Daniel Snider

PhD Student, U of Toronto, Scalable Machine Learning Systems

[Daniel Snider @DanielSnider]

@ubi_works Hell yeaaah Kim Pate absolutely slaying on the floor

[Daniel Snider @DanielSnider]

@LinusEkenstam @AzeAlter @musoscientific S

[Daniel Snider @DanielSnider]

@cafreeland Well done! Be bold. We need something like DOGE in Canada

[Floyd Marinescu 🔰 @floydmarinescu]

Many people including CEOs have come to me asking about Sam Altman's UBI pilot.

Unlike most critics, we actually read the research.

Basic income ended poverty with minimal impact on work beyond single parents. People took risks to find better jobs, retrain & escape abuse.

1/🧵

[Daniel Snider @DanielSnider]

@nisten Set a course for the eclipse! 🌙

[Waterfront Toronto @WaterfrontTO]

The new bridges in the Port Lands are now open! We’ve loved seeing your enthusiasm as you followed along. Check out these great photos sent in by Brian Tao as the bridges were getting ready to open.

[Chris Anderson @TEDchris]

<long post, alert> Incredibly important conversation going on right here. I know so many admirers of @elonmusk - including me, I'll admit it - who are perplexed and dismayed by his recent relentless drift toward people whose opinions he himself would have regarded as extreme a year ago.
@chuckcook has been a tireless champion for Tesla and its FSD roll-out. When you start losing people like him, it surely... surely... should be some kind of reality check. Elon argues that Chuck is ignoring the existential risk to American civilization. But I think the biggest risk to that civilization comes from unbridgeable political division, and Elon's latest provocations here are actually making the problem worse. There was a time, early in the Twitter/X takeover when Elon was arguing convincingly that the loudest voices where at the extremes, and he, along with 80%+ of America and the world did not share those opinions.

But now, to many of us, it seems that he's being pulled closer and closer to extreme opinion. In today's world, dismay and disgust are more powerful motivators than love and support. So this move can only be negative-sum in terms of Elon's overall influence. He is gaining a few new rabidly enthusiastic supporters. But losing far, far more. It's pretty heartbreaking to see.

I really think this is a moment when those who care most about Elon - his friends, his family, board members, anyone he might listen to - need to step up and urge him to rediscover the possibility that people like Chuck have his best interests at heart.

Elon, I happen to believe you could be on the verge of achieving astonishing things with X. In many ways the platform has advanced under your ownership (Community Notes my favorite example). It continues to innovate at incredible speed, despite - or because - of having only a quarter of the workforce it had. This is a genuinely incredible achievement and flies in the face of much of the snarking among progressives.

But I really think you're in danger of handing your enemies victory after all, just by making it harder and harder for anyone remotely in the center to listen to you. You said something brilliant when you first took over this platform. You said you wanted to maximize unregretted user-minutes. This is exactly the right metric.

But the tragedy right now is that for literally millions of people, the minutes they regret most when they come to X are... your posts. It's not just what they say, nor their aggressive tone. It's the fact they come from you, the powerful owner of the platform. To be clear, I'm not talking about all of your posts... there are still plenty featuring imaginative humor and exciting tech revelations. But too many come over as ignoring basic human values of respect and kindness. You won the admiration of countless millions around the world by opening up new possibilities for a genuinely exciting future. Many of those millions, I'm pretty certain, feel heartbroken and betrayed by your latest moves.

There is a way to articulate the beliefs you have about the risks of unfettered immigration, and the over-reach of wokeness, without needlessly antagonizing so many.

You own the playground. You cannot become the playground bully.

Please don't write this off, Elon. There's a better way where you can champion vigorous debate on this platform from all sides without letting yourself constantly get dragged into the mire.

Meanwhile... I'm with Chuck.

[nisten @nisten]

What almost everyone does not get about artificial intelligence is that this is more of a microbiomes-on-a-chip thing.

You are basically talking to yoghurt.

[Alex Glista 🇨🇦🇪🇺 @AlexanderGlista]

Instead of reinventing the wheel, Metrolinx is importing ETCS, or European Train Control System. This state-of-the-art signalling system is standard in the European Union, and will enable more frequency by allowing GO trains to run closer together.

[Daniel Snider @DanielSnider]

What going to see the Raptors in SF feels like #Raptors905 #WeTheNorth #RTZ #Raptors

[UBI Works 🇨🇦 @ubi_works]

BREAKING: President Obama joins the call for UBI

[rohan anil @_arohan_]

First of all, massive congratulations are in order to @zacharynado @GeorgeEDahl @naman33k and co-authors on this massive work spanning multiple
years on benchmarking neural network training algorithms! 🎉🍾

I have a horse 🐴 in the race and its called distributed shampoo 🦄

[NorfolkPlantSciences @BigPurpleTomato]

Imagine these two delicious combo in your dinner plate.

Left is golden rice enriched with beta carotene 🌾. Right is purple tomato enriched with anthocyanin, all products of agricultural biotechnology.

We love to see science improve lives!🍛😍

[H⊶a⊶c⊶k⊶L⊶a⊶b⊶T⊶o @hacklabto]

ANNIVERSARY ALERT!

Already 15 incredible years of hacking and tinkering have passed! Come celebrate with us Hacklab's 15th anniversary on Saturday, August 19th.

Food, music, and a lot of fun are guaranteed! 🪩

Find more below ⬇️
https://t.co/ATFKrnpoNl

[Shalev @Shalev_lif]

Exciting advancement by Google DeepMind with the newly introduced Robotic Transformer 2 (RT-2)!

Main idea: train a single agent on 1) web-scale text/image data and 2) robot data, to create a single Vision-Language-Action (VLA) model. This VLA model can directly output actions to control a robot, using knowledge learned from the web!

The elegance of the approach is compelling: represent actions as language-like tokens.

Key result: RT-2 can solve robot tasks that require knowledge transfer from the web-data. For example, to "move banana to the sum of two plus one", RT-2 has to use its object manipulation skills (learned from the robot dataset) and it has to know the answer to "two plus one" (learned from the web-based dataset).

These results support recent trends, namely that 1) scale allows for more elegant solutions, and 2) diverse, multi-modal data improves performance.

Amazing work from the Google DeepMind team! Excited to dive even deeper into the work after our poster session for STEVE-1 at the SPIGM workshop today.

https://t.co/4UfSJTRfkF

[Floyd Marinescu 🔰 @floydmarinescu]

Tax buildings, you discourage development.
Tax paycheques, you discourage work.
Tax capital, you discourage investments.

Tax land... you encourage better use.
One of these is not like the others.

[Elon Musk @elonmusk]

Many have offered rich incentives for X (fka Twitter) to move its HQ out of San Francisco.

Moreover, the city is in a doom spiral with one company after another left or leaving.

Therefore, they expect X will move too.

We will not.

You only know who your real friends are when the chips are down.

San Francisco, beautiful San Francisco, though others forsake you, we will always be your friend.

[Daniel Snider @DanielSnider]

@zacharynado This has been on my mind since you tweeted. Why not have LLMs make database queries? Latency is an interesting concern. An LLM might have less latency variation as compared to an unknown number of database round trip times added up.

[Daniel Snider @DanielSnider]

@IsochronicMike I loved your talk at MLSys'23 on neuromorphic computing. It got me excited about the bright future of neuromorphic systems!