FrenchDevRemote

7 ms·

31.
by FrenchDevRemote 2y ago
A while ago I wrote a shell script to open like 100 instances on AWS, running a script on each that would be in charge of downloading a part of the files. using curl or wget could probably be a bit faster than python, you should also run mu
32.
by FrenchDevRemote 2y ago
Really depends on where you buy it. Some real cheese maker, or some random crappy industrial brand? In France we eat cheese made from raw milk all the time, it's almost never a problem, it's so rare that when it does it makes the
33.
by FrenchDevRemote 2y ago
Not an expert, but maybe using RAG/embeddings on the on-disk wikipedia would be better than finetuning on wikipedia? Most decent LLMs probably were already trained on wikipedia, that doesn't stop them from hallucinating when asked
34.
by FrenchDevRemote 2y ago
Just so you know the whole reports are readable without paying from the console through the network requests, might wanna change that ;)
35.
by FrenchDevRemote 3y ago
Is this anything new? seems like it was an obvious use case right from the start and just an ad for your product.
36.
by FrenchDevRemote 3y ago
What would be the benefits of this?Hundreds of man hours+maintenance, for what purpose exactly? It's working perfectly as it is. Adding a shortcut to your homepage is really easy if you're that addicted to HN: https://l
37.
by FrenchDevRemote 3y ago
Multiple options, but any bot detection can be bypassed, if you offer free GPUs you'll run into serious motivated people. Only accept professional emails(i.e name@company.com) from a domain that is older than X days, you can find APIs
38.
by FrenchDevRemote 3y ago
Most AIs doing image recognition/generation use a very low resolution. Training models on higher resolution would become exponentially more expensive, while still not being that likely to succeed. And even if you had the schematics, a
39.
by FrenchDevRemote 3y ago
The prompt in my post was just an example it probably would work.(the prompts I actually used included proprietary/private code/data, so I cannot share them here). Codellama is likely a lot less powerful than GPT.
40.
by FrenchDevRemote 3y ago
Like I said, yes, I've tried dozens of different prompts and custom instructions, most of the time he just ignores them, and refuses to go further after he made a few It also happens when I ask him to write class or functions, he puts
41.
Ask HN: How do you stop ChatGPT from being lazy?
6 points by FrenchDevRemote 3y ago|9 comments
42.
Ask HN: How would you learn AI/ML/DL/DS today?
4 points by FrenchDevRemote 3y ago|0 comments
43.
by FrenchDevRemote 3y ago
It's not experimentation, it's probably one of the only things that allowed them to make gpt 3.5 turbo 10 TIMES cheaper than the previous model.
44.
by FrenchDevRemote 3y ago
That may be because they have a lot more budget and the support of one of the largest cloud providers on earth. Offering free GPU compute to millions of people is really not an easy or cheap task.
45.
by FrenchDevRemote 3y ago
If you mean through the user friendly chat GPT website, they're probably making it output as few tokens as possible to cut costs
46.
by FrenchDevRemote 3y ago
Anthropic models have way inferior capabilities.
47.
by FrenchDevRemote 3y ago
Yes it was pretty easy even though it took like 2 weeks. You just have to make it sound like you could maybe potentially spend money on them one day(instead of just being a curious nerd trying things out)
48.
by FrenchDevRemote 3y ago
They are actually extremely advanced autocomplete. But when you think about it this way, your brain is also just autocomplete. You give it input, tons of input, and your next thought/action/sentence is really just what your brain
49.
by FrenchDevRemote 3y ago
Not only that: Google's recaptcha will also ask you for dozens of captchas on firefox android (which is probably the best) way to watch youtube/use google site without ads on mobile) f** google
50.
by FrenchDevRemote 3y ago
There are free scripts and or tools with free trial like phantombuster. So the response is 0$.
51.
by FrenchDevRemote 3y ago
Do you know how the model accuracy compares to GPT-4 vision?
52.
by FrenchDevRemote 3y ago
you're not buying books when going to a library
53.
by FrenchDevRemote 3y ago
Oh it definitely can. I tried with GPT-4 and Anthropic(claude 2), it was relatively good, although I had to tweak the prompts to get everything I wanted(sometimes it forgot images or urls, because my prompt was too vague) You could also wri
54.
Ask HN: Have you used a good general purpose LLM-powered scraper?
2 points by FrenchDevRemote 3y ago|3 comments
55.
by FrenchDevRemote 3y ago
They've been doing it for years. It's rootkits/RATs, just malware developed by intelligence services and/or some technical branches of police, although they sometime hire external contractors for this. They use exploits
56.
by FrenchDevRemote 3y ago
Believe? Propaganda? Did you ever have a single history lesson in your life? Where do you think Afro-americans came from? Why do you think half the continent is speaking french or English? Why do you think 14 African countries use a money c
57.
by FrenchDevRemote 3y ago
>but I'd really wish for them to come through some community coordinated list of botnet infected IPs or something. Using any kind of community coordinated IP ban is useless and would hurt a lot of people, millions(or even billions)
58.
by FrenchDevRemote 3y ago
If any middle schooler had the same destructive power as a squad of marines embedded clandestinely in a foreign country the world would be in shambles.
59.
by FrenchDevRemote 3y ago
Tailored propaganda, scams, spams, and harassment at a scale that was never seen before. Plugging metasploit into an unfiltered GPT-5 with a shell and a few proxies could be devastating. Undetectable and unstoppable bots would be available
60.
by FrenchDevRemote 3y ago
IMO the next gen AI are going to be tiny nukes that middle schoolers could play with on their iPhones. AI regulation is as needed as radioactive material regulation. Nuclear energy is great, Hiroshima not so much.

More ›