338
Subscribers
No data24 hours
No data7 days
No data30 days
Posts Archive
338
Every Noise
Every Noise at Once is an ongoing attempt at an algorithmically-generated, readability-adjusted scatter-plot of the musical genre-space, based on data tracked and analyzed for 6,043 genre-shaped distinctions by Spotify as of 2023-03-04. The calibration is fuzzy, but in general down is more organic, up is more mechanical and electric; left is denser and more atmospheric, right is spikier and bouncier.
(https://everynoise.com) — @Ethea
338
If you are using iOS, there is a cool way to enhance your voice input with Whisper AI: Shortcut.
You can also combine it with DeepL shortcut to make a transition from any language to a preferred one. You need to make a bit of customization with this one. (I'll create such a shortcut if there are any requests).
Lastly, you can run a shortcut by tapping the back of your iPhone (double or triple tap). The result of Whisper shortcut will be copied automatically.
338
Spoiler alert: AI isn't going to take your design job.
But there's one company that's sitting on a hidden mountain of data that could radically change the way products are designed. The rest are likely vaporware or toys.
To capitalize on the hype, several companies have teased "ChatGPT for UI". Some are calling it the end of "website designers". 🙄
Most of these tools either won't ever ship (or won't be useful) for one reason: they aren't trained on a sufficiently large data set.
Why does GitHub Copilot work so well? It was trained on the billions of lines of code stored inside Github.
There's one place where an enormous pile of UI data exists in a way that could be used to train a large neural network: Figma.
Behind every Figma design is a secret language: a proprietary data format that specifies everything about a file inside Figma. It's intentionally undocumented for public consumption.
Large neural networks like GPT3 are sophisticated pattern-matching algorithms. They work by looking at a prompt and answering the question "based on everything I've seen before, what's most likely to come next?"
A model trained on this "secret language" can work in exactly the same way. For instance, the model could suggest a correction for inconsistent spacing because it's seen millions of similar examples where the spacing was consistent.
I predict Figma is training an AI on its large, proprietary dataset that will work like GitHub Copilot. Copilot does not write entire programs from a text prompt. It suggests snippets of code based on context.
Similarly, Figma AI will not be a text box to prompt for entire designs (ex: "Design a social network for cats that looks like Linear").
It will make inline suggestions scoped to the context of what you're already working on.
Because Figma has to build and train custom models on proprietary data, their AI product is not as simple as integration with GPT3.
But they are currently hiring ML/AI engineers and I predict we'll see something announced in 2023, possibly at the next Config conference.
In summary:
1. Figma is the only company poised to launch a major AI product in the design space because of its large proprietary dataset.
2. Other AI products in the design space are vaporware or will not be useful.
3. AI will augment designers, not replace them.
(Twit) — @Ethea
338
Repost from Brodetskyi. Tech, VC, Startups
Якщо ви ще не використовуєте LLMs, я рекомендую витратити час на те, щоб розібратись з ними. Якщо ви працюєте з текстом/кодом/листуванням/документами, це точно має сенс.
Я раджу вам спробувати новий Bing, який інтегрує ChatGPT зі свіжими даними з інтернету. Це бомба порівняно зі звичайним GPT. На десктопі працює тільки через Edge, на мобайлі — через апку Bing.
Ось вам список цікавих промптів для експериментів. І результати свіжих досліджень:
Two early papers find the effects of generative AI on knowledge work are completely unprecedented in modern history.
Separate studies of both writers and programmers find 50% increases in productivity with AI, and higher performance and satisfaction. And this is just the start.
338
Repost from Ai ✧ flavour of qual ✧ iA
Boom! ChatGPT API + Whisper API
The new Chat API calls gpt-3.5-turbo, the same model used in the ChatGPT product. It’s also our best model for many non-chat use cases; we’ve seen early testers migrate from text-davinci-003 to gpt-3.5-turbo with only a small amount of adjustment needed to their prompts. Learn more about the Chat API in our documentation.
Pricing
It’s priced at $0.002 per 1K tokens, which is 10x cheaper than the existing GPT-3.5 models.
Whisper API
The API is priced at $0.006 / minute, rounded up to the nearest second.
https://openai.com/blog/introducing-chatgpt-and-whisper-apis
338
#GPT: Here we go again [+ Soundtrack for the post] ——— ## Integrations - Raycast (Link) [Integrating OpenAI to the Raycast application] - Notion (Link) [Notion provided GPT service to their service] ——— ## Semantic Search - Metaphor (Link) [Semantic search with unusual outputs] - Perplexity (Link) (Chrome ext.) [Search engine while you browse] [Search and ask on the page/document content] - Phind (Link) [The AI search engine for developers] - arXiv Explorer (Link) [The Semantic Search Engine for ArXiv] - Andi Search (Link) [Chat for search] - Kagi Search Engine (Link) [Beta tested contextual search with GPT] [In February 2023 will be implemented in Kagi Search] - CrowdView (Link) [Search content from Forums] - Kaito (Link) [Search in Web3 area - crypto and blockchains] [Waitlist] - You Search Engine (Link) [The AI Search Engine] ——— ## Semantic Research - Elicit (Link) [Ask a research questions and get a Research Papers insights] - Researchrabbit (Link) [Surf through the research papers] [not a GPT] (Link) - Consensus (Link) [Ask a question, get conclusions from research papers] - Omniscience (Link) [All-in-one platform for research] - Explainpaper (Link) [Get GPT help while reading the research paper] - Humata (Link) [Ask GPT based on your uploaded files] ——— ## Chrome plugins with GPT - writeGPT (Link) - CasperAI (Link) - Cmd+J (Link) ——— ## Newsletters - ⭐️ Ben's Bites (Link) (Bot) — @Ethea
