Co-lead / Director @Google AI x Economy Program. AP (on leave) @Wharton. Cofounder @workhelix. Everyone can just do stuff and that's {good, bad}.
Philadelphia, PA
- Tweets 3,256
- Following 2,362
- Followers 6,156
- Likes 16,530
really cool
What can Astra do when given a humanoid embodiment?
We built HomeBody to find out. Controlled by GPT Astra, it carries out long-horizon tasks in a previously unseen kitchen—from tidying up across the room to retrieving remembered objects from ambiguous requests—without environment-specific training data or additional policy learning.
Here's how we did it 👀: tml.stanford.edu/homebody/
4
625
Daniel Rock retweeted
Underappreciated fact in this chart’s discourse - is not just that it peaked at almost 10% of GDP, but it ran for a really long time… the area under the curve is what matters
1
4
869
new reason to keep AI from taking @ModeledBehavior's job:
We've gone through a cheesesteak industrial revolution recently, but I actually think the cheesesteak can be disrupted further. Mostly the crunchy seeded crust of the bread is what you want on the outside, you could slim that down to 90% crust. You really need very little of the white bread. Something in between a pretzel and naan perhaps? Jerusalem bagels come close too.
That then allows, indeed compels, you scale down the meat and cheese to keep the ratio with the bread constant. As a result you have a much slimmer sandwich. Dare I say, almost a wrap? For most people, the huge volume of a cheesesteak is an overkill, and lower costs would allow you to cut down on what has become a very high price point. Thus your customer base is expanded in two directions.
Mostly a bakery innovation is needed, the rest is simple.
Anyway, when AI takes my job this is what I'll be working on.
3
555
Daniel Rock retweeted
I have been fascinated by old science fiction ideas about using stars for computation since I was a kid.
But it was only about three years ago, during a convening on the future energy demands of AI, that I began to think about it seriously: what if we actually moved our machine learning hardware closer to its ultimate power source, the Sun?
5
13
92
6,068
Daniel Rock retweeted
The popular narrative of AGI often centers on a single, isolated super-intelligence. My colleagues James Manyika, @bratton, and I have a different model. In our new DeepMind Institute essay, we argue that the AGI transition will be a highly social event.
bit.ly/artificial-symbiotic-…
28
106
526
47,259
Daniel Rock retweeted
It was really great to chat with Ben, and this is quickly becoming one of my favorite pods to listen to. Thank you for having me on!
I could listen to @annastansbury all day, one of the best communicators and thinkers on worker power, inequality, and big macro labor questions. Check out our conversation on The Economics of Work podcast!
reveliolabs.com/podcasts/the…
5
24
5,693
Daniel Rock retweeted
➕➕ New research ➕➕
On the design of agents for markets and markets for agents.
AI agents are increasingly entering markets, acting on behalf of people—
How do we make sure they represent their people well, and that these markets are set up safely and fairly for everyone?
17
50
279
41,430
Daniel Rock retweeted
I know we’re focused on looming AI-driven extinction now, but can you spare a thought for what’s going on with labor demand? I have a new paper about that
2
24
78
8,767
Daniel Rock retweeted
Who owns the benefits of AI? In this essay I co-authored with @IasonGabriel for the DeepMind Institute, we offer four moral arguments — grounded in rights, reciprocity, fairness, and beneficence — that together make the philosophical case that the benefits of AI should reach everyone.
Read the piece here and let us know what you think: institute.deepmind.com/essay…
12
30
110
28,197
was lucky to be an early reader of this and it's awesome. buy this one to make @emollick finish the winds of AI winter.
I just got the first copies of my new book, Co-Existence (out October 20) & they look great!
Also, there is a fun pre-order bonus: if you pre-order, you get a code to an AI interview that will help you figure out how to use your human advantages with AI. co-existence.ai/
1
2
27
3,478
Daniel Rock retweeted
Nice to see AlphaFold and other bio ML models being used in Anthropic's novel enzyme system research.
Illustrates the complementarities between LLMs and specialized models that we described in our recent AI for science paper. www-cdn.anthropic.com/225736… ai.google/static/documents/A…
2
6
17
1,784
Daniel Rock retweeted
A 37-year-old man with a rare blood disease was being sent to hospice.
An AI model searched thousands of existing drugs and ranked a combination no one had tried for his condition.
He responded within a week. He is in remission.
The drugs were already on the shelf. The AI model found the match.
208
1,392
10,004
1,884,621
Daniel Rock retweeted
Super interesting!
"Does AI Assistance Enhance or Erode Expertise? Evidence from a Three-Month Field Experiment in Patent Drafting" by David Autor, Tanya Rodchenko, Josh Martin, Zanna Iscenko, Scott Strand, David Pearl, and Melissa Ferere.
"Whether AI assistance builds or erodes professional expertise is unsettled. In a pre-registered three-month randomized controlled trial, we gave 133 practicing patent lawyers at eleven U.S. intellectual property law firms access to a custom AI drafting assistant and measured both their performance while using AI and their professional judgment afterward without it. All work was scored by blinded expert patent attorneys. Paralleling findings from other white-collar domains, AI access raised the quality of work delivered on benchmark patent drafting tasks at 10 days (0.34 SD, p = 0.03) and 90 days (0.38 SD, p = 0.01), with larger gains among junior lawyers. After three months, all subjects redlined an existing patent application without AI, a core task of patent practice requiring expert judgment. Treated lawyers outperformed controls by 0.32 SD (p = 0.04), but this advantage was concentrated entirely among senior lawyers (0.45 SD, p = 0.02). Junior lawyers showed no average gain; their scores instead bifurcated, with sharply fewer mediocre scores offset by more poor and more good ones. The largest gains from AI thus accrued to the lawyers who retained the least. Foundational expertise may be a prerequisite for extracting durable skill from AI-assisted practice."
nber.org/papers/w35720
10
18
967
Daniel Rock retweeted
Muse bought me a car yesterday! 🤯
For the last few days, I had it send me a spreadsheet of top deals from every nearby car dealership.
Yesterday, an unusual deal popped up on a car I liked. It wrote up a bunch of negotiation text messages until I got an even better all-in price.
I ubered to the dealership, signed some papers, and drove off. Pretty wild!
4
2
20
2,829