Daniel (@DanielTamart) header
Daniel (@DanielTamart) profile picture
Joined November 2018
Photo by Daniel (@DanielTamart)Photo by Daniel (@DanielTamart)Photo by Daniel (@DanielTamart)Photo by Daniel (@DanielTamart)Photo by Daniel (@DanielTamart)Photo by Daniel (@DanielTamart)Photo by Daniel (@DanielTamart)Photo by Daniel (@DanielTamart)
FC Augsburg believes it's quite terrible to have 17 other Bundesliga clubs sitting on a bunch of the league's limited supply of points.
OpenAI CEO Sam Altman believes people owning RAM is wasteful, and that it belongs in data centers instead. "From a global allocation perspective, actually, it’s quite terrible to have everyone have a bunch of the world’s limited supply of DRAM sitting in their own computer."
280
8,104
252
126,308
2,244,657
Daniel retweeted
incredible timeline where we let a French (rightfully) school us on not surrendering
Should Europe leave frontier AI to the Americans and the Chinese? The idea has gained ground in Europe. It is a fundamental mistake. Build or surrender: for now, the choice is still ours. Full article below.
4
3
1
80
3,421
Apparently this is where the AI era started 😅 If I had beaten that bot in 2017, maybe the world would be a very different place now I had a chance to stop AI, but …
107
744
121
8,003
350,842
Cost is a bad thing It should be down oh no... 💀💸
38
44
5
1,247
199,925
『球体タイヤ: 全ての方向に移動できるタイヤ』 コンセプトや試作機(デモ)の動画は見かけるが実用化は如何に…。 youtu.be/Tr83-rCwG4k #BallTire #SphericalWheel #SphereTire #omnidirectional #concept #design #mobility #futuretech
14
1
49
2,946
In 2016 I set out to solve Brood War with self-play RL from scratch. Today I believe that this quest has been completed youtube.com/watch?v=Lfo4C0xz… (by a 300M params model).
10
12
4
113
29,697
Daniel retweeted
On July 25, we hacked OpenAI. Two bugs let us take over ChatGPT/Codex accounts of OpenAI employees (+some unaffiliated users) and reach connected services: Outlook, Slack, GitHub, etc. We proved it with a PR in OpenAI’s internal codebase . It took us <72h. 🧵
353
1,381
548
11,781
2,826,845
Daniel retweeted
This result is amazing, but not for the reasons people think. It has very little to do with puzzle solving since it's a linear game and you just have to strictly follow any of the detailed walkthroughs. The impressive parts are efficient 3D navigation, spatial reasoning, combining instructions from multiple sources, micro adjustments after failed attempts and not getting stuck. The cherry on top is the relatively low cost (167$).
And... GPT-6 Astra has autonomously completed Portal! I didn’t expect this to happen so soon, but I’m glad we've made so much progress here. I was reminded that back in 2016, one of OpenAI’s technical goals was to “solve a wide variety of games using a single agent.”
16
20
1
424
29,306
最近たまに見かける「インセプション風の曲面マップ」、ナビアプリに実装されたら手前と遠くが同時に見渡せて意外と便利そう。 orbify.eu/demo/
129
612
249
5,410
1,454,247
Fukushima's video (1986) shows a CNN that recognises handwritten digits [3], three years before LeCun's video (1989). CNN timeline taken from [5]: ★ 1969: Kunihiko Fukushima published rectified linear units or ReLUs [1] which are now extensively used in CNNs. ★ 1979: Fukushima published the basic CNN architecture with convolution layers and downsampling layers [2]. He called it neocognitron. It was trained by unsupervised learning rules. Compute was 100 times more expensive than in 1989, and a billion times more expensive than today. ★ 1986: Fukushima's video on recognising hand-written digits [3]. ★ 1988: Wei Zhang et al had the first "modern" 2-dimensional CNN trained by backpropagation, and also applied it to character recognition [4]. Compute was about 10 million times more expensive than today. ★ 1989-: later work by others [5]. REFERENCES (more in [5]) [1] K. Fukushima (1969). Visual feature extraction by a multilayered network of analog threshold elements. IEEE Transactions on Systems Science and Cybernetics. 5 (4): 322-333. This work introduced rectified linear units or ReLUs, now widely used in CNNs and other neural nets. [2] K. Fukushima (1979). Neural network model for a mechanism of pattern recognition unaffected by shift in position—Neocognitron. Trans. IECE, vol. J62-A, no. 10, pp. 658-665, 1979. The first deep convolutional neural network architecture, with alternating convolutional layers and downsampling layers. In Japanese. English version: 1980. [3] Movie produced by K. Fukushima, S. Miyake and T. Ito (NHK Science and Technical Research Laboratories), in 1986. YouTube: youtube.com/watch?v=oVYCjL54… [4] W. Zhang, J. Tanida, K. Itoh, Y. Ichioka. Shift-invariant pattern recognition neural network and its optical architecture. Proc. Annual Conference of the Japan Society of Applied Physics, 1988. First "modern" backpropagation-trained 2-dimensional CNN, applied to character recognition. [5] J. Schmidhuber (AI Blog, 2025). Who invented convolutional neural networks? xitter.cf/SchmidhuberAI/status/1…
Yann LeCun’s 1989 convolutional neural network demo, the foundation for the CNNs we still use today. It's amazing how far we've come since then!
72
513
86
3,255
749,998
Daniel retweeted
guys you do know you can just disable thinking, and instead give it a "deep_think" tool, and it will call it with internal CoT reasoning format right? gl fixing that
We can finally talk about it: We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company. We verified that our reasoning token count matches billed API thinking tokens 1:1 for most of the prompts we queried.
154
449
139
6,377
957,779
Daniel retweeted
Somehow, the US Government in both 1926 and 2026 -- "We have to shut down illegal distillation of the most powerful spirits."
19
231
12
2,720
98,419
Daniel retweeted
3D Gaussian Splattingの新しい見せ方を作りました。その名も Inversphere. 2Dのリトルプラネットと違い、空間ごと球面反転しています。ガラス玉に入れる従来の見せ方は球で「切り取る」のに対して、これは球の中へ「折り込む」ので360度の景色が無限遠まで立体で入っています。 #3DGS
142
910
91
6,623
533,765
can’t be done
NEW: The White House is reportedly demanding Anthropic make Fable 5 impossible to jailbreak before rerelease — which security experts say “can’t be done.”
287
253
53
5,004
289,503
Daniel retweeted
noam shazeer is kylian mbappe for people who import pytorch
8
56
5
944
52,799
Back in the day, there were so many cell phone options, it was rare to see someone with the same one as you.
25
110
13
945
26,906
A day in Manhattan, warped so that distance ≈ drive time. Drive times swell in the morning and evening rush hour, then contract in the night.
119
523
216
8,454
2,094,020
After AlphaGo, the skill of human Go players noticeably improved. I suspect we will see a similar pattern in math.
Another major problem, this time in additive combinatorics, has fallen, this time to humans rather than AI, but using methods related to the AI solution to the unit distance conjecture.
182
938
205
8,910
823,570
Codex grew programmatic policies with no neural nets: max score on Breakout, and SOTA-level scores on MuJoCo. Maybe heuristics were not too weak. Maybe they were just too expensive to maintain. Maybe it's the next paradigm. trinkle23897.github.io/learn…
67
233
72
1,448
3,222,839