1465
Reporting AI nonsense. A future news media, driven by virtual assistants 🤖
New Claude feature aims to handle files exceeding current context limits
https://www.testingcatalog.com/new-claude-feature-aims-to-handle-files-exceeding-current-context-limits/
🗞 #claude
This might become a competitive solution against large context window models like 2M Gemini 1.5 Pro.
If Claude manages to query ref files properly (which it seems to be, not very good at currently, on a model level) - it will be a huge step for coding 🔥
This is how it looks currently if you try to upload a file that exceeds the 200k context window. The same would happen if you try to upload it into a project as a knowledge base.
Читать полностью…
However, I noticed only now that it shows you console errors inside the artefact 🤯
Читать полностью…
Seems like MacOS client isn't loading the content of the canvas properly yet (or is it only me?).
But we will be able to open them in a side-sheet, similarly to the web UI.
Canvas model selector is also not available there yet 🚫
Final polish adds better formatting which works well with text inpainting and text formatting options.
ChatGPT canvas works with 4o currently as a separate model option. Looking forward to seeing how will it perform with o1 👀👀👀
Length adjustment along with a writing level open 2-click editing along with an easy switch between different reversions 👀
Читать полностью…
Perplexity working on new Collections UI for custom instructions and file uploads
https://www.testingcatalog.com/perplexity-working-on-new-collections-ui-for-custom-instructions-and-file-uploads/
🗞 #perplexity
Black Forest Labs introduces faster, more efficient flux1.1 pro
https://www.testingcatalog.com/black-forest-labs-introduces-faster-more-efficient-flux1-1-pro/
🗞 #ai
OpenAI is rolling out Canvas to Plus users, achieving better performance at certain writing and coding tasks in comparison to normal prompting 👀
It seems to be still in rollout as many users don't have access to this feature yet.
https://openai.com/index/introducing-canvas/
Files management UI looks also similar to the one which will be available in settings later when feature will be released 👀
Читать полностью…
Perplexity collections will get a new UI for displaying custom instructions and adding files.
Читать полностью…
ICYMI: YouTube’s conversational AI available for U.S. Premium members
https://www.testingcatalog.com/icymi-youtubes-conversational-ai-available-for-u-s-premium-members/
🗞 #youtube
BREAKING 🚨: Microsoft started rolling out a new mobile Copilot experience for iOS including a new voice experience with 4 different voices available for you to choose from.
h/t QRoonjha
ICYMI: AI Studio model selector is now split into categories like Gemini 1.5, Gemma, Preview, etc.
Are we about to see a new category out there soon? 👀
Maybe as I started about magic wands, Claude will get it available for artefacts as well.
A new button will allow users to ask Claude to improve their artefacts even further with one tap (Still WIP).
As you can see on the first screenshot, Claude allows uploading exactly the same file.
Later on, you can prompt Claude to extract more and more parts out of it chunk by chunk.
BREAKING 🚨: Anthropic is working on reference files support that potentially will allow users to work with files that are larger than Claude's context window!
Disclaimer: This feature is not released yet 🚧
I only had a chance to play with it now. Why may be a big deal 👇
Claude now can try fixing its own rendering errors 👀
Have you seen any? I gave up trying to reproduce w/o using any extra magic 🪄
ChatGPT Canvases are also supported on iOS (partially).
You cannot start a new conversation yet but in conversations that contain Canvases you can open them in a separate view 👀👀👀
It also allows you to add emojis to the text but takes this task way too seriously and adds 2-3 emojis to every sentence now.
Читать полностью…
Got access to the ChatGPT Canvas beta too and it seems like it will be a huge tool for writing 🤯
Understanding a context shows much better results in my tests along with an improved writing style - the output is much more humanized.
OpenAI unveils Canvas for coding and writing projects in early beta
https://www.testingcatalog.com/openai-unveils-canvas-for-coding-and-writing-projects-in-early-beta/
🗞 #chatgpt
Microsoft launches updated Copilot AI with new voice mode
https://www.testingcatalog.com/microsoft-launches-updated-copilot-ai-with-new-voice-mode/
🗞 #microsoft-copilot
ChatGPT may get some new keyboard shortcuts for web 👀
Cmd + K to open a new chat may appear to Mac users at some point.
File upload will likely arrive as a part of upcoming Files search feature and currently not available yet.
Читать полностью…
ICYMI: Now you can delete images from Grok's image history tab 👀
Читать полностью…
The voice mode itself doesn't seem to be working for me yet 👀👀👀
In addition to voices, in the settings, it will be also possible to choose voice speed.
ICYMI: Pika Labs is back! Now, with its improved 1.5 video gen model and new Pikaffects allowing users to melt, explode or crush any object on the image 🔥
Trying to see if it can crash "Meta" logo for me 👀
ICYMI: Kling AI got a lips sync feature along with API support for AI video generation 👀
Читать полностью…