, nearly half the cost of electricity to the consumer was allocated towards distributing it where (and when) you want it. AI data centers want upwards of hundreds of megawatts of power in one place, most of the time. (and for training they don't care about latency) The most constrained thing isn't even the energy itself -- there's a surplus during the solar peak -- it's often the sheer availability & the interconnection. It now can take multiple years to get a grid interconnect to draw power of this magnitude, if the location can even handle it. The capital cost of the power infrastructure is just a tiny fraction (like 3%!) of the capex of the compute, and even just the depreciation of the compute exceeds the cost of even premium power. Hence, AI hyperscalers and those that aspire to be in their class are traveling to where the power is, are building where power has been (and there's legacy transmission to support it, like old nuclear plants) and are getting into the business of actually building powerplants and reactors. Utilities and transmission and distribution companies, interconnection queues, all are used to react much slower -- over many years -- unlike the top technology companies, now vying to compete at the highest levels of AI performance. Since ~99% of energy technologies previously died withering while waiting for utilities to consider them bankable, this represents an extremely fertile, attractive new state of play if you're bringing a new energy technology to market (💁🏻♀️🔆). Whereas before you'd have to brave what was once called a "green valley of death", now you have teracap companies like microsoft bidding on: - Conventional, large nuclear fission - Small modular nuclear reactors - Engineered Geothermal - Fusion! - Stirling Engines 😬 - Who knows what else - Maybe you, Anon! Even Tesla is standing up gas generator arrays to burn fossil fuels for their AI power supply. Shit is getting real
English
@curiouswavefn The Coming Plague, by Laurie Garrett, Pulitzer Prize-winning author.
English
I'm looking for a history of the Black Death that traces the disease through its major outbreaks, the discovery of the causative organism and successful efforts to find drugs against it. Easy to find books on the first part but not on all three. Recommendations?
English
@ylecun This is why Stephen Wolfram developed a Mathematica ChatGPT plugin. It's the same problem in biomedical AI. Impossible AI derived drug-like molecules don't conform to physical properties of matter. Physics-informed NNs are the way to go.
English
To people who claim that "thinking and reasoning require language", here is a problem:
Imagine standing at the North Pole of the Earth.
Walk in any direction, in a straight line, for 1 km.
Now turn 90 degrees to the left.
Walk for as long as it takes to pass your starting point.
Have you walked:
1. More than 2xPi km
2. Exactly 2xPi km
3. Less than 2xPi km
4. I never came close to my starting point.
Think about how you tried to answer this question and tell us whether it was based on language.
English
@AndrewJSauer Empagliflozin is USD$40/mth on patent in New Zealand. Generic would be ~$20/mth.
English
@demishassabis When will AlphaFold-latest be open sourced?
English
We have a long history of supporting responsible open source & science, which can drive rapid research progress, so we’re proud to release Gemma: a set of lightweight open models, best-in-class for their size, inspired by the same tech used for Gemini blog.google/technology/dev…
English
@bindureddy LLMs may be very clever but the fact that you cannot search your chat history is dumb.
English
A number of rumors of GPT-5 being released after Super Bowl!
OAI is still working on the new model and it won’t be released until Q2 at the latest
The next big model drop will be Llama-3 which is rumored to be very good 🔥
English
@Hragy Blade of a wind turbine, ironically passing a petrol station in Saudi Arabia
English
