crakenzak

10 ms·

1. 91.▲
   by crakenzak 3y ago

   I really like how it has the open ai whisper speech-to-text model built in so you can use your voice to talk to the model.
2. 92.▲
   How Is AI Affecting the Future of Productivity Interfaces?(warp.dev)

   4 points by crakenzak 3y ago|2 comments
3. 93.▲
   ChatGPT plugins now available for all Plus subscribers(twitter.com)

   2 points by crakenzak 3y ago|0 comments
4. 94.▲
   by crakenzak 3y ago

   Code: https://github.com/facebookresearch/ImageBind Blog Post: https://ai.facebook.com/blog/imagebind-six-modalities-bindin... Demo: https://imagebind.metademolab.com/
5. 95.▲
   by crakenzak 3y ago

   Code: https://github.com/facebookresearch/ImageBind Blog Post: https://ai.facebook.com/blog/imagebind-six-modalities-bindin... Demo: https://imagebind.metademolab.com/
6. 96.▲
   by crakenzak 3y ago

   Code: https://github.com/facebookresearch/ImageBind Blog Post: https://ai.facebook.com/blog/imagebind-six-modalities-bindin... Demo: https://imagebind.metademolab.com/
7. 97.▲
   Meta AI releases ImageBind demo: Holistic AI learning across six modalities(imagebind.metademolab.com)

   4 points by crakenzak 3y ago|4 comments
8. 98.▲
   by crakenzak 3y ago

   The 6 modalities that the model are images, text, audio, depth, thermal, and motion data. Thermal and motion data completely caught me by surprise. Huge step towards laying down the framework for AIs that understand the world around them li
9. 99.▲
   ImageBind: Holistic AI learning across six modalities(ai.facebook.com)

   17 points by crakenzak 3y ago|4 comments
10. 100.▲
    ImageBind: Open-source 6-modal generation ML model released by Meta AI(github.com)

    5 points by crakenzak 3y ago|2 comments
11. 101.▲
    by crakenzak 3y ago

    https://archive.ph/KtCxg
12. 102.▲
    OpenAI Shap-E: 3D Shape Generation(arxiv.org)

    6 points by crakenzak 3y ago|2 comments
13. 103.▲
    by crakenzak 3y ago

    Very excited to see how this comes along! Hopefully we have a JS->TS style jump here! Some of the performance gains look absolutely ludicrous.
14. 104.▲
    by crakenzak 3y ago

    I'm curious if future developments in LLMs will enable them to extract significant/noteworthy info from their context window and incorporate it into their underlying understanding by adjusting their weights accordingly. This could
15. 105.▲
    by crakenzak 4y ago

    > On the downside, you now need to heat the cabin manually, rather than reusing the ICE heat. Most modern airliners do not use bleed air for climate control in the cabin anymore.
16. 106.▲
    by crakenzak 4y ago

    Eid Mubarak all! So grateful for this community -- feels like I've learnt more through here than all of university ;)
17. 107.▲
    by crakenzak 4y ago

    Every single member of the research team that invented the transformer architecture have left Google to go to OpenAI, or make their own startups (character.ai, anthropic, cohere)
18. 108.▲
    by crakenzak 4y ago

    good. the era of overinflated real estate in SF is hopefully coming to an end.
19. 109.▲
    by crakenzak 4y ago

    Can anecdotally attest to this, working at Microsoft is slowly becoming more sought after than Google amongst me and my peers (a few of which work at Google with me).
20. 110.▲
    by crakenzak 4y ago

    https://web.archive.org/web/20230418182954/https://www.nytim...
21. 111.▲
    by crakenzak 4y ago

    hmm, could increasing the amount of data in other languages potentially improve performance within the english portion of the model? from it building deeper understandings of concepts that are found within other languages.
22. 112.▲
    FerretDB: A truly open-source MongoDB alternative(github.com)

    3 points by crakenzak 4y ago|0 comments
23. 113.▲
    by crakenzak 4y ago

    very interested to see where this goes, hopefully this promotes more competition with OpenAI, as I'm getting a little uncomfortable with them having a near monopoly on high quality LLM models.
24. 114.▲
    by crakenzak 4y ago

    I agree with your general statement, but if one thing the recent DeepMind Chinchilla research paper showed us is that the size of the model (number of parameters) is much less of the determinant of model performance & quality than the a
25. 115.▲
    by crakenzak 4y ago

    This is going along with the new Segment Anything Model paper Meta AI just released: Paper: https://scontent-sea1-1.xx.fbcdn.net/v/t39.2365-6/10000000_6... Announcement: https://ai.facebook.com/blo
26. 116.▲
    Segment Anything Model (SAM) can "cut out" any object in an image(segment-anything.com)

    725 points by crakenzak 4y ago|152 comments
27. 117.▲
    by crakenzak 4y ago

    Wow, glad they decided to provide all the model weights (not behind a request form this time!): https://github.com/facebookresearch/segment-anything#model-c... very impressive work by Meta AI
28. 118.▲
    by crakenzak 4y ago

    what is the benefit to such splitting of text based on seemingly meaningless lines? isn't there a better way to do it?
29. 119.▲
    by crakenzak 4y ago

    relevant research paper: https://arxiv.org/abs/2302.12813
30. 120.▲
    by crakenzak 4y ago

    Is there any chance that the transcripts can be published on the website? I'd love to be able to search through them. love the course btw!