XOutlier
Download .zipAdd to Chrome

Miles Brundage

@Miles_Brundage79k followers

AI policy researcher, @lfschiavo wife guy, Executive Director of @AVERIorg, Substacker, fan of animals and sci-fi, views my own

Open on X
Top outlier
56.2×
Best organic post
Peak views
621k
Single post
Bookmarks
7.5k
Across 100 organic posts
Last scan
Oct 17
Community cache
Posts 100 posts
#DatePost↓ OutlierViewsLikesRepostsBookmarks
1Jun 29, '24The New Yorker cannot keep getting away with this56.2×621k7.8k212358
2May 3, '25Not sure why the gutting of American science funding isn’t a bigger story. No one voted for it, it reduces American innovation and economic competitiveness in the near-term and lon40.1×443k4.4k904249
3Oct 23, '24I just sent this message to my colleagues, and elaborate on my decision and next steps in a blog post (see next tweet):29.8×329k1.2k100293
4Feb 7, '25Few remember this but this is what language models sounded like a few years ago like 10% of the time, before RLHF25.6×282k2.5k37171
5Nov 3, '25Kalshi's Twitter account spreads a lot of misinformation, such as this21×232k9902848
6Nov 18, '25There was never any evidence that pre-training had hit a wall and there continues to be no such evidence20.1×222k4583149
7May 29, '25Been doing some writing lately that touches on the pace of AI progress, and I arrived at this concise summary of my views: By the end of 2027, almost every economically valuable ta19.9×220k65678280
8Dec 20, '24I've been saying recently that completely superhuman AI math and coding by end of 2025 was plausible - 50/50 or so. Now I'd say it's much more likely than not (o3 is already better18×199k58261116
9Mar 31, '23Men will literally propose air strikes instead of engaging with existing debates about AI regulation16.8×186k4644128
10Jun 27, '25Not enough Ilya t-shirt appreciation on Twitter13.9×154k1.1k27118
11Oct 14, '25OpenAI has provided no evidence it has mitigated the mental health risks associated with its products other than announcing some advisors and reducing sycophancy from a high starti13.4×148k5246166
12Aug 7, '25Dear OpenAI friends: I'm sorry to be the one to tell you but I can say confidently now, as an ex-insider/current outsider, that the vagueposting is indeed off-putting to (most) out13.3×146k1.6k30142
13Sep 1, '25Yep we're on the same page, AI 2027 is indeed super unrealistic - the timeline are way off ᵇᵉᶜᵃᵘˢᵉ ᵗʰᶦⁿᵍˢ ʷᶦˡˡ ᵍᵒ ᶠᵃˢᵗᵉʳ ᵗʰᵃⁿ ᵗʰᵃᵗ12.9×143k56733128
14Jun 18, '23People laughed at Kurzweil but in retrospect his whole thing was more empirical + coherent than much current AI discourse, and he's basically right re: things getting progressively11.7×130k6545176
15Jan 19, '23Boston Dynamics videos basically haven’t changed much in the past few years. Comparing this to AI is pretty revealing.11.6×129k4761317
16Dec 20, '24In my new blog post, I share a starter pack for feeling the AGI: resources and arguments indicating that AI is virtually certain to exceed human performance in most areas in the ne11.4×126k43273216
17Dec 21, '24Sure, o3 shows that throwing thousands of dollars of compute can allow you to solve hard problems. But AI never gets cheaper over time, so this is the cheapest it'll ever be. And n11.3×124k1.2k37121
18Mar 13, '25Some people have asked me some variant of this question lately: "When will we get really dangerous AI capabilities that could cause a very serious incident (billions in damage / hu11.1×122k39760153
19Mar 6, '25One of the most distressing things I've learned since leaving OpenAI is how many people think something along the lines of: "Anthropic seems to care about safety - so Anthropic 'wi10.8×120k73042216
20Nov 20, '24Unsurprising that DeepSeek announced capabilities that are likely beyond what all but ~1-3 frontier developers in the US have. We're in the early stages of a new paradigm, which is10.8×120k42540136
21Aug 15, '25I was quoted here but only the first half of my quote was included and it gives the exactly opposite impression of my intent. In retrospect I should have asked to only be quoted in10.7×118k65522123
22Jan 24, '25Not bad (Operator making memes about itself)10.5×116k1.3k87147
23Sep 10, '19🧐 from:10.2×02.5k946154
24Dec 27, '24Glad to see OpenAI sharing their latest thinking re: corporate structure, and a lot of the explanation makes sense. However, there are some red flags here that need to be urgently 10×111k37248203
25Feb 4, '25Man goes to Deep Research, asks for help with the literature on trustworthy AI development. Deep Research says, "You are in luck. There is relevant paper by Brundage et al." Man: "9.1×101k1.9k102203
26Nov 10, '25View post on X9.0×99k5591040
27Sep 20, '25OpenAI alone is reportedly planning to spend ~20 billion next year on training alone, which is about as much as the entire Manhattan Project. Again, the Manhattan Project for AI an8.9×99k76766127
28Apr 9, '25In the long-run, this is far more important news than the stock bounce today. Few policy questions are as clear cut. If this continues, the US will essentially be forfeiting its AI8.8×97k5195584
29Nov 21, '24The real wall is an unwillingness to believe that human intelligence isn’t that hard to replicate and surpass8.4×92k4645669
30Aug 5, '25Is this "high-end laptop" in the room with us right now8.3×92k1.6k3358
31Jan 22, '25View post on X8.2×90k1.5k161209
32Oct 12, '25View post on X8.2×90k1.1k18
33Jan 24, '25Unfortunately this narrative won't die, and I'm extremely concerned that the Trump administration might believe it + be pressured by NVIDIA to believe it. To be clear, the US rever7.8×86k3573289
34Sep 1, '25TIL OpenAI sent a letter to Governor Newsom filled with misleading garbage about SB 53 and AI policy generally. Unsurprising if you follow this stuff, but worth noting for those wh7.8×86k4352496
35Oct 4, '24With very rare exceptions, the economics profession is completely discrediting itself by either ignoring or badly misunderstanding AI7.6×83k4013567
36May 16, '25Isn't this the second time they've blamed a rogue employee for changing the prompt?7.1×78k9934356
37Sep 12, '24View post on X7.0×77k7424685
38Oct 15, '24Some personal news:7.0×77k648015
39Sep 12, '24My team + many others across OpenAI have been thinking for a while about the implications of AIs that can think for a while. More to say later (and do read all the o1 blog posts, s6.3×70k51733163
40Dec 29, '24RL on chain of thought (i.e. the o1/o3 series of models from OpenAI and similar ones elsewhere) is relevant to non-math, non-coding problems. It’s nice to have an infinite source o6.2×68k43739309
41Oct 28, '25Update: this might be the first Thanksgiving in which brussels sprouts don't even briefly overtake deep learning5.8×64k1.2k4955
42May 7, '25Right or wrong, o3 outputs are never slop. These are artisanal, creative truths and falsehoods5.8×64k4391740
43Oct 31, '25I’m @beffjezos. We must accelerate5.8×64k480422
44Jan 21, '25If r1 being comparable to o1 surprised you, your mistake was forgetting the 1 part. This is the early stage of a new paradigm, and SOTA is the cheapest it will ever be. That does N5.7×63k3571877
45Aug 8, '25TFW you're one of the experts in a mixture-of-experts model and a query comes up that is relevant to your expertise5.3×59k1k5076
46Oct 13, '25Me: maybe I’ll take a break from AI for a few days United:5.2×58k529622
47Jan 5, '23Keeping up with AI progress like5.1×57k3455136
48Dec 20, '24The thing about all of the economics papers saying there will be like .01% or .3% etc. impact on GDP from AI in 10 years, is that they're just obviously false + embarrassingly so5.1×56k5353559
49Dec 19, '24Trying to imagine aspirin company CEOs signing an open letter saying "we're worried that aspirin might cause an infection that kills everyone on earth - not sure of the solution" a5.1×56k4844054
50Aug 1, '25Some of y’all need this5.0×56k1.2k7584
51Jun 23, '21For sale: neural network, never trained5.0×02k17231
52Nov 15, '24If you’re a journalist covering AI and think you need leaks in order to write interesting/important/click-getting stories, you are fundamentally misunderstanding what is going on. 5.0×55k4272987
54Jan 26, '25The DeepSeek iOS app does gather data btw. Says it right there4.6×51k3612851
55Jan 22, '25Stargate + related efforts could help the US stay ahead of China, but China will still have their own superintelligence(s) no more than a year later than the US, absent e.g. a war.4.6×51k4173285
56Nov 13, '24Betting against AI scaling continuing to yield big gains is a bad idea. Would recommend that anyone staking their career, reputation, money etc. on such a bet reconsider it.4.6×50k40053100
57Mar 18, '25The next phase of my career is starting to come into focus, and I'll be sharing occasional updates about it in the coming weeks/months. Today, I'm excited to share that I recently 4.5×49k6412347
58Nov 21, '20Lady Gaga as diagrams about AI systems (thread).4.4×01.3k272103
59Nov 18, '24I don’t understand the whole five hour podcast thing, from anyone’s time allocation perspective4.0×44k367514
60Nov 6, '24Relaxing by thinking about how prepared society is for AGI3.8×41k5743633
61Aug 22, '25I think people massively underestimate how much national security risk the US is creating for itself by prioritizing personal loyalty over competence in government hiring + retenti3.5×39k6199441
62Oct 30, '24Saying “X times smarter than Y” is a telltale sign of incoherent thinking about AI progress. Means nothing3.4×38k4041222
63Dec 20, '24I'm old enough to remember when getting double digit scores on FrontierMath was considered super hard I'm 6 weeks old3.4×38k7794737
64Aug 27It was an honor and a ton of fun to work on this unique collaboration between AVERI, Google DeepMind, OpenMined, and MLCommons. Getting AI governance right will involve both techni3.4×37k1261718
65Sep 7, '25Type of Bay Area guy who joins Meta Superintelligence Labs in order to donate his income to Anthropic3.3×37k604932
66Apr 29, '25*taps the sign*3.2×35k4813065
67Sep 28, '19Totally crazy that you can squeeze BERT down to 7 MB without much drop in performance - "Extreme Language Model Compression with Optimal Subwords and Shared Projections," Zhao et a3.0×0842209124
68Mar 28, '25View post on X3.0×33k4001040
69Feb 7, '22I'm old enough to remember when protein folding, text-based image generation, StarCraft play, 3+ player poker, and Winograd schemas were considered very difficult challenges for AI2.9×01.1k8364
70Dec 10, '24Bard times create strong Gemini2.9×32k6944120
71Jan 25, '25Freaking out about r1 is inversely proportional to familiarity with recent developments in AI. Where were y’all when QwQ was open sourced, or, ya know, when r1 itself launched last2.7×30k3891566
72Aug 26, '17View post on X2.6×07422550
73Nov 13, '25Don't worry about the cyber thing, state-sponsored hackers in China are all effective altruists who just pretend to hack banks etc. in order to help Anthropic with regulatory captu2.5×28k3671924
74Jul 22, '25The last thing you see before you realize your alignment strategy doesn’t work2.5×27k5503129
75Jan 20, '25"AGI isn't here literally this second" is the new standard for being an AGI "pessimist"2.4×26k4762531
76Nov 28, '17"Ms. Sabour [CapsuleNet paper lead author] is an Iranian researcher who wound up in Toronto after the United States government denied her a visa to study computer vision at the Uni2.4×06122600
77Mar 30, '18View post on X2.3×06772025
78Oct 6, '25View post on X2.3×25k4862351
79Aug 7, '18I'm super excited to share that I'll soon be joining the policy team at @OpenAI! I'll be working on a range of policy issues related to OAI's vital mission of "discovering and enac2.0×0832355
80Apr 22, '21Pretty wild Google put out a long paper on "Carbon Emissions and Large Neural Network Training", with a focus on LMs, and didn't cite @emilymbender / @timnitGebru et al. 2021, from2.0×06199053
81Jul 8, '21Excited to finally share a paper on what a huge chunk of OpenAI has been working on lately: building a series of code generation models and assessing their capabilities and societa1.8×04839593
82Apr 15, '25Reward hacking has begun, unfollowed1.8×20k40922
83Sep 10, '19StyleGAN trained on a few thousand real + drawn/generated images:1.8×053511829
84Apr 16, '21We've reached the stage of AI ubiquity where I'm just like "cool, makes sense" when seeing a deep learning paper published by researchers at Home Depot:1.7×05706353
85Mar 28, '21View post on X1.7×06169913
86Oct 24, '18"Superhuman Starcraft Play with Deep Reinforcement Learning":1.6×047210518
87Mar 7, '25View post on X1.6×18k4261730
88Jul 24, '19VictorianStyleGAN, trained on Pinterest photos.1.6×05428720
89Feb 12, '21View post on X1.6×05937412
90Feb 15, '25Fellas, is it woke to like laws1.5×17k4972112
91Dec 5, '20Hard to imagine a way that Google could have called more attention to this paper.1.4×05743210
92Dec 7, '22Good thing there is no further room for AI systems to improve, so we can take a breath and figure stuff out now—otherwise things would get weird1.2×04384826
93May 29, '22AI Twitter in 2025 is going to be like: “OK this book/movie/videogame it generated is decent but clearly it was cherrypicked. Show me all the bad outputs”1.2×04673415
94Feb 3, '18Old McDonald had a farm, AI, AI, oh...1.2×0380910
95Jun 29, '21Excited to share something GitHub, Microsoft, and OpenAI have been working on for a while--GitHub Copilot. Folks are going to be pretty blown away by it IMO. We've also put lots of1.2×03724836
96Aug 20, '21Professor: let's come up with a bite-sized first research project First year PhD student:1.1×04363711
97Feb 23, '21Everyone is getting into deep learning these days...1.1×04382618
98Aug 24, '17Find someone who loves you as much as this guy dislikes a hypothesis about Babylonian math:1.0×0342659
99Aug 3, '21CS people: "Excited to announce this $25 million AI center" Humanities people: "Excited to announce NSF reimbursed me for a cup of coffee 2 years ago"1.0×0413296
100Sep 3, '21Best $11 I have spent in recent history:1.0×03961818
101May 30, '21View post on X0.9×0384333

Scan another creator

@
@karpathy@levelsio@paulg@naval@marclou@illyism

Run this on any account.

XOutlier builds this table for every public profile you open on x.com, with live data and every post, not just the top 50.

Add to Chrome Download .zip