@[email protected]

Recovering politico. Minnesota transplant. Opinions are my own.

[John Oliver @iamjohnoliver]

Here's just a few thoughts ahead of tomorrow...https://t.co/YVcZGJoZdR

[Jon Favreau @jonfavs]

The most powerful message to young men in the entire campaign

[Ben Lesh @BenLesh]

Do you like debugging bad, undocumented code, with no tests, written by a developer that left the company?

Then LLM-built apps are for you!

[southpaw @nycsouthpaw]

Charlie Brown says Lucy wants to humiliate him by pulling the football away while he’s in the act of kicking it. Mostly false.

While Lucy has always pulled the football away every time the two have found themselves in this scenario, Lucy says she probably won’t do it this time.

[Tommy Vietor @TVietor08]

This graph from the NYT is important. Israeli hostage families blame Netanyahu for blocking a ceasefire deal, as do the Israeli negotiators, intelligence officials, the military, and broad swaths of the Israeli public. But the US keeps giving Bibi cover by only blaming Hamas.

[Star Tribune Pitch Bot @BothSidesDoItMN]

Opinion | If national republicans want the playbook for defeating Tim Walz, they should come talk to us.

By Scott Jensen & Matt Birk

[Haley Britzky @halbritz]

- May 2005, Tim Walz retires from the National Guard
- July 2005, his unit receives alert orders for deployment
- September 2005, unit goes to Camp Shelby to prepare for deployment
- March 2006, unit deploys

w/ @jeremyherb
https://t.co/IcEyKSoGPf

[Wedge LIVE!™ @WedgeLIVE]

In the final days of a lackluster campaign, the Don Samuels campaign tries to gin up excitement by putting his opponent's name on his yard sign.

[François Chollet @fchollet]

The question of whether LLMs can reason is, in many ways, the wrong question. The more interesting question is whether they are limited to memorization / interpolative retrieval, or whether they can adapt to novelty beyond what they know. (They can't, at least until you start doing active inference, or using them in a search loop, etc.)

There are two distinct things you can call "reasoning", and no benchmark aside from ARC-AGI makes any attempt to distinguish between the two.

First, there is memorizing & retrieving program templates to tackle known tasks, such as "solve ax+b=c" -- you probably memorized the "algorithm" for finding x when you were in school. LLMs *can* do this! In fact, this is *most* of what they do. However, they are notoriously bad at it, because their memorized programs are vector functions fitted to training data, that generalize via interpolation. This is a very suboptimal approach for representing any kind of discrete symbolic program. This is why LLMs on their own still struggle with digit addition, for instance -- they need to be trained on millions of examples of digit addition, but they only achieve ~70% accuracy on new numbers.

This way of doing "reasoning" is not fundamentally different from purely memorizing the answers to a set of questions (e.g. 3x+5=2, 2x+3=6, etc.) -- it's just a higher order version of the same. It's still memorization and retrieval -- applied to templates rather than pointwise answers.

The other way you can define reasoning is as the ability to *synthesize* new programs (from existing parts) in order to solve tasks you've never seen before. Like, solving ax+b=c without having ever learned to do it, while only knowing about addition, subtraction, multiplication and division. That's how you can adapt to novelty. LLMs *cannot* do this, at least not on their own. They can however be incorporated into a program search process capable of this kind of reasoning.

This second definition is by far the more valuable form of reasoning. This is the difference between the smart kids in the back of the class that aren't paying attention but ace tests by improvisation, and the studious kids that spend their time doing homework and get medium-good grades, but are actually complete idiots that can't deviate one bit from what they've memorized. Which one would you hire?

LLMs cannot do this because they are very much limited to retrieval of memorized programs. They're static program stores. However, can display some amount of adaptability, because not only are the stored programs capable of generalization via interpolation, the *program store itself* is interpolative: you can interpolate between programs, or otherwise "move around" in continuous program space. But this only yields local generalization, not any real ability to make sense of new situations.

This is why LLMs need to be trained on enormous amounts of data: the only way to make them somewhat useful is to expose them to a *dense sampling* of absolutely everything there is to know and everything there is to do. Humans don't work like this -- even the really dumb ones are still vastly more intelligent than LLMs, despite having far less knowledge.

[Jordan Schneider @jordanschneider]

what a classic VC move to buy Trump at the peak

[@[email protected] @danweinand]

@LindaIHiggins This is an ad from 2020

[Brian Rosenwald @brianros1]

This is an excellent point — the Framers envisioned people like Joe Biden holding the presidency. And they feared/loathed/recoiled at the idea of someone like Trump holding it.

[Tommy Vietor @TVietor08]

It's never too late to uninvite Bibi Netanyahu to Washington!

[Kamala Harris @KamalaHarris]

On behalf of the American people, I thank Joe Biden for his extraordinary leadership as President of the United States and for his decades of service to our country.

I am honored to have the President’s endorsement and my intention is to earn and win this nomination.

[Ben Rhodes @brhodes]

In a world increasingly filled with leaders who have changed laws, killed people, and stormed parliaments to cling to power, Joe Biden just flipped the script.

[The Daily Show @TheDailyShow]

Does Joe Biden realize that there’s no participation trophy in end-game democracy?

[Sarah Longwell @SarahLongwell25]

I slept on it, but still can’t believe this. Biden’s answer is breathtaking in its self-centeredness. This race isn’t about Biden giving it his all. It’s not about Biden period. It’s about beating Donald Trump. It’s about the future of liberal democracy. It’s about the country.

[@[email protected] @danweinand]

Time to donate to @AngieCraigMN

[Angie Craig @AngieCraigMN]

As an elected leader, I feel a responsibility to be honest about what I believe, even when it’s hard to hear.

President Biden is a good man & I appreciate his lifetime of service.

But I believe he should step aside for the next generation of leadership.

The stakes are too high.

[@[email protected] @danweinand]

This is silly. $150k is the cost of a single mid-level FTE engineer fully loaded. Building a new ride-share app would cost millions and take months to build. Expecting a start up to save us from Uber and Lyft leaving in a month is nonsensical.

[Kyle Stokes @kystokes]

INBOX: At a Minneapolis City Council committee meeting today, members will “consider allocating $150,000 for supplemental small business financing that could be used to support new and emerging rideshare companies.”

(Item isn’t currently on the agenda, but that can be changed)

[@[email protected] @danweinand]

@DeRushaJ The reality is that many cities tried to regulate them. But Uber in particular followed a strategy of using their huge amounts of capital to either just pay fines or fight the laws in court. They believed rightly it turned out that there was enough demand to force gov hands.