Harshveer Singh @stringharsh

.. AI Researcher .. I truly believe that degrees of consciousness are hierarchical at all levels and originate due to superposition collapses. Views my own. llstringll.github.io Joined June 2018

Tweets

1K
Followers

55
Following

249
Likes

2K

Harshveer Singh @stringharsh

2 weeks ago

Today I discovered in experiment: By applying Cepstral-Mean-Variance Norm on inputs, the spectral convergence of a reconstruction loss behaves more like a RMSE of standardized shapes. *this is in terms of frequency bins, per uttterance. Although it's evident from the SC formula…

0 0 1 6 0

Harshveer Singh @stringharsh

3 weeks ago

I'm coining a new word. "Blabbergasted" Its when I'm flabbergasted by the amount of blabber some people spew out of their mouths, w/o saying anything valuable.

0 0 0 8 0

Andrej Karpathy @karpathy

a month ago

@lawrennd I almost had this happen yesterday, but my LLM suggested the right thing as `rm -rf '~'`. (with 'quotes'). Still very scary. Example of how this happens, e.g. if something like this sneaks into your .bashrc: export WANDB_DIR="~/.cache/wandb" The ~ isn't expanded inside "quotes"…

35 24 686 71K 127

Prime Intellect @PrimeIntellect

a month ago

Introducing the Environments Hub RL environments are the key bottleneck to the next wave of AI progress, but big labs are locking them down We built a community platform for crowdsourcing open environments, so anyone can contribute to open-source AGI

120 404 3K 1.4M 2K

Download Video

Harshveer Singh @stringharsh

a month ago

If @nsitharaman @nsitharamanoffc is gonna enter our emails and chats with the new bill they passed... Lets switch back to private IRC channels.

0 0 0 12 0

Elon Musk @elonmusk

a month ago

This may trigger some people

13K 50K 531K 40.4M 23K

Download Image

Sebastian Raschka @rasbt

a month ago

I’ve been working on something new: 📚 Build a Reasoning Model (From Scratch). The first chapters just went live! (The book will cover topics from inference-time scaling to reinforcement learning)

111 364 3K 155K 2K

Download Image

François Chollet @fchollet

a month ago

Saying that deep learning is "just a bunch of matrix multiplications" is about as informative as saying that computers are "just a bunch of transistors" or that a library is "just a lot of paper and ink." It's true, but the encoding substrate is the least important part here.…

134 321 3K 206K 620

pdawg @prathamgrv

a month ago

hot take: if you want to learn ML foundations, skip the tutorials. watch recorded lectures from top uni professors instead.

36 68 1K 56K 655

Santiago @svpino

a month ago

I attended a meeting last week where the customer told me point-blank: "We were waiting for the GPT-5 announcement before deciding whether to hire you. Well, here we are." GPT-5 was a massive shock to many people. Everyone is now realizing that AI is far from replacing software…

68 90 914 98K 152

Harshveer Singh @stringharsh

a month ago

All extrapolative learning is what we forgot about Love

0 0 0 15 0

Harshveer Singh @stringharsh

a month ago

We are agents that are internally driven towards mapping out new/uncharted loss landscapes.

0 0 0 15 0

Autistic Rambam Enjoyer @SwissBankingJew

2 months ago

I wrote a thing 👇 Today about ultra-low-latency applications, spooky concurrency bugs and the brilliant engineering of the LMAX Disruptor

1 1 8 467 1

Deedy @deedydas

2 months ago

Indians are making millions of dollars in the US and coming back home more than ever before. They're buying property 99% of salarymen in India couldn't afford in a lifetime all while keeping the US option open. It's hard to financially justify never leaving the country.

205 250 4K 435K 1K

Download Image

Ruben Veidt @RubenVeidt

2 months ago

@kawkoiharu from my limited retard knowledge ,it's can be true because we don't know about many parts of the cpu, or even some parts of the gpu, for some parts we only have apis to use, so we don't know any of the implementation details. one of main suspicion comes not from the main part of…

23 34 1K 139K 218

Harshveer Singh @stringharsh

2 months ago

So wholesome

Elliot @ElliotEvNo

2 months ago

So wholesome

45 29 988 244K 449

Download Image

0 0 1 37 0

BURKOV @burkov

2 months ago

First, these clowns waited for GPT-5. It was a flop. Now they wait for Gemini 3 as if it were something not transformer-based. Dear cardboard AI influencers: The circus is over. Now engineers will just do their work as they always did. You can switch your short attention span…

64 90 1K 68K 157

Harshveer Singh @stringharsh

2 months ago

Why doesn't @PMOIndia @nsitharaman and their government use the same "advanced ai tools" to detect electoral fraud, that they use to snoop our personal messages to track apparent tax fraud in every citizen's personal digital media.