sinuhe69

37 ms·

661.
by sinuhe69 4y ago
How authentic is the result compared to what John Mayer did with Steve Job's voice? https://twitter.com/BEASTMODE/status/1637613704312242176
662.
by sinuhe69 4y ago
Oh, thanks. I thought somebody got a fine-tuning version of GPT-4. However, the author used the exact same prompts to demonstrate the ability of his interface, so I don't think it says much about its fitness for the purpose.
663.
by sinuhe69 4y ago
Anybody knows how it was done? Did GPT-4 generate all the codes, or did the author fine-tune GPT-4 on Blender code?
664.
by sinuhe69 4y ago
Oh, I didn't know they even have a physical store! Thanks for pointing out. Yeah, I know they also archived digital audio and other stuffs, but never thought book archiving lies in their realm as well.
665.
by sinuhe69 4y ago
I thought the IA only archives websites and digital contents, which are in danger of being lost? eBooks are seldom the only form of publishing, thus I believe there is always at least an archive copy of them. Making them accessible for all
666.
by sinuhe69 4y ago
If Javascript is the target and the baseline benchmark, I wonder what is the rationale not using Javascript directly? Surely, the Python syntax is not the sole motivation, isn’t it?
667.
by sinuhe69 4y ago
Probably the cheapest solution would be letting GPT monitors user feedbacks from various social media channels and alert human engineers to check for the summative problem. GPT can even engage with users to request more details or reproduce
668.
by sinuhe69 4y ago
If I want to watch a thread, I simply favorite or upvote it, so I can always come back quickly using my account. It’s all built-in, no need for other techs :) Yes, even with that, sometimes a thread is so convoluted and so long, that is har
669.
by sinuhe69 4y ago
It’s not exaggerated to say DPReview is an important bit of history. Wonder if anyone will go archive them before it’ll vanish forever? It’s really sad.
670.
by sinuhe69 4y ago
It’s August 2022, no wonder the information is outdated. Please change the title to include the year. Today, if you want to run SD locally, you have ways more options, particularly Automatic1111
671.
by sinuhe69 4y ago
That is a basic deployment scene for OpenAI GPT API: you have a repository of text, GPT will work on it then answers your question. Just take a look at the OpenAI API page (and share your experience here :) )
672.
by sinuhe69 4y ago
OK, I repeated the same experiment, though with Google instead of GPT. I just translate the emoji to text (each emoji has associated keywords): movie girl fish clownfish pufferfish And the first result? Finding Nemo! Thus the association is
673.
by sinuhe69 4y ago
I copied the opinion of Yann LeCun, one of the authorities on deep learning: (Feb 13,2023) My unwavering opinion on current (auto-regressive) LLMs 1. They are useful as writing aids. 2. They are "reactive" & don't plan no
674.
by sinuhe69 4y ago
So it continues to reaffirm what we’ve known: generative LLM does not have a model of the world, can not reason and can not plan. It generates text by mix-matching remembered texts and thus it can not generate truly new content. No surpris
675.
by sinuhe69 4y ago
I think we should not undervalue DeepL. Not only its default-translation is already very good, it allows users to select different alternatives and remember these preferences, too. Which is not possible, at least not easy with GPT. And as w
676.
by sinuhe69 4y ago
All the features it mentioned is already implemented in F#, of course without the silly curly brackets. And of course it has also remembered the pro and contra in the design choices of making F# and similar languages, thus I would not read
677.
by sinuhe69 4y ago
At worst, it’ll be kind of infinite recursive with psychopathic tendency. If more lucky, the law of dismissing return will strike unmercifully.
678.
by sinuhe69 4y ago
It all comes down to labelled data. There are millions images of VGA connectors and lightning cables on the internet with description, where CLIP model and similar could learn to recognize them relatively reliably. On the other hand, I&#x27
679.
by sinuhe69 4y ago
If the Star Trek computer hallucinated like ChatGPT, Captain Picard and his crew would end up inside a star long ago!
680.
by sinuhe69 4y ago
It's a surprise that GPT-4 could improve so much in maths, especially GPT 3.5 didn't seem to understand many word problems correctly. For example, ChatGPT still struggle with this very simple problem, how GPT-4 could do much bette
681.
by sinuhe69 4y ago
For anyone who interests, A Little History of Science of William Bynum is recommended. A book written for school children, but IMO highly accurate and of course accessible.
682.
by sinuhe69 4y ago
On the first diagram, the y-axis is the absolute brightness of the stars in the visible spectrum (V) and the x-axis is the difference between magnitude in blue (B) and the entire magnitude of the star in visible spectrum (V). Thus on the le
683.
by sinuhe69 4y ago
Yeah, as a non-native speaker, I was a bit confused about “demographics” as well. What did I expect? Nothing. It’s only felt “off” with stars and demographics. A “characteristics” or distribution on the other hand would click immediately. (
684.
by sinuhe69 4y ago
It can not guarantee a unique solution, though. Which in terms of logic puzzles is considered bad. In this particular game, you can choose any of the four 3s and the correspondent 3 and the game will end successfully. Another problem is of
685.
by sinuhe69 4y ago
I don’t know thuy you need that but in many chess programs, you can set that when you touch a figure, it will display all possible moves. Thus, you only need to touch all of them :)
686.
by sinuhe69 4y ago
Got it. The article didn’t go into such details, only mentioned “Pic, or it didn’t happen” part, thus my question. PS: I’d say the article was poorly written, but now I wonder why it’s flagged? It seems HN users nowadays are ways more hot-t
687.
by sinuhe69 4y ago
Don’t know about the other humans. But for identifying of AI-generated texts (intermixed with human written texts), I use mostly logic. AI don’t truly understand text, that’s why by extending the text, it often makes plausible and convincin
688.
by sinuhe69 4y ago
What was the “mocking” part? “Pics, or didn’t happen”? Ps: I’m no native speaker and I don’t use Twitter. So I’m not sure about the tone and the behaviors here. That’s why my asking. For me, it doesn’t sound like mocking but disbelieve (but
689.
by sinuhe69 4y ago
You can use CLIP to build a database to search (offline) for keywords like “nude” or “naked”. Specifically, I use clip-anytorch with the ViT-B/16 pertrained model and find the result very good. Just go to pypi and the corespondent Git
690.
by sinuhe69 4y ago
Some are pretty old. I wonder why there is a “now” in the list. By the way, not all are hard science fiction. Some are merely fantasy.

More ›