alecco

41 ms·

781.
Lutnick/Fischer want to stop Meta publishing Llama(twitter.com)
1 points by alecco 2y ago|0 comments
782.
by alecco 2y ago
I'm pretty sure Middle-Eastern investors are having second thoughts on sending more money to OpenAI, Anthropic, xAI, etc. And plenty of AI companies are priced out of the market. And a week ago Google priced their new models along the
783.
by alecco 2y ago
This is quite bad blogspam appealing to the open source crowd. They didn't even bother to read a bit. Deepseek is open source because the founders are part of the new generation of Chinese graduates who relate more to the global youth
784.
by alecco 2y ago
Deepseek is remarkable, but they explicitly say they built on top of Meta's Llama and Alibaba's Qwen. They scored the goal but there were other players involved to get there.
785.
by alecco 2y ago
Even if all that about training is true, the bigger cost is inference and Deepseek is 100x cheaper. That destroys OpenAI/Anthropic's value proposition of having a unique secret sauce so users are quickly fleeing to cheaper alterna
786.
by alecco 2y ago
What about Brazil.
787.
by alecco 2y ago
That would require stealing the model weights and the code as OpenAI has been hiding what they are doing. Running models properly is still quite artistic. Meanwhile, they have access to Meta models and Qwen. And Meta models are very easy
788.
by alecco 2y ago
In my experience most event-sourcing was implemented as storing versions of objects (it came from the OOP camp). All the consistency checks had to be done manually in imperative code across countless classes. Many large investment banks use
789.
by alecco 2y ago
TL;DR https://x.com/INArteCarloDoss/status/1884217440189702194 > Guess what Nassim is trying to say is that given the absurd amount of concentration on the US market, over a third of SPX is Mag, it will take a
790.
by alecco 2y ago
Not OP, but it's been topping app downloads everywhere. A lot of non-tech people are trying it. Note they offer a free reasoning model while OpenAI charges $200. And the reasoning is a lot more transparent. My opinion is probably both
791.
by alecco 2y ago
Event sourcing is a PITA, from experience. Something like Datomic makes a lot more sense: https://vvvvalvalval.github.io/posts/2018-11-12-datomic-even...
792.
by alecco 2y ago
No weights, no proof.
793.
by alecco 2y ago
https://archive.ph/Z3w9H
794.
Chinese Quant Whiz Built DeepSeek in the Shadow of a Hedge Fund Rout(bloomberg.com)
5 points by alecco 2y ago|1 comments
795.
by alecco 2y ago
OpenAI and others are valued for expected future revenue of running the models. And they were also valued as having magic "secret sauce" in their closed source models. Investors are now pulling back from this kind of company. Deep
796.
by alecco 2y ago
It's not a CISCO moment, more like a "Wavelength-division multiplexing" [1] moment of 2000 where the fiber optic craze was over capacity and crashed, causing a lot of "Dark Fibre" [2] left around. Deepseek found a w
797.
by alecco 2y ago
They also did bandwidth scaling to handle work around the nerfed H800 interconnects. > efficient cross-node all-to-all communication kernels to fully utilize IB and NVLink bandwidths > The key idea of DualPipe is to overlap the comput
798.
by alecco 2y ago
How is this very high signal vs noise post out of the front page in 2hs? Are people so upset with the stock market crash that they are flagging it?
799.
DeepSeek limits registrations to Chinese numbers, citing 'large-scale attacks'(businessinsider.com)
5 points by alecco 2y ago|1 comments
800.
by alecco 2y ago
Source: trust me, bro.
801.
by alecco 2y ago
More like a tiny startup from a second tier Chinese city schools tech giants both in America and China. High-Flyer has only 160 employees and only some of those are working on Deepseek. They are beating companies 1000x bigger in budget and
802.
by alecco 2y ago
Big Tech is full of grifter CEOs.
803.
by alecco 2y ago
Hold the bag.
804.
by alecco 2y ago
Deepseek API works fine and it's dirt cheap.
805.
by alecco 2y ago
H800s were made to match Biden's export restrictions. They were banned in late 2023 but a lot were sold to China. Having 2k is quite small compared to the bigger players like BABA (200k employees) and Tencent (100k employees). And thos
806.
by alecco 2y ago
I think part of the problem was many/most previous models coming out of China were trained on eval data to cheat in the rankings. Quite an uphill battle for Deepseek.
807.
by alecco 2y ago
Deepseek shows you details of the reasoning so you can trust its answers more and you can correct it when it makes a bad turn.
808.
by alecco 2y ago
The limit is high quality data, not compute.
809.
by alecco 2y ago
Alice gets a 400k mortgage at bank A, so she gets 400k in credit at her (new?) account at the bank. Alice then pays to Bob for the house by transfering the 400k to Bob's account at bank B. No real money or gold is moved. Alice owes ban
810.
by alecco 2y ago
Professional content creators need to get paid reasonably. AFAIR Tik Tok spent a lot of money getting good content creators in the beginning. YouTube and X pay peanuts but due to scale it's a significant amount of peanuts for top earne

More ›