Mike Lawrence

Scientist/statistician & fan of evidence-based/policy | he/him/Dr. | Aspiring ally. Privileged & biased & welcome suggestions for growth.

[Mike Lawrence @MikeLwrnc]

. @mmcintire given the suspicious behaviour surrounding the "study" your piece on The Daily covered, sounds like it might be worth the folks at @DataColada taking a look at the data

[Mike Lawrence @MikeLwrnc]

@DavidVognar8 @tegmark Are you under the misunderstanding that AI is currently created in a manner for which the term “programmed” is suitable?

[Max Tegmark @tegmark]

I found this a captivating, balanced and remarkably up-to-date book on the most important issue of our time: 🎅https://t.co/UJj9JLbL23

[Mike Lawrence @MikeLwrnc]

@DavidVognar8 @tegmark @kuki_ai Sentience is wholly orthogonal to risk.

[Mike Lawrence @MikeLwrnc]

@tamborrino_m Are these talks recorded and available afterwards anywhere?

[Mike Lawrence @MikeLwrnc]

From what I’ve gleaned so far from the approach taken to “prediction”, it seems concerningly naive of prediction uncertainty, decision theory considerations (attends costs of only misses and not false alarms), and the tricky nuances of causal modelling.

[Mike Lawrence @MikeLwrnc]

Any recommended papers on predicting harm risk akin to how some Canadian child protection services might use? I’m informally advising a friend who is serving as a non-stats expert in a class action related to birth alerts and similar “protective” state intervention.

[Mike Lawrence @MikeLwrnc]

@a_hartikainen @ajordannafa @mjskay @mcmc_stan N.b. The code there has expectations that diverge from the ID spec slightly (though I can’t recall now precisely how).

Also I saw big write bottlenecks with hdf5/netcdf4, so have been planning on switching to zarr (considering parquet too)

[Mike Lawrence @MikeLwrnc]

@a_hartikainen @ajordannafa @mjskay @mcmc_stan aria::class_score() is where to find reading code:

https://t.co/qDrYqigAvW

[Mike Lawrence @MikeLwrnc]

@YHZweatherguy Predictions from the latest ATCF model set, aggregated:

[Dropout @dropout]

Dropout's new password-sharing rule: you should share your password with a friend who can't afford Dropout, because it would be nice of you 🤗

[Mike Lawrence @MikeLwrnc]

Or do you actively want to discourage families from using your presumably-higher-margin digital games? This seems truly ill-considered.

[Mike Lawrence @MikeLwrnc]

@nintendo for being the more family-oriented platform, your actual support for family use is a joke. A multi-console house with a single buyer account should be expected. Limit simultaneous play of the same game sure, but forcing use of the buyer account on all but one console?!

[Mike Lawrence @MikeLwrnc]

If both a console and user are registered as part of my family, they should be able to play the games I purchased for the family using their own accounts and thus with their own distinct game saves.

[Mike Lawrence @MikeLwrnc]

@bbrsntd In the case of hierarchical models, there are ways to encode the hierarchy such that regions of the parameter space imply non-hierarchy, thereby implicitly representing at least the choice between hierarchical & non.

[Mike Lawrence @MikeLwrnc]

@bbrsntd More generally, I wonder if the feeling of structure implying framework here may reflect appreciation that choice of structure for a particular data context itself reflects use of prior knowledge that can be both important & difficult to account for properly. …

[Mike Lawrence @MikeLwrnc]

@EvpokPadding @bbrsntd Really? Fascinating. The fragility of MLE in the typically-high-dimension hierarchical models was my original motivation for converting to Bayes