i've been looking into doing this, the basic algorithms you can get from research, but building a top of the line FPGA design for this is beyond my capabilities as a single engineer, does anyone know of anyone doing this with just some really well optimized c++?
You can't compete with HFT firms even if you have this equipment because you need to install your equipment in a colocation, aka a data center that exchanges sending their raw data and your equipment trades there.
In this kind of trading receiving the data nanoseconds late from the other players matters.
This is only true for arb opportunities on exchanges that offer colocation (which leaves almost all crypto exchanges as fair game). But if you can't get colocation, there isn't much point in having the FPGA anyway since the network latency is going to be so much higher than any savings from FPGA vs low level programming.
yes, this would be non-colocated on AWS. wouldn't everyone have the same network latency if they had a node in the same AWS region? and wouldn't that even the playing field, making FPGAs relevant again (you can rent an FPGA there)? i guess the importance of low latency also depends on the update rate of the exchange server.
the original poster said that this video was sent to him by a fellow researcher who received it on telegram from 'a colleague'. no further details given. if this is truly lk99 showing flux pinning, then that would be very strong evidence that it's also a superconductor. (physics bsc here, my only experience with condensed matter is 1 course and nothing more)
It looks like that permanent magnet it's hovering over is a simple dipole magnet, rather than an array. A diamagnet would be unstable in that position. Look at any video of floating graphite and you'll see it hovering on top of a 2x2 array of magnets.
Of course, you could fake it for a video by buying a special disc magnet with a multipolar magnetization pattern, but magnets like that are pretty hard to find.
If you already had a multipolar magnet for some reason, on the other hand, this wouldn't be too far fetched. You've got to think that with all the interest in this right now, the confluence of having access to a handy collection of peculiar magnets, knowing what to do with a bit of graphite, and being enough of a troll to publish a video, all in a single individual, isn't going to be that unlikely.
It looks (could be faked) like flux pinning is happening. Essentially, it's not moving but staying in place. Imagine trying to balance a magnet on another. It would not inherently stay in place.
Yes - Meissner effect is also called Flux Pinning. Superconductor will remain pinned to the magnet in any orientation. Diamagnet will only repel. The video shows horizontal hover - that confirms Meissner effect.
Flux Pinning is separate from the Meissner effect. Flux pinning only happens in Type II superconductors, where the Meissner effect of expelling external magnetic fields happens with both Type I and Type II superconductors.
From the article: A "stochastic parrot", according to Bender, is an entity "for haphazardly stitching together sequences of linguistic forms … according to probabilistic information about how they combine, but without any reference to meaning."
It seems to me that the great success transformers are now enjoying is precisely due to the fact that 'probabilistic information about how they combine' _is_ meaning.
It's really not. Read the National Library of Thailand thought experiment to understand the difference. But this isn't saying that AGI is impossible, only that it can't come purely from LLMs, and that pure LLMs will remain stochastic parrots no matter how they are scaled up.
I agree. There's a quote in that paper about how ML models can never access meaning (semantics of words) because they only see the form (syntax and letters) and the two are somehow completely divorced.
It's obvious nonsense. I can describe a new concept to you using only words and letters and you can understand it. Therefore you can build up knowledge using only syntax.
Nobody is saying that LLMs understand the layout of a bus or the feel of leather, but they understand that buses are vehicles with four wheels that transport people etc.
sorry, I'm aware that climate change is a serious problem requiring decisive action, but could you explain exactly your claim that 'humanity is done for' by 2026? this sounds somewhat exaggerated and extraordinary claims require extraordinary evidence.
Only thing I disagree with on that blog is that the feedbacks will happen over a longer timeline. But 2026 is the tipping point year where climate change starts accelerating so fast it won't be able to be stopped. Then over the next 5-10 years things like "Clouds Tipping Point" (https://arctic-news.blogspot.com/p/clouds-feedback.html) will be reached. So 2026 is the beginning of the end of human civilization IMO. Of course some people may live underground, in the ocean or other places, but billions of people will die within 10 years following the 2026 tipping point. Billions is the low estimate, more like 99.999999% of humans.
as long as an LLM is a black box (i.e. we haven't mapped its logical structure) then there can always be another prompt injection attack you didn't account for.
Of course it is irrelevant to depositors, that is why the bank failed.
However, it also proves that holding bonds to maturity is different from selling them at market rates. It demonstrates the distinction between solvency and liquidity. Every financial publication has made this point when discussing SVB in order to educate their readership about the problems of duration risk and explain how a bank with enough assets to cover liabilities can still fail.
Now, the nice thing is, the government has time to wait for the bonds to mature. So the government can take the bonds, pay off the depositors, and get the money back when the bonds mature. The government won't lose money if they do it right -- just like they didn't lose money with TARP in 2008.
Algorithmic trader, quantitative researcher, 'retail trader' here. When your strategies are trend-heavy, most of the time you're slowly bleeding out, waiting and doubting your simulations, wondering whether the market has fundamentally changed etc, but a few times a year it's like you're printing money and you finally don't feel like an idiot anymore.
This is usually a reference to the types of securities he trades in - stuff that's available to the general public. It's normal to focus on certain security types even when working at larger firms/groups.
If so, doesn’t that make it trivial for anyone to make a ChatGPT-like app using the GPT-3 API? If so, I would’ve expected several ChatGPT-like tools coming out in the days following GPT-3’s release. I don’t understand why it took so many months - and then, why it was OpenAI themselves that finally released the first interactive thing on top of it. It’s puzzling.
i tried nearly drug under the sun back in the day, played in bands, wrote the best songs during drug comedowns and partied like an animal until suddenly growing up around 30 and getting my bachelor's in physics, a family and starting a business. i have no regrets. mdma, lsd, ketamine and nitrous were my favourites. i still smoke weed now and then on weekends, never drink because i don't really enjoy it. i'm from the netherlands in case that matters.
