novaRom

11 ms·

121.
by novaRom 3y ago
you can run opnellama on a 300 Euro laptop with ~500 ms/token performance in fact you can run any decent model of 3B-7B size on any contemporary hardware with enough RAM
122.
by novaRom 3y ago
quote: Germany ranks only eighth in the Global Innovation Index Is it so bad? For 83 millions people economy.
123.
by novaRom 3y ago
Sorry, ReplayWeb.page won't work in this browser as Service Workers are not supported. Please try a different browser. (Service Workers are disabled in Firefox in Private Mode. If Using Private Mode in Firefox, try regular mode)
124.
by novaRom 3y ago
There are tens of similar projects on GitHub, most are Python-oriented and use OpenAI API.
125.
by novaRom 3y ago
"Laptops don’t change" - this is most important reason if we're talking about Apple laptops. My 10 years old MacBook Pro is almost absolutely the same as MacBook Air of 2023. Performance, screen, OS, weight, battery, keyboar
126.
by novaRom 3y ago
I know some friends of mine stopped using it because it requires login, and it's quicker to Google or go to Wiki for a quick answer. Most people have a laptop at work, home devices, phone, etc. So it's just you ve tried it say on
127.
by novaRom 3y ago
Multiple reasons, but most important: 1. ChatGPT is "play twice and forget" for 90% of users, driven by curiosity. 2. ChatGPT’s answers have actually gotten significantly worse over time. 3. No fun. It's too restricted, hesit
128.
by novaRom 3y ago
TL;DR When I create I learn. When I consume I just relax.
129.
by novaRom 3y ago
How small can be a LLM transformer in order to be able to understand basic human language and search for answers on the internet? It should not contain all the facts and knowledge, but must be quick (so, it's a small model), understand
130.
by novaRom 3y ago
Important addition to your partially right statement: "they’re trained to generate ‘likely’ text" is they are trained to produce most probable next word so that the current context look as "similar" to training data as p
131.
by novaRom 3y ago
I would rather say there is existential threat to get all human knowledge spoiled, mixed with fabricated high quality facts, faked textbooks, faked movies, faked replicas of original scans, etc.
132.
by novaRom 3y ago
Always use containers! Any big project should be sandboxed, having completely isolated environment, dependencies, and you have all installation steps well described and reproducible. Containers in Linux are lightweight citizens and using th
133.
by novaRom 3y ago
I don't speak Danish but I remember how amazing is their encyclopedia "Lille Salmonsen", one I "read" every evening in my hotel room in Kopenhagen. It was 1940 edition. Somehow with my German and English, and a bit
134.
Hollywood Writers Strike AI(theguardian.com)
1 points by novaRom 3y ago|0 comments
135.
by novaRom 3y ago
Yes. It's pretending. The best way to think about chatGPT answers is that it always invents the most plausible reply. With some different temperature it can provide slightly different chain of thought, but it's making it up based
136.
by novaRom 3y ago
“Alice does not know that she does not know the painter of the Mona Lisa.” I (human) need to read it few times to understand.
137.
by novaRom 3y ago
> ML hasn't done anything new since the 80s Schmidhuber, is it you? But seriously, it's partially true most techniques we use today could be found initially envisaged in 90th ANN-related papers. I think Geoffrey Hinton summariz
138.
by novaRom 3y ago
Germany has almost no hydropower, no nuclear power anymore, but German government is continuing to insist a similar ban on natural gas in all new buildings should be introduced (or maybe already in action).
139.
by novaRom 3y ago
Interesting dialogue, it's a kind of computer game, or better call it AI game and LLMs are incredibly good playing any character. It's fun to time travel, transform into different beings, thinking like an alien, etc.
140.
by novaRom 3y ago
Quote: I'll be surprised if this event occurs before 2005 or after 2030 Actually, my initial thought was this text generated by a LLM pretending it's from 1993
141.
by novaRom 4y ago
The problem is interdisciplinary. We have to unite first and make a clear plan which should answer following questions: 1. How to get a stable non-monopolistic semiconductor industry with multiple players without critical dependencies on a
142.
by novaRom 4y ago
and it will work on any smartphone because: The image encoder has 632M parameters. The prompt encoder and mask decoder have 4M parameters.
143.
by novaRom 4y ago
Never say never. Next 2 or 3 years it will assist, but chances are very good AI will completely automate software and hardware development in several years.
144.
by novaRom 4y ago
> Even the mighty Google Since the release of the Attention paper, they havent come up with any groundbreaking idea, that was five years ago. Where is their research? All they seem to have are technical descriptions with scarce details,
145.
by novaRom 4y ago
We need to build even bigger and better LLMs ASAP "primus inter pares". Better models will help us finding better solutions.
146.
by novaRom 4y ago
2B weights is not too big, so with some compression + sparsity + down-sampling it will run on any device offline
147.
by novaRom 4y ago
Unfortunately, LLMs are still prone to make facts up, and very persuasively. In fact, most of non-trivial topics I tried have required double-/triple- checking, so it's sometimes not really productive to use chatGPT. You are corre
148.
by novaRom 4y ago
Is there any estimate how much training data someone like MindGeek has?
149.
by novaRom 4y ago
Yes, you can run any inference with any model on CPU, but here some examples: To create a single frame with Stable Diffusion 4-5B parameters, 512x512, 20 iterations takes 5-30 minutes depending on your CPU. On any modern GPU it's only
150.
by novaRom 4y ago
chatGPT is great to rewrite certain text in grammatically correct and simplify or make it more formal, etc.

More ›