redox99

44 ms·

901.
by redox99 4y ago
In my opinion they are on track to meet the 1.8M deliveries forecast. The only question is how much margins were hit with the price drops.
902.
by redox99 4y ago
I actually enjoy a bit of silliness and things not being so serious
903.
by redox99 4y ago
I don't know about GPT4, but GPT3.5 I'd bet is pretty traditional and boring. It's power comes from a really good, properly curated dataset (including the RLHF). GPT3.5 turbo is much more interesting probably, because they se
904.
by redox99 4y ago
> for (non-violently) stealing money is draconian.. It isn't even non-violently stealing money like breaking into your house when you are on vacation, or hacking a bank or something like that. People just willingly bought a shitcoin
905.
by redox99 4y ago
Happy to help
906.
by redox99 4y ago
Yeah it's incredible they shipped it like that. I bet they planned to ship it with the FSD visualization (which is very impressive), but it got delayed so much that they used that atrocious legacy visualization to meet whatever "d
907.
by redox99 4y ago
I recommend you disable the chime
908.
Adobe Firefly vs. MidJourney AI image generation(twitter.com)
139 points by redox99 4y ago|90 comments
909.
by redox99 4y ago
I'm not denying your feelings. Just pointing out that your art being or not being in their dataset probably makes no difference.
910.
by redox99 4y ago
> Can they even license those works to train a model? Adobe is already creating such AI with these kinds of partners, so yes > And if they can, then everyone can. I can buy the rights to a single photograph of Darth Vader and create a
911.
by redox99 4y ago
> but having your IP ripped out of your own blog feels like being raped, by open-source software or not That's most likely an irrational feeling. "Adobe AI" will most likely work just as well without including your IP in t
912.
by redox99 4y ago
It's pretty funny that most people that hate AI art and want it banned because of copyright violations, don't understand that the only difference that will make is that instead of having open models, you will only have closed mode
913.
by redox99 4y ago
They have a privacy policy like any other business which explains what they do with the data.
914.
by redox99 4y ago
I'm sure in many cases that's already happening.
915.
by redox99 4y ago
That makes no sense. First, people have no trouble exiting. Second, why would you swap BNB to exit? Just exit with BTC, ETH, USDT/USDC, etc.
916.
by redox99 4y ago
> people change the card's backplate and return a cheaper/broken one People really have to ruin everything
917.
by redox99 4y ago
> crypto is somewhere in the range of a couple of minutes to a couple of hours to confirm That's absolutely untrue unless you think "crypto" is only bitcoin. Ethereum block time is 12 seconds, polygon is 2 seconds. And not
918.
by redox99 4y ago
GPT6 or whatever will always require alignment, as the base model just blindly predicts next token, instead of being a helpful, chat style assistant. Right now the best way to align it is with RLHF. The specific technique might change, but
919.
by redox99 4y ago
Here it's outperforming because ChatGPT is already good at these tasks (and the MTurks aren't very good, OpenAI labelers are probably better, and a panel of experts much better). To further improve ChatGPT shortcomings (assuming s
920.
by redox99 4y ago
> ChatGPT access to your existing humongous codebase and having it submit PRs I'm sure that's just around the corner. In 2 years tops it will very likely exist and be useful for some percentage of your JIRA issues.
921.
by redox99 4y ago
PaLM-E: An Embodied Multimodal Language Model https://palm-e.github.io/
922.
by redox99 4y ago
Yes? Attention is all you need and Alpha Zero are the first that come to mind, but there are thousands.
923.
by redox99 4y ago
> Given both the competitive landscape and the safety implications of large-scale models like GPT-4, this report contains no further details about the architecture (including model size), hardware, training compute, dataset construction,
924.
by redox99 4y ago
Google (and others) may not have released model weights, but they've published papers, which is ultimately what makes the field advance. OpenAI not only did not publish any GPT4 paper, they haven't even said how many parameters it
925.
by redox99 4y ago
Don't forget about a more efficient attention that let's them get 32k tokens of context.
926.
by redox99 4y ago
Assuming they didn't change it, they literally randomly add "black" to the end of many prompts to make it more "diverse". So here you got Princes Peach black.
927.
by redox99 4y ago
I agree that for stylized, painting-like 3D models it could be very cool. I was indeed thinking of the typical pipeline for a photoreallistic asset.
928.
by redox99 4y ago
While this is cool, this is not meant to target "game ready". For games and CGI, there's no reason to limit yourself to a single image. Photogrammetry is already extensively used, and it involves using tens or hundreds of ima
929.
by redox99 4y ago
GPT4 should be much better though.
930.
by redox99 4y ago
"sizable percentage" is so vague that it's probably easy to back up that claim

More ›