emporas

23 ms·

361.
by emporas 3y ago
>LLMs are really smart, and also really, really dumb Machines will always be dumb. The notion that machines are smart, probably stems from people inexperienced using engines. In chess we use engines for more than 20 years and humans alwa
362.
by emporas 3y ago
Centralization or decentralization is a mute point. The problem is solved using PKI, which was defined as a theory back in the 90's. PKI can be supported by any government or official with some computers, even private companies can off
363.
by emporas 3y ago
I have heard similar stories of Amazon fraud like you describe. I don't know if everything is true, but this is a problem which can be solved with PKI. With PKI, the customer reviews a product, and puts the hash of his review somewhere
364.
by emporas 3y ago
>Model merging is, to me, one of the most counterintuitive empirical results in modern deep learning. It turns out that you can actually more-or-less naively merge the weights of two different models and produce a new one that captures s
365.
by emporas 3y ago
The proposition coding assistants like Copilot and Sourcegraph make, is that they can help in writing code for every programming language, and speed up a lot the process of software development. While this is true, i consider as a lot more
366.
by emporas 3y ago
I agree with most of your points, but why not encrypt sensitive information yourself before it gets uploaded/shared on your dropbox account? Sure, it's not end to end encryption but it prevents the company from using the encrypted
367.
by emporas 3y ago
If the cells lack arteries, proper arteries with nutrients, leukocytes, immune system etc, then their lifespan will be a lot less than 7 years. Pretty amazing actually that everything else is easy, or not difficult at least, and that's
368.
by emporas 3y ago
The thought emporium channel is great. There is one really good video with an explanation of the process, brain cells to computing devices. https://www.youtube.com/watch?v=67r7fDRBlNc And one more video, not very relevant,
369.
by emporas 3y ago
Thanks for letting me know, i didn't use GPT-4, but i was under the impression that the cutoff data between all GPT's was the same, or almost the same. The code is correct, yes. I do not have a GPT4 subscription, i did not bother
370.
by emporas 3y ago
From the article: "In contrast, Rust's explicitness in this area not only made things simpler for us but also more correct. If you want to set a file permission code in Rust, you have to explicitly annotate the code as Unix-only.
371.
by emporas 3y ago
I didn't know about RAG, thanks for sharing. I am not sure, if outdated information can be tackled with RAG though, especially in coding. Just today, i asked GPT and Bard(Gemini) to write code using slint, neither of them had any idea
372.
by emporas 3y ago
Given that 50% of all information consumed in the internet is produced in the last 24 hours, smaller models could hold a serious advantage over bigger models. If an LLM or a SmallLM can be retrained or fine-tuned constantly, every week or e
373.
by emporas 3y ago
That's smart, but very insidious as well. Following in the footsteps of dark UI patterns designed for users to misclick on ads, now dark suggestions on conversations with an LLM will be the next big thing. Like a conservative talking w
374.
by emporas 3y ago
Definitely Forth and Factor. Every programmer should get a little bit familiar with the concatenative style. 3 days ago i discovered a channel on YT which talks about Forth, even Chuck Moore himself did a talk there talking about GreenThrea
375.
by emporas 3y ago
As someone who created over 500.000 images last time i checked 6 months ago, i don't know where that notion comes from, that A.I. images need no work. Key here is that the artist wants to express an idea, an emotion something like that
376.
by emporas 3y ago
Algorithms and data structures however? Still important! Rust implements some algorithms [1] for the same API with the sole purpose to be interchangeable and readily replace one with another. Does a computer know how to do that? I find that
377.
by emporas 3y ago
>Clearly, they aren't as likely to generate "Colorless green ideas sleep furiously" as "It's important to feed your dog". There is a setting right there called temperature, no?
378.
by emporas 3y ago
There is not much of a need for humans to get closer to the machine long term, when with new datasets for training the machine will get closer to humans. Magic keywords like "step by step" won't be as necessary to know. One o
379.
by emporas 3y ago
I would generalize even more and say that any scripting language is going to be deprecated very soon, like Python etc. They are going to be replaced by safe, type-checked, theorem-proved verbose code, like Rust or something similar. What do
380.
by emporas 3y ago
In general, one should not instruct GPT to solve a problem. The instructions should be about generating code, after a human thought process took place, and then generate even more code, then even more, and after merging all the code togethe
381.
by emporas 3y ago
Meaning, we should focus on our desires and what products or services satisfy them better. Money is an automation of what desires need to be produced more in the future, so as to be offered in bigger quantities and/or cheaper. Edit: Nv
382.
by emporas 3y ago
I have listened a lot of Varoufakis's speeches some years back, in Greek as well as in English, when i had no idea of economics. Varoufakis was always very critical of the tendency of capitalism and market forces to create powerful com
383.
by emporas 3y ago
Cryptography can be used to hide something, or prove something. The word cryptography encapsulates two different disciplines, cryptography and provegraphy. People who use the term crypto-* to refer to blockchain, do not know that blockchain
384.
by emporas 3y ago
So, statically link the new sudo with the old sudo, and when the user tries to use less commonly used features fallback to the older version in C. The binary executable will be bigger, but disk space is a lot cheaper nowadays, compared to 4
385.
by emporas 3y ago
More like, "wasm is the new Emscripten". A much more formalized standard across all browsers. It goes to the same direction as LLVM but a far cry from that.
386.
by emporas 3y ago
Well said. Is technology that much more than ideas? Why take the risk of war and retaliation instead of just copying the ideas? The implementation of ideas is not trivial, but given the right combination of people and specialized labor, ide
387.
by emporas 3y ago
Back in 2008, i remember reading books thousands of pages long, about genetics in biology, and i was impressed by how easy the subject is. I was an amateur in programming at the time, but programming, regular programming of web servers, web
388.
by emporas 3y ago
The OSes want programmers to handle resources correctly, and the Rust compiler makes that task a breeze. We are in a more abusive relationship with our OSes, than the Rust compiler. How about the hardware? Doesn't that need to run asse
389.
by emporas 3y ago
I put GPT and Llama 70B to write me comments about the Rust language source code. They are both superb at this task. I haven't used any other weaker LLMs in the same fashion, with less billions of parameters, but i can imagine many o
390.
by emporas 3y ago
Yep, Asianometry puts information out there, of another level. One more channel i really like, is NightHawkinsLight. He did a video yesterday about multivariate analysis, i was blown away.

More ›