I've shared a virtual stage with Sir Mix-a-Lot, been selected to handle the follow-up tasting after a keynote by Matthew McConaughey for his Bourbon brand and even been selected as the alternate presenter for Jimmy Russell, but - Kermit the Frog?!
Listening to yesterday's KEXP's Drive Time for the second time today, and l have a new life goal: intro track by Kermit the Frog. I mean, how did Evie get that? 😲
So, yeah, I’m definitely presenting at a Bourbon industry conference, clearly. 🤷🏻♂️
My graphic designer, @hilarybaumann.com : “I have accomplished doom & gloom!”
Me, thinking about my presentation: awesome.
I would like to at least collect them into a smaller number of locations …
Well, that’s annoying. Just found two more media samples and a gift sample in the garage. I swear they’re like Pokémon around the house.
In my defense, I left two bottles of vermouth in the office. (As one does)
Yep, definitely can confirm.
Wasn't there, can confirm.
Full liqueur. Took me a minute to realize the abv is under 40%. It’s likely a very, very young whiskey a lot of sugar plus very artificial flavorings. If it was otherwise, that would be loudly proclaimed on the label.
Only three people in the audience actually ‘know’, though, and the rest are nodding along ‘knowingly’ in fear.
Current mood:
Turning off Memories has made a substantial improvement for me.
If a student submitted that to me as their reasoning and answer, I would believe they were thinking. I see nothing here which convinces me the models are using language to world-model and create outputs in any meaningfully way different than I do.
I do not believe that to be correct based upon what I have ready about how they work (mostly via TheZvi) and my own personal experience using them, but I'm open to changing my mind if you can provide evidence to support your claim.
I think you have not used current-gen AI with chain-of-thought reasoning models.
That analogy would mean the AI is doing creative writing, not lying.
This is where the confusion comes in regularly regarding AI. It is not programmed. AI is not 'designed' in the way you are using that word. Describing specific errors as "lying" is easy but wrong, which is why people who understand this better than you do write about it differently than you.
But the US is actually 4% of the world population, although middle- and upper-income white people are 15% of the US population …. oh.
If I’m understanding this correctly, privatizing Social Security would change it from group risk to individual risk, is that correct?
Making this a candlestick chart would make it even more obvious.
Now do the chart showing changes against disposable income.
Perhaps it’s because they don’t know they’re broligarchs - they think they’re average bros, just special.
What is stopping them from just requesting a 24 hour extension every 24 hours? It seems like the judges are extremely handicapped by this specific need to appear existentially unbiased.
I don’t want journalists to tell me two people are arguing about whether it’s raining and report on both sides in an equal and unbiased way. I want them to look out the damn window and tell me what’s actually happening and who is lying about it.
Thank you for putting succinctly what I have been trying to conceptualize.
I take credit for that cocktail. Dirty martini, quite good. Photo credit to Hilary.
Best explanation: bsky.app/profile/carl...
Okay, thank you. Previous comments were much lower indirect rates. This makes more sense.
I was just using that figure because it's a round number, not as a specific measure of a grant. I could have easily used 1MM or 10MM to achieve the same percentages.
