rnosov

9 ms·

61.
by rnosov 4y ago
This project might actually seriously poison any future datasets for AI training. Some conversations are really fun though
62.
by rnosov 4y ago
I think they should fix missing hands, legs, fingers etc first
63.
by rnosov 4y ago
Requires signup to even look at the content. That's a pass for me...
64.
by rnosov 4y ago
You might want to at the very least specify compensation (or lack thereof)
65.
by rnosov 4y ago
Video unavailable The uploader has not made this video available in your country
66.
by rnosov 4y ago
How is it different from ChatGPT? The only example on the landing page is "Write email to welcome new customers". ChatGPT wrote me a nice email in response to this prompt without any follow up questions.
67.
by rnosov 4y ago
I assume lots of high end Nvidia cards like A100. Time to buy NVidia stock... ChatGPT response for the reference: "As an AI language model, I am a software program that runs on computer servers hosted in data centers. The specific hard
68.
by rnosov 4y ago
Just for the avoidance of doubt "AI Files" is not my project and I'm not affiliated with the author in any way. osascipt line does look a bit dodgy to me but perhaps is safe. But I can see how things might go downhill quickly
69.
by rnosov 4y ago
Answering to myself as I can't edit my response any more. I've investigated a bit more and the attack as described by Greshake[1] seems to be much more realistic to me than I initially thought. [1] https://github.com&#x
70.
by rnosov 4y ago
PDF reader hack will be indeed dangerous. I've looked at your link and they suggest that LLMs will fetch random files of the internet. At the moment, no LLM will do that. Network requests are currently off limits. In the paper they see
71.
by rnosov 4y ago
If a user can be tricked into downloading files and then running them and why not trick a user into downloading and running actual malware? With your pdf scenario it would be a possibility, but at the moment OpenAI davinci doesn't foll
72.
by rnosov 4y ago
I won't describe it as malware. Your link describes prompt injection which applicable to any software that currently employs LLMs (including this package). To successfully exploit it an attacker would need to place a file with maliciou
73.
by rnosov 4y ago
Wikipedia is your friend https://en.wikipedia.org/wiki/Worker_cooperative Not many success stories so far. A startup can be a cooperative but there will be no stock to grant. One worker = One vote. You can probably see
74.
by rnosov 4y ago
What would alternative be? Worker run cooperatives have not been particularly successful so far. Serf had no choice of "changing a job" but a modern worker presumably has a lot more freedom to choose their employer.
75.
by rnosov 4y ago
It should be feasible with GPT-J. You should be able to run it locally if you have GPU with more than 16gb of video memory. Output quality might not match OpenAI offerings though. https://playground.helloforefront.com/models
76.
by rnosov 4y ago
More examples on the AudioLM page. Some are pretty impressive (assuming they are cherry picked). https://google-research.github.io/seanet/audiolm/examples/
77.
by rnosov 4y ago
I think you might want to correct ChatGPT to just GPT. As far as I know, there is no public API access point for the ChatGPT. OpenAI davinci model that you're likely using is based on InstructGPT ( a different GPT based beast ). Also,
78.
by rnosov 4y ago
In a same vein, you can argue then you can't empirically test that every proton in your hand has, say, three quarks or whatever, as you loose your hand. You then might say that all knowledge about protons in your particular hand isn&#x
79.
by rnosov 4y ago
> I had some existential concerns about not knowing whether I was just AI You might want to dive into simulation argument by Nick Bostrom https://www.simulation-argument.com/simulation.pdf https://en.wikipedia
80.
by rnosov 4y ago
Good catch! The author's name "Heorhii Skovorodnikov" sounds Ukrainian and he describes himself as a graduate of NYU Abu Dhabi on his Github page. Looks like he is not a native english speaker and that might explain the need
81.
by rnosov 4y ago
There are plenty of LLMs that are completely open source that you can use (like BLOOM for example). There's also a massive original use case of machine translation for LLMs which is a real world problem. AI image generation seems WAY m
82.
by rnosov 4y ago
As of now, I'd argue ChatGPT has a lot of value for an average student who has to write an essay "by tomorrow".
83.
by rnosov 4y ago
It's been 14 years since bitcoin mined its first block. Still not clear what ( if any ) the use case for it is.
84.
by rnosov 4y ago
I think it's "we" as in "real humans" vs them "AI chatbots". It would be easy to avoid them at the moment but might be a lot harder in the future (in a way it becomes difficult to use physical cash in many
85.
by rnosov 4y ago
Well I've addressed it by specifying December timeframe (height of the polar night on the North Pole without any sun to speak of). Sometimes ChatGPT even addresses the issue of the lack of sunlight in it's reasoning and still reco
86.
by rnosov 4y ago
What strikes me the most about ChatGPT is how convincing it is at justifying patent nonsense. For example, I've noticed it has some strange obsession with sunscreen. Even for absurd questions such as "Should I wear sunscreen on th
87.
by rnosov 4y ago
The problem here is that most people just don't have suitable hardware. Ideally, you'd want to load the entire model into a GPU and most consumer grade GPUs just don't have nowhere near enough video memory. You'd need to
88.
by rnosov 4y ago
One way to think about it is that the model needs to essentially encode the entirety of human knowledge. If you can do it with just 175b parameters then it looks quite efficient to me. GPT-3 is about 400gb in size which would even fit in so
89.
by rnosov 4y ago
Y axis are Github stars. They sort of mention it in the preceding paragraph.
90.
by rnosov 4y ago
I think they are correctly referring to ChatGPT as GPT-3 + RLHF. In other words ChatGPT = GPT-3 + RLHF. So, 80GB A100 GPU would be required for both GPT-L AND RLHF (PyTorch version). And it looks to me from the TFA that the main thing that

More ›