Warren W. Pettine

Academic/founder. As much skiing as humanly possible.

[Warren W. Pettine @WarrenWPettine]

@Dan_Jeffries1 How do we pay to develop/maintain open?

[Warren W. Pettine @WarrenWPettine]

@axios Missing from this is the role funding from orgs like the #NSF needs to play in developing/promoting US open-weight models. For decades, they've been a pillar of open source. It's hard for a company to pay for a common good with its own revenue. See META and Llama models.

[Warren W. Pettine @WarrenWPettine]

https://t.co/xeGcnWEPd7

[Warren W. Pettine @WarrenWPettine]

Open-weight Chinese models really have become solid components in our stack. For a number of technical reasons (e.g., measures of model confidence), open-weights are essential.

Missing is the critical role of grant funding. Hard to open-source when revenue is the focus. #NSF

[Warren W. Pettine @WarrenWPettine]

@morgancheatham How does that address ontological specificity? For example, if you have trouble consistently diagnosing patients with ADHD versus ASD versus OCD, then how can one train a model that is based on that classification?

[Warren W. Pettine @WarrenWPettine]

@ben_mathes @APEXropeInd Also, it's hilarious I triggered an autopost from APEX Rope Industries on this...

[Warren W. Pettine @WarrenWPettine]

@ben_mathes It might be time to put the console away for the day...

[Warren W. Pettine @WarrenWPettine]

@ben_mathes Yeah, sorry, was parsing through the potential cause. We've been talking about that on the team, particularly in the context of pricing.

[David Sussillo @SussilloDavid]

Honored to share my story of growing up in group homes, and my work in neuroscience, on Armchair Expert!!
@ArmchairExpPod

[Warren W. Pettine @WarrenWPettine]

@ben_mathes @paulg @SiliconHBO And here I am like 80% Ed Chambers.

[Warren W. Pettine @WarrenWPettine]

@ben_mathes @paulg @SiliconHBO Huh, so Gilfoyle really was based on you. I always just thought your cat-eye contacts were a coincidence.

[🅱🅔🅝 @ben_mathes]

Did this over the weekend 💍.

[Brian Locke, MD MSCI @doc_BLocke]

Our review summarizing the epidemiological evidence of OSAs impact on the occurence of hypercapnic respiratory failure is now freely available on PMC

https://t.co/Z5dIvpj5bv

[Scientific Computing & Imaging Institute @uusci]

👏 Congrats to SCI Director @parasharmanish for receiving ACM's Distinguished Service Award!

Read more on SCI's site: https://t.co/KfE5grjS9A

[Brian Locke, MD MSCI @doc_BLocke]

Why are RCTs indispensable for medical science, but rare in sciences like physics or chemistry?

A great paper by Cartwright and Deaton argues that RCTs power comes from the fact that a skeptical reader needs to accept few theoretical assumptions to be convinced by a result.

If you really, correctly understood the mechanisms by which disease causes harm, you could learn more by cleverly probing your theories with active patient allocation.

In physics, we have robust, predictive theories. In medicine, we do not - physiologic understanding is rudimentary at best. Hence, RCTs.

I worry that by moving from simple, RCT to complex Bayesian designs, we lose a bit of that simplicity.. and skeptics are likely to be a bit less persuaded by results.

We talk about this in the context of why I predict REMAP-CAP's finding of harm in steroids for CAP will be mostly ignored on Last Week in Medicine's last episode

[Farrukh @implausibleblog]

Mark Carney, "The lessons of Brexit are beginning to be applied to the United States"

"When you break off or substantially rupture trade relationships with your major trading partners, Canada included - the most important trading partner of the United States"

"You end up with slower growth, higher inflation, higher interest rates, volatility, weaker currency, a weaker economy"

"We're seeing the early stages of that in the United States"

[Kenway Louie @kenwaylouie]

Not all noise is the same!

Excited to be a part of this fantastic new work from @BoShenNeuro using modeling and choice experiments to show differential effects of early versus late noise on decision making.
https://t.co/AJo49IiMRS