Claude 3 Haiku turns thousands of physical documents into structured data

Поделиться
HTML-код
  • Опубликовано: 26 авг 2024
  • Introducing Claude 3, our next generation of AI models.
    The three state-of-the-art models-Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku-set new industry benchmarks across reasoning, math, coding, multilingual understanding, and vision.
    Learn more in our blog post: www.anthropic....
    Build with Claude: www.anthropic....

Комментарии • 60

  • @mchusma
    @mchusma 5 месяцев назад +24

    Of all the models released, Haiku was honestly the one I was most excited for for some applications. Speed and cost really matters for some applications!

    • @leandrewdixon3521
      @leandrewdixon3521 5 месяцев назад +1

      For those of us actually building things with the API, this is where it's at!

    • @mchusma
      @mchusma 5 месяцев назад +3

      Basically half price gpt 3.5 with ballpark gpt 4 abilities and faster. It crushes almost any other option for most applications (if they release publicly soon).

    • @codelucky
      @codelucky 5 месяцев назад +3

      Can you tell me the use cases and applications of Haiku, Sonet, and Opus? Which models are suitable for Copywriting, Brainstorming, and Article Writing?

    • @InspiredScience
      @InspiredScience 5 месяцев назад

      @@codelucky - Difficult to explain in short comments and YT doesn't like URL's. Google (w/o quotes) *"Anthropic news Claude 3 family"* and click the first link. This is possibly the best comparison overview.

    • @rolandcucicea6006
      @rolandcucicea6006 5 месяцев назад +1

      @@mchusma If only I waited more time before loading openai with another 50bucks

  • @MultiMojo
    @MultiMojo 5 месяцев назад +8

    Excellent way to scan and digitize old textbooks

  • @zhalberd
    @zhalberd 5 месяцев назад +15

    Holy moly! This is exactly what Ive been looking for.

    • @computerscientist9980
      @computerscientist9980 5 месяцев назад +2

      What do you want to do with this?

    • @straylight7116
      @straylight7116 4 месяца назад

      it's biznis secret bro, he ain't gonna tell ya@@computerscientist9980

  • @LaHoraMaker
    @LaHoraMaker 5 месяцев назад +9

    Three brief questions:
    - Is the prompt you used available to compare the performance with other vision/multimodal models?
    - Did you test accuracy of the transcriptions? Are the results available somehow?
    - Do you plan to offer access to Claude to European users before the end of the decade?
    I really enjoyed using Claude but I lost all hope of seeing it deployed to people outside USA.

    • @Regenperf
      @Regenperf 5 месяцев назад

      Why?

    • @doctormoobbc
      @doctormoobbc 5 месяцев назад +6

      I'm using Claude in the EU. What are you talking about?

    • @toonix
      @toonix 5 месяцев назад +2

      @@doctormoobbc Interesting. How?

    • @doctormoobbc
      @doctormoobbc 5 месяцев назад

      @@toonix just by going to their website. They say they're available in the EU.

    • @toonix
      @toonix 5 месяцев назад

      ​@@doctormoobbcHuh. I can register but am capped to about 10 messages. No chance to buy pro. Doesn't feel like it's really available.

  • @theeternalnow6506
    @theeternalnow6506 5 месяцев назад +5

    This would be really interesting to use on the new National Archives UAP documents once they're all there. Great stuff Anthropic.

  • @ashleighlillith1154
    @ashleighlillith1154 5 месяцев назад +2

    Claude is amazing... truly inspirational

  • @aldorodriguez7310
    @aldorodriguez7310 2 месяца назад

    I never knew what model to use. Thank you for this tutorials.

  • @YouLoveMrFriendly
    @YouLoveMrFriendly 5 месяцев назад +13

    Anthropic hitting homeruns.

  • @LiebsterFeind
    @LiebsterFeind 5 месяцев назад +1

    What's the approximate cost of doing that operation you described in the video with the Claude API?

  • @ahmadzaimhilmi
    @ahmadzaimhilmi 5 месяцев назад +3

    Tested vision model with sonnet to extract pdf content. It works fine although it needs further prompt refinement. The only downside is it's expensive 😅. When are you guys releasing the haiku model?

    • @LiebsterFeind
      @LiebsterFeind 5 месяцев назад

      What do you think would be the approximate cost of doing the operation described in the video with the Claude API and the Sonnet engine?

  • @sudeepand
    @sudeepand 5 месяцев назад

    This would be extremely useful.

  • @Vodki7777
    @Vodki7777 5 месяцев назад +7

    It would be super cool if it could be released in Canada someday, using a VPN works, but my card isn't accepted and so I'm restricted to 10 or so messages per 8 hours.

  • @TeamRonDi
    @TeamRonDi 5 месяцев назад

    When will Claude be available in EU countries like Spain or Belgium? It's pretty much available everywhere else but here 🥺

  • @BabakRezai
    @BabakRezai 5 месяцев назад +1

    Is the code used in the video going to be added to the Anthropic cookbook?

  • @isleepy6801
    @isleepy6801 5 месяцев назад

    Awesome work, is there a demo available?

  • @USONOFAV
    @USONOFAV 5 месяцев назад +1

    I unsubscribe to Google Gemini and subscribe to Claude 3. Personally Claude response is the best so far and the most based.

  • @pythontok4192
    @pythontok4192 5 месяцев назад +5

    How about costs? How much per document

    • @jpet11
      @jpet11 5 месяцев назад +3

      Less than than $0.0005 per 1mb image

  • @OzGoober
    @OzGoober 5 месяцев назад

    great demo

  • @chriswatts1361
    @chriswatts1361 5 месяцев назад +1

    Do you have a demo of the code for this somewhere ?

  • @dliedke
    @dliedke 5 месяцев назад +1

    Why Brazil is not supported?

  • @MinhajMalik
    @MinhajMalik Месяц назад

    Intelligence augmentation in action. This is what Klaus Schwab talked about in his book.

  • @GeorgeDonnelly
    @GeorgeDonnelly 5 месяцев назад

    This is awesome.

  • @deeplearningpartnership
    @deeplearningpartnership 5 месяцев назад

    Nice.

  • @escerbo
    @escerbo 5 месяцев назад +1

    It would be great if it can do more then 200k characters

  • @footballuniverse6522
    @footballuniverse6522 5 месяцев назад

    Make it available in Italy too please, I want to subscribe to the Pro but i have to stick with openai :(

  • @LRSKWTKWSK
    @LRSKWTKWSK 5 месяцев назад

    When will Claude 3 will be available in Germany?

  • @vipulvyas7600
    @vipulvyas7600 5 месяцев назад +2

    I expected lot more detailed and more videos..

  • @H1kari_1
    @H1kari_1 5 месяцев назад +1

    Not gonna lie, we are a blue collar company with lots of hand written service protocols and Claude 3 OCR will make our archiving process fully automated.

    • @dg-ov4cf
      @dg-ov4cf 5 месяцев назад

      aint no replacement for human reading skills

  • @HiddenPalm
    @HiddenPalm 5 месяцев назад +6

    Goodbye Google Notebook. Bubbye!

  • @tzenmatteo
    @tzenmatteo 5 месяцев назад

    yuess yes

  • @StrandedKnight84
    @StrandedKnight84 5 месяцев назад

    When is Haiku being released? I need this badly for my startup!

  • @MihikChaudhari
    @MihikChaudhari 5 месяцев назад

    He sounds a bit like Sam Altman

  • @Pavan_AI
    @Pavan_AI 4 месяца назад

    please release mobile app

  • @dg-ov4cf
    @dg-ov4cf 5 месяцев назад +4

    so what? i could do that in a week tops

    • @GamerBath
      @GamerBath 5 месяцев назад +20

      "The AI can do this in a few minutes but I can do this in a week. so what?"

    • @ankitnmnaik229
      @ankitnmnaik229 5 месяцев назад

      Weeks vs minutes..
      That's what...

    • @dg-ov4cf
      @dg-ov4cf 5 месяцев назад

      @@GamerBath @JohnDoe-ru8le
      why are you bozos just repeating exactly what i said

    • @illuminated2438
      @illuminated2438 5 месяцев назад

      Surely this is a troll comment.

  • @datacruncher1564
    @datacruncher1564 5 месяцев назад +1

    Its a great product but the narration is bad. The speaker kept pronouncing many words with low stress and I had to pause and rewind throughout the video.

    • @ashleighlillith1154
      @ashleighlillith1154 5 месяцев назад +1

      that's an AI voice. one of the "uhm"'s was unnatural. it was just like the new GPT robot voice. it's actually really incredible how realistic it sounds, but as you noted, it's still uncanny

    • @datacruncher1564
      @datacruncher1564 5 месяцев назад +1

      ​@@ashleighlillith1154 Thanks for letting me know! I never expected that a human would ever read this comment, let alone respond! Or you could be Claude?

    • @ashleighlillith1154
      @ashleighlillith1154 5 месяцев назад

      @@datacruncher1564 the web is a dark forest now

  • @aquelaquelaquelaquel
    @aquelaquelaquelaquel 5 месяцев назад

    Man.. make it available for EU also.. this is so dumb I can not subscribe