ankeshanand

8 ms·

1.
Gemini 2.5 Pro tops LiveBench, +6 pts overall over Claude 3.7 Sonnet Thinking(livebench.ai)
1 points by ankeshanand 1y ago|0 comments
2.
Gemini 1.5 Flash-8B(simonwillison.net)
2 points by ankeshanand 2y ago|0 comments
3.
by ankeshanand 2y ago
If you're an individual developer and not an enterprise, just go straight to Google AIStudio or GeminiAPI instead: https://aistudio.google.com/app/apikey . It's dead simple getting an API key and calling with
4.
by ankeshanand 3y ago
We've done extensive comparisons against GPT-4V for video inputs in our technical report: https://storage.googleapis.com/deepmind-media/gemini/gemini_... . Most notably, at 1FPS the GPT-4V API errors out aroun
5.
by ankeshanand 3y ago
Has anyone in this subthread actually read the papers and compared the benchmarks? LLama2 is behind PALM-2 on all major benchmarks, I mean they spell this out in the paper explicitly.
6.
by ankeshanand 4y ago
https://aipaygrad.es/
7.
by ankeshanand 5y ago
You can also rent a cloud TPU-v4 pod ( https://cloud.google.com/tpu ) which 4096 TPUv-4 chips with fast interconnect, amounting to around 1.1 exaflops of compute. It won't be cheap though (excess of 20M$/year I beli
8.
by ankeshanand 5y ago
It's important in the context that RL does not have performance ceilings.
9.
by ankeshanand 5y ago
Looks like any Github pages served with CloudFlare are getting blocked, I am trying out a fix.
10.
Reinforcement Learning as a fine-tuning paradigm(ankeshanand.com)
22 points by ankeshanand 5y ago|7 comments
11.
by ankeshanand 5y ago
Yep, Karpathy has mentioned this multiple times in their AI talks.
12.
by ankeshanand 6y ago
If you carefully curate who you follow, Twitter can be more like a bunch of subreddits, with the added signal of knowing who's posting. So it ends up a being great way to keep up with small communities.
13.
by ankeshanand 7y ago
Sorry if it wasn't clear, I do mention the linear classification protocol several times in the post. If you want to evaluate performance on a classification task, you have to show it labels during evaluation, otherwise it's an imp
14.
by ankeshanand 7y ago
I didn't mean to convey that we should abandon generative self-supervised methods, but I can see how comparing them gives that impression. Agree that using them in conjunction would make sense, since generative methods could capture so
15.
Contrastive Self-Supervised Learning(ankeshanand.com)
97 points by ankeshanand 7y ago|13 comments
16.
by ankeshanand 7y ago
https://twitter.com/karpathy/status/868178954032513024
17.
by ankeshanand 7y ago
Great piece, you might want to update the article with the mention of PyTorch Mobile that released today: https://pytorch.org/mobile/home/
18.
by ankeshanand 7y ago
There's active research in Model-Based RL right now that tries to tackle 1) and 2) together.
19.
by ankeshanand 7y ago
Bear has no Android app, and no LaTeX support either, so this works better for me.
20.
Sam Altman’s leap of faith(techcrunch.com)
161 points by ankeshanand 7y ago|162 comments
21.
by ankeshanand 8y ago
Neat idea, but isn't the $5k amount for 3 months kinda low?
22.
by ankeshanand 8y ago
One thing to note is that the camera viewpoint (it's position, roll, pitch, and yaw) is fed along with the images during training. Requiring access to this ground truth makes this method very constraining to use in practice.
23.
by ankeshanand 8y ago
"Our method still has many limitations when compared to more traditional computer vision techniques, and has currently only been trained to work on synthetic scenes. However, as new sources of data become available and advances are mad
24.
by ankeshanand 9y ago
The new version does use MCTS, you should read the paper again. :)
25.
by ankeshanand 9y ago
Twitter, Reddit, Facebook, HN, Twitch
26.
Adversarial Neural Cryptography in TensorFlow(github.com)
4 points by ankeshanand 10y ago|0 comments
27.
by ankeshanand 10y ago
Probably referring to the Master-Slave terminology
28.
by ankeshanand 10y ago
We could always use pre-trained word embeddings, the few observations won't matter then.
29.
by ankeshanand 10y ago
Bag of Words is not actually a great approach to understand text because it ignores the semantics of the word. For example, 'hotel' and 'motel' which are similar words have completely different vector representations in
30.
by ankeshanand 11y ago
Location: Kharagpur, India Remote: No Willing to relocate: Yes (San Francisco / Bay area preferably) Technologies: Python, Javascript, Django, Flask, MySQL, MongoDB, Redis, HTML, CSS, ReactJS (Full stack); scikit-learn, Pandas, Numpy,

More ›