Simon Willison
Simon Willison
  • Видео 30
  • Просмотров 125 761
Datasette comments, pins and write UI with Alex Garcia
Alex Garcia demonstrates three plugins for Datasette: datasette-write-ui, datasette-pins and datasette-comments.
github.com/datasette/datasette-write-ui
github.com/datasette/datasette-pins
github.com/datasette/datasette-comments
Просмотров: 96

Видео

Simon Willison on new Datasette Enrichments
Просмотров 1969 часов назад
Simon Willison demonstrates the latest improvements to Datasette Enrichments - enrichments.datasette.io/ and github.com/datasette/datasette-enrichments
llm-questioncache with Nat Knight
Просмотров 599 часов назад
llm-questioncache builds on top of llm.datasette.io/ to cache answers to questions, using embeddings to return similar answers if they have already been stored. github.com/nathanielknight/llm-questioncache
Congressional Travel Explorer with Derek Willis
Просмотров 1139 часов назад
Derek Willis describes a project at the University of Maryland using datasette.io/ and AWS Textract and Claude to analyze congressional travel. cnsmaryland.org/interactives/fall-2024/congressional_travel_explorer/index.html
llm-consortium with Thomas Hughes
Просмотров 14612 часов назад
Thomas Hughes presents a collection of his plugins for llm.datasette.io/ - including llm-model-gateway and llm-consortium. github.com/irthomasthomas/llm-model-gateway adds an OpenAI-API-compatible web serving feature to LLM. github.com/irthomasthomas/llm-consortium uses multiple LLMs to collaboratively solve complex problems.
llm-logs-feedback with Matthias Lübken
Просмотров 19912 часов назад
llm-logs-feedback is a plugin by Matthias Lübken for llm.datasette.io/ which adds the ability to store feedback on prompt responses, using new "llm feedback 1" and "llm feedback-1" commands. These also accept an optional comment, and the feedback is stored in a "feedback" table in SQLite. github.com/luebken/llm-logs-feedback
Gemini 2.0 Flash multi-modal streaming demo
Просмотров 6 тыс.Месяц назад
Try this out at aistudio.google.com/live I wrote more about the new Gemini 2.0 Flash here: simonwillison.net/2024/Dec/11/gemini-2/
Civic Band, presented by Philip James during Datasette Public Office Hours, 15th November 2024
Просмотров 6952 месяца назад
Detailed notes: simonwillison.net/2024/Nov/16/civic-band/
VERDAD - tracking misinformation in radio broadcasts using Gemini 1.5
Просмотров 1,3 тыс.2 месяца назад
VERDAD - verdad.app/ - is a new project from Rajiv Sinclair and Public Data Works that aims to identify misinformation broadcast on US radio stations by archiving their audio, transcribing and translating it and hunting for potential misinformation topics using LLMs. In this interview we dive deep into how the project works and what they've learned from building it so far. 00:00 Introduction to...
“Teresa T” the juvenile humpback whale - in Pillar Point Harbor, Half Moon Bay
Просмотров 2,1 тыс.4 месяца назад
“Teresa T” the juvenile humpback whale - in Pillar Point Harbor, Half Moon Bay
Extracting unstructured text and images into database tables with GPT-4 Turbo and Datasette Extract
Просмотров 12 тыс.9 месяцев назад
Demonstrating datasette-extract, a new Datasette plugin that uses GPT-4 Turbo and GPT-4 Vision to extract structured data. github.com/datasette/datasette-extract datasette.io/ www.datasette.cloud/ The events table created in this video: simon.datasette.site/content/events
Datasette Enrichments
Просмотров 1,6 тыс.Год назад
More details here: simonwillison.net/2023/Dec/1/datasette-enrichments/
Embeddings: What they are and why they matter
Просмотров 26 тыс.Год назад
Extensive notes to accompany this talk: simonwillison.net/2023/Oct/23/embeddings/
When Zeppelins Ruled The Earth
Просмотров 2,7 тыс.Год назад
When Zeppelins Ruled The Earth
Prompt Injection, explained
Просмотров 20 тыс.Год назад
Prompt Injection, explained
Datasette ChatGPT Plugin
Просмотров 3 тыс.Год назад
Datasette ChatGPT Plugin
Bellingcat Hackathon: Action Transcription
Просмотров 4,3 тыс.2 года назад
Bellingcat Hackathon: Action Transcription
Datasette: a big bag of tricks for solving interesting problems using SQLite
Просмотров 3,1 тыс.2 года назад
Datasette: a big bag of tricks for solving interesting problems using SQLite
How to build, test and publish an open source Python library (without sign language)
Просмотров 7763 года назад
How to build, test and publish an open source Python library (without sign language)
Datasette Desktop initial demo
Просмотров 1,2 тыс.3 года назад
Datasette Desktop initial demo
Using Datasette with Jupyter to publish your data (JupyterCon 2020)
Просмотров 4793 года назад
Using Datasette with Jupyter to publish your data (JupyterCon 2020)
Datasette - an ecosystem of tools for working with small data
Просмотров 1,2 тыс.3 года назад
Datasette - an ecosystem of tools for working with small data
Joining CSV and JSON data using the "sqlite-utils memory" command
Просмотров 1,7 тыс.3 года назад
Joining CSV and JSON data using the "sqlite-utils memory" command
Django SQL Dashboard
Просмотров 3,1 тыс.3 года назад
Django SQL Dashboard
Git scraping: tracking changes to a scraped data source using GitHub Actions
Просмотров 5 тыс.3 года назад
Git scraping: tracking changes to a scraped data source using GitHub Actions
Introduction to Datasette and sqlite-utils
Просмотров 16 тыс.3 года назад
Introduction to Datasette and sqlite-utils
Barn the Spoon makes a wooden spoon at Monki Gras 2013
Просмотров 3,7 тыс.11 лет назад
Barn the Spoon makes a wooden spoon at Monki Gras 2013
Czech muscle-bus doing press-ups outside the Business Design Centre in Islington
Просмотров 1,8 тыс.12 лет назад
Czech muscle-bus doing press-ups outside the Business Design Centre in Islington
How to use OpenID
Просмотров 8 тыс.12 лет назад
How to use OpenID

Комментарии

  • @cal4
    @cal4 3 дня назад

    Thomas is doing a lot of interesting work. I appreciated catching this during the live cast.

  • @bernardcrnkovic3769
    @bernardcrnkovic3769 4 дня назад

    you are building amazing stuff, i am following the development very closely and using datasette myself. i am building a similar product myself but with event sourced sqlite instance. keep up the good work!

  • @jandy1
    @jandy1 23 дня назад

    Rum based old fashioned huh?

  • @manuelkoerner
    @manuelkoerner Месяц назад

    Nice

  • @rnalexander
    @rnalexander Месяц назад

    Nice selections of booze Simon! (the Rittenhouse 100 is one of my staple bottles.)

  • @pleka
    @pleka Месяц назад

    Amazing capabilities. Ask it to pronounce "daiquiri"... Dye-queer-eee

    • @MrNoipe
      @MrNoipe Месяц назад

      thats how it was pronounced at the company I worked at (daiquery)

  • @dun623
    @dun623 Месяц назад

    Thanks for this superb subject matter. I can really relate to a shelf of bottles ❤

  • @chrisogonas
    @chrisogonas Месяц назад

    Great resource! Thanks

  • @xnbet
    @xnbet Месяц назад

    💥💥💥

  • @gautame
    @gautame 2 месяца назад

    Brilliant stuff!

  • @user-kt1iz4vc3x
    @user-kt1iz4vc3x 2 месяца назад

    This guy’s like Jack Black meets Nicholas Hoult.

    • @kurnaikent
      @kurnaikent 29 дней назад

      I love his enthusiasm at 16:42! He actually has really great charisma.

  • @sadburger
    @sadburger 2 месяца назад

    This was a really nice interview and interesting project. It’s incredible the superpowers that we developers have gained over the last two years. Things that you could’ve asked for 10 years ago and I would’ve said maybe with a year and a few million dollars worth of headcount are now an API call away. I have LLM‘s integrated into nearly every part of my workflow and my tooling. The way I work now looks almost nothing like the way it used to. I want to know more about the price difference with Gemini flash versus Whisper for transcription particularly with all the many flavors of local whisper that are available. I’ll have to do some research on this.

    • @swillison
      @swillison 2 месяца назад

      OpenAI charge $0.006 / minute for their Whisper API - so an hour of audio would cost 36 cents. Gemini 1.5 Flash is $0.075 for 1 million tokens and every second of audio is charged as 25 tokens, which means an hour is 90,000 tokens and hence costs just 0.675 cents - so it's over 50x cheaper!

    • @ftk525
      @ftk525 2 месяца назад

      @@swillison If you use GPU spot instances yourself you can run whisper large v3 turbo at about a penny per hour. Since this project only requires timestamping, and appears to have a high tolerance for timestamps not being exactly accurate, I would think your guest would be well served with just whisper tiny, which you can run at roughly 10x on a single CPU - basically free.

  • @scottieapplseed
    @scottieapplseed 3 месяца назад

    Fantastic tool and fun examples that actually demonstrate fun little use cases.

  • @arpitgarg5172
    @arpitgarg5172 3 месяца назад

    Is there a place where one could explore all the published datasettes?

  • @andrewrecchia4103
    @andrewrecchia4103 5 месяцев назад

    ruclips.net/video/CQbkhYg2DzM/видео.htmlsi=bBBGXe5F6RoeMtS_

  • @andrewrecchia4103
    @andrewrecchia4103 5 месяцев назад

    三 1

  • @mikecourian
    @mikecourian 5 месяцев назад

    Amazing work Simon! Thank you!

  • @alokranjan
    @alokranjan 5 месяцев назад

    Can i use this Plugin along with MySQL .. getting errors.

  • @MatthewTerry-suade
    @MatthewTerry-suade 6 месяцев назад

    Thank you, very informative

  • @bbcc2960
    @bbcc2960 6 месяцев назад

    Thank you for sharing.

  • @energyexecs
    @energyexecs 6 месяцев назад

    Thank you Simon Wilson. Great information. I especially like how you demonstrated the development of your own tools. Finally my thoughts is your presentation in an executive summary format will educate policy makers in both the enterprise and government sector who seem to have fear of AI. For example my company has an existing early policy that employees are not allowed to use AI or ChatGPT. At the same time my Use Case to leverage RAG was to augment our LLM was accepted by our AI Review Committee. My thought is the enterprise companies will be careful and prudent in the rollout of LLMs and AI tools because they will want “security rails” in place. Thank you.

  • @canadianrepublican1185
    @canadianrepublican1185 6 месяцев назад

    Thank you .

  • @schalkdormehl3057
    @schalkdormehl3057 7 месяцев назад

    Mask, in 2023...

  • @Tony_Indiana
    @Tony_Indiana 7 месяцев назад

    I just took a poll. And people said if you could show us OSINT using a model like mistral that is mostly uncensored (Dolphin/Instruct) or whatever your preference is then gpt4. Everyone who responded agreed that would be something we would pay for. 2024 tips and tricks LLMs and OSINT. But there are advantages to uncensored.

  • @tutacat
    @tutacat 7 месяцев назад

    This is the best fundamental way of describing embeddings.

  • @tutacat
    @tutacat 7 месяцев назад

    This is what microsoft recall wants to do

  • @tutacat
    @tutacat 7 месяцев назад

    This man is truly based.

  • @codenocode
    @codenocode 7 месяцев назад

    this is really nice! thanks for sharing.

  • @Clammer999
    @Clammer999 8 месяцев назад

    I’m totally new to embeddings and this video inspired me to want learn even more!

  • @codenocode
    @codenocode 8 месяцев назад

    I've recently stumbled across your work which I read about in Gergely's book "Software Engineering Guidebook". Fantastic find. Love the creativity here.

  • @enigmeta
    @enigmeta 9 месяцев назад

    Love this! Would be useful to mention you need to run datasette in --root mode in order to make modifications, it took me a while to find this.

  • @Speejays2
    @Speejays2 9 месяцев назад

    Is it possible to replace the OpenAI API key with local vision model instead?

  • @monKeman495
    @monKeman495 9 месяцев назад

    there should be some kind of authorized base restriction on internal llm tokens to normal public

  • @_ramen
    @_ramen 9 месяцев назад

    very great demo, thanks for sharing! this is an excellent example of practical use of embeddings.

  • @brcosmin
    @brcosmin 9 месяцев назад

    Thanks for linking yourself on ycombinator, very interesting talk and quite engaging delivery.

  • @QINGCHARLES
    @QINGCHARLES 9 месяцев назад

    The future is wild. Imagine how good this will be 6 months or a year from now.

  • @MichelBinkhorst
    @MichelBinkhorst 9 месяцев назад

    New to Datasette. Just installed it on OSX with Homebrew, and added the Extract plugin, but I'm not seeing the 'database actions' button. Am I missing something?

    • @jmottishaw
      @jmottishaw 9 месяцев назад

      same here on Windows in a fresh venv

  • @AP-hv5dh
    @AP-hv5dh 9 месяцев назад

    🔥

  • @subinalex88
    @subinalex88 9 месяцев назад

    Nice

  • @ecosse64
    @ecosse64 9 месяцев назад

    That's fantastic. Does it work across multiple websites and in different languages? For example, if you wanted to provide a list of specific events in a country where both English and Spanish or Italian are spoken but have a single database in English.

  • @kai.diefenbach
    @kai.diefenbach 9 месяцев назад

    Awesome!

  • @anne-marieroy8812
    @anne-marieroy8812 9 месяцев назад

    Thanks very interesting and useful.

  • @sebastianwagner5843
    @sebastianwagner5843 9 месяцев назад

    Things start to become magical.

  • @muddasirkhan805
    @muddasirkhan805 10 месяцев назад

    This was so good! Please do more of these - i am still in awe!! Thank you!

  • @zgintasz2
    @zgintasz2 10 месяцев назад

    "vibes-based search" lol. love the term you invented.

  • @korolyovPavel
    @korolyovPavel 11 месяцев назад

    Cool