Simon Willison @simonw Twitter profile | Pikagi

Pikagi

Simon Willison

@simonw

71,072

Followers

5,468

Following

2,950

Media

50,662

Statuses

Creator @datasetteproj , co-creator Django. PSF board. @nichemuseums . Hangs out with @natbat + @cleopaws . He/Him. Mastodon:

San Francisco, CA

https://t.co/p4R0XiEYEc

Joined November 2006

Don't wanna be here? Send us removal request.

Pinned Tweet

@simonw

Simon Willison

24 days

I gave a 50m talk at the Story Discovery at Scale data journalism conference at Stanford a few weeks ago. The video is now out, and I've written an extensive annotated version AI for Data Journalism: demonstrating what we can do with this stuff right now

Tweet card media

AI for Data Journalism: demonstrating what we can do with this stuff right now

I gave a talk last month at the Story Discovery at Scale data journalism conference hosted at Stanford by Big Local News. My brief was to go deep into the …

simonwillison.net

7

35

166

Last Seen Profiles

@parkerlendrum22

@sm5gp

@T3RMm2KevGqnfKb

@ramonorteg

@mr_sahil110

@MusaSeriki2

@SSeawards

@Remajobe11

@Nad_ryan

@inclusionafric2

@extrimrace

@eilishdata

@gohhshurricanes

@luvstrizzy

@GolDeVicky

@kotobacchi

@kkyukirby

@IRC_NIC

@MSPBJnews

@nicolawalker_

@vincxmx

@maxiikits

@Swrve_Inc

@history_alice

@honami_DM

@dmandisi

@DivineSorcerer_

@ctewc

@RollFUNGO

@AngelicaNoboaP

@amitisinvesting

@princessashleiv

@KashviGupta

@LoveNb8c

@celestisflights

@UNLincoln

@simonw

Simon Willison

2 years

New hobby: prototyping video games in 60 seconds using a combination of GPT-3 and DALL-E Here's "Raccoon Heist"

Tweet media one

Tweet media two

119

1K

8K

@simonw

Simon Willison

2 years

If someone gives you a CSV file with 100,000 rows in it, what tools do you use to start exploring and understanding that data?

2K

894

7K

@simonw

Simon Willison

3 years

Here's a piece of information that will send a chill down the spine of anyone who's ever designed a database schema: Our new house that we just moved into... has two zip codes!

273

760

6K

@simonw

Simon Willison

1 year

Leaked Google document: “We Have No Moat, And Neither Does OpenAI” The most interesting thing I've read recently about LLMs - a purportedly leaked document from a researcher at Google talking about the huge strategic impact open source models are having

126

1K

5K

@simonw

Simon Willison

2 years

TIL you can run SQL queries directly against CSV files as a one-liner using the default sqlite3 command line utility

Tweet media one

57

779

5K

@simonw

Simon Willison

3 years

"Hosting SQLite databases on Github Pages" is absolutely brilliant: it adds a virtual filesystem to SQLite-compiled-to-WebAssembly in order to fetch pages from the database using HTTP range requests

46

966

4K

@simonw

Simon Willison

1 year

If you're just starting to learn software engineering right now but you're considering dropping it because you think the field might be made obsolete by AI, I have an alternative approach to suggest for you: Start learning now, and use AI tools to learn FASTER

69

367

3K

@simonw

Simon Willison

16 days

"Do stuff and then blog about it" remains one of the most underrated pieces of career advice

@vboykis

vicki

17 days

An absolutely fantastic way to increase this is to start a blog. Almost all the cool fun stuff in my professional life for me has come from doing stuff then blogging about it.

11

141

1K

25

319

3K

@simonw

Simon Willison

5 months

@d_feldman It's got me to the point where I can read Spanish language news articles and understand ~80% of them - spoken Spanish is much harder

25

30

3K

@simonw

Simon Willison

1 year

@tomgara That is the most "Society of Human Resource Management" story I could possibly imagine

5

37

3K

@simonw

Simon Willison

6 years

How about if, instead of ditching Twitter for Mastodon, we all start blogging and subscribing to each other's Atom feeds again instead? The original distributed social network could still work pretty well if we actually start using it

102

644

2K

@simonw

Simon Willison

6 months

Found the system prompt that drives this thing here: It works by generating a base64 encoded PNG of the drawn components, then passing that to GPT-4 Vision with that system prompt and instructions to "Turn this into a single html file using tailwind"

@tldraw

tldraw

6 months

let's go

148

1K

8K

23

244

2K

@simonw

Simon Willison

1 year

OK, I'm calling it: Large language models are having their Stable Diffusion moment right now

Tweet card media

Large language models are having their Stable Diffusion moment

The open release of the Stable Diffusion image generation model back in August 2022 was a key moment. I wrote how Stable Diffusion is a really big deal at the …

simonwillison.net

45

426

2K

@simonw

Simon Willison

5 months

I tried this out yesterday and it's incredible: download a 4GB binary, child 755 it and now you have a full LLM and the software needed to run it ready to go, with multiplied operating system platforms supported by that single file

@JustineTunney

Justine Tunney

5 months

I spent the last month building llamafile which is the fattest executable file format ever. It lets you turn LLM weights into runnable llama.cpp binaries using cosmo libc. Blog post:

49

449

3K

15

190

2K

@simonw

Simon Willison

3 months

I tried feeding a 7s video of my bookshelf into Gemini Pro 1.5 to get back a JSON array of books... and it worked!

Tweet card media

The killer app of Gemini Pro 1.5 is video

Last week Google introduced Gemini Pro 1.5, an enormous upgrade to their Gemini series of AI models. Gemini Pro 1.5 has a 1,000,000 token context size. This is huge—previously that …

simonwillison.net

128

222

2K

@simonw

Simon Willison

1 month

I built a new tool! It's a single page web app that runs OCR against images and PDFs entirely in your browser (no file upload needed) using Tesseract.js and PDF.js You can drop files onto it, or you can click to select and open them (which works on Mobile Safari as well)

36

230

2K

@simonw

Simon Willison

1 year

MiniGPT-4 is pretty astonishing: an AI chatbot you can use to ask questions about an image (a feature that's been promised but not yet shipped by GPT-4), building on top of the Vicuna-13B LLM (derived from LLaMA) and BLIP-2 vision-language model

Tweet media one

36

238

2K

@simonw

Simon Willison

2 years

This is really grim, if not entirely unexpected: apparently the Instagram mobile app injects additional JavaScript into every page that's loaded using the in-app embedded browser - here's the tool @KrauseFx built to track changes made to the DOM when loading a page

Tweet media one

@KrauseFx

Felix Krause

2 years

💥 New Post: Instagram & Facebook tracks everything you do on any website in their in-app browser

Tweet media one

95

2K

5K

26

695

2K

@simonw

Simon Willison

2 months

TIL about binary vector search... apparently there's a trick where you can take an embedding vector like [0.0051, 0.017, -0.0186, -0.0185...] and turn that into a binary vector just reflecting if each value is > 0 - so [1, 1, -1, -1, ...] and still get useful cosine similarities!

68

158

2K

@simonw

Simon Willison

3 years

GitHub issues tip: if you paste in a link to an issue or PR in another repo it will display it as a truncated URL, but if you instead add it in a hyphenated bullet point it will display the title of the issue and and indicate if it is open or closed

11

324

2K

@simonw

Simon Willison

3 months

AI may enable anyone to produce code, but that's not the same thing as enabling anyone to develop software The typing-out-code bit is one of the least challenging parts of building useful software that solves real problems

@Carnage4Life

Dare Obasanjo🐀

3 months

Jensen Huang, CEO of Nvidia, argues that we should stop saying kids should learn to code. He argues the rise of AI means we can replace programming languages with human language prompts thus enabling everyone to be a programmer. AI will kill coding.

1K

5K

21K

124

165

2K

@simonw

Simon Willison

2 years

Whoa. runs a full Debian VM entirely in your browser via WebAssembly... and it ships with working Perl, Python, Ruby and Node.js!

Tweet card media

WebVM - Linux virtualization in WebAssembly

Server-less virtual machine, networking included, running browser-side in HTML5/WebAssembly. Code in any programming language inside this Linux terminal.

@leaningtech

Leaning Technologies

2 years

We have made a server-less virtual Linux environment that runs unmodified Debian binaries in the browser. This is powered by CheerpX, a WebAssembly virtualization platform. Feel free to play with it and report bugs:

Tweet media one

16

151

488

26

394

2K

@simonw

Simon Willison

1 year

I expect GPT-4 will have a LOT of applications in web scraping The increased 32,000 token limit will be large enough to send it the full DOM of most pages, serialized to HTML - then ask questions to extract data

60

104

1K

@simonw

Simon Willison

2 years

Reddit conversation about using GPT-3 to write your homework. A teacher comments: "Grading something an AI wrote is an incredibly depressing waste of my life."

61

188

1K

@simonw

Simon Willison

4 years

SQL is a better API language than GraphQL. Convince me otherwise!

128

174

1K

@simonw

Simon Willison

2 years

Fascinating HN comment from someone who's company built a custom distributed data warehouse using compressed SQLite DB files in S3 that were queried using Lambda functions orchestrated by PostgreSQL running a custom foreign data wrapper

Tweet media one

23

210

1K

@simonw

Simon Willison

1 year

Notes on how I ran Facebook's 7B LLaMA model on my 64GB M2 MacBook Pro using llama.cpp by @ggerganov It's genuinely possible to run a LLM that's hinting towards the performance of GPT3 on your own hardware now! I thought that was still a few years away

Tweet card media

Running LLaMA 7B and 13B on a 64GB M2 MacBook Pro with llama.cpp

See also: Large language models are having their Stable Diffusion moment right now .

til.simonwillison.net

41

239

1K

@simonw

Simon Willison

4 years

15 years ago today on my blog: Introducing Django

Introducing Django

You may know that I spent a year working in Kansas for a local newspaper—the Lawrence Journal-World. I’m delighted to announce that a decent chunk of the software I worked …

simonwillison.net

25

184

1K

@simonw

Simon Willison

11 months

Understanding GPT tokenizers: I wrote about how the tokenizers used by the various GPT models actually work, including an interactive tool for experimenting with their output

Tweet card media

Understanding GPT tokenizers

Large language models such as GPT-3/4, LLaMA and PaLM work in terms of tokens. They take text, convert it into tokens (integers), then predict which tokens should come next. Playing …

simonwillison.net

24

256

1K

@simonw

Simon Willison

3 years

A lesson I re-learn on every project: always have an automatically populated "created_at" column on every single database table. Any time you think "I won't need it here" you're guaranteed to want to use it for debugging something a few weeks later.

46

131

1K

@simonw

Simon Willison

2 years

The stuff happening on the Stable Diffusion subreddit right now is pretty wild - since the model can be run by anyone on their own machine if they have a decent GPU

25

250

1K

@simonw

Simon Willison

7 months

Prompt injection comes to GPT-4V

@mn_google

Patel Meet 𝕏

7 months

In GPT-4V Image content can override your prompt and be interpreted as commands.

Tweet media one

18

44

247

23

120

1K

@simonw

Simon Willison

2 years

@kingrolo @gvanrossum @tech_update0 @ethereum @solana We hate it because we understand it

16

60

1K

@simonw

Simon Willison

1 year

We accidentally invented computers that can lie to us and we can't figure out how to make them stop

83

264

1K

@simonw

Simon Willison

3 years

PostgreSQL 14 adds new syntax for accessing JSON data: SELECT * FROM shirts WHERE details['attributes']['color'] = '"neon yellow"' I like this so much more than the -> operators, which stubbornly refuse to stick in my head

@craigkerstiens

Craig Kerstiens

@craigkerstiens

3 years

Y'all this new JSON subscript syntax in Postgres 14 is sweet. Super excited to see Postgres just getting better bit by bit -

7

105

415

26

251

1K

@simonw

Simon Willison

5 months

llamafile is the new best way to run a LLM on your own computer

Tweet card media

llamafile is the new best way to run a LLM on your own computer

Mozilla’s innovation group and Justine Tunney just released llamafile, and I think it’s now the single best way to get started running Large Language Models (think your own local copy …

simonwillison.net

28

195

1K

@simonw

Simon Willison

7 months

If you haven't tried Claude yet it's a absolutely worth spending time with - I lean on it a lot for working with longer documents, since it can handle 100,000 tokens (GPT-4 is only 8,000) at a time Plus you can upload PDFs to it - I've used it with 100+ page documents

@AnthropicAI

Anthropic

7 months

We’re rolling out access to to more people around the world. Starting today, users in 95 countries can talk to Claude and get help with their professional or day-to-day tasks. You can find the list of supported countries here:

158

362

2K

57

94

1K

@simonw

Simon Willison

3 years

We really need to start teaching web developers how to use links

25

202

1K

@simonw

Simon Willison

1 year

A new post about prompt injection attacks, which I'm increasingly concerned about now that people are hooking LLMs up to external tools through Auto-GPT, ChatGPT Plugins etc

Tweet card media

Prompt injection: What’s the worst that can happen?

Activity around building sophisticated applications on top of LLMs (Large Language Models) such as GPT-3/4/ChatGPT/etc is growing like wildfire right now. Many of these applications are potentially...

simonwillison.net

27

250

1K

@simonw

Simon Willison

2 years

Not sure if this is a controversial opinion or not: unit tests should make up a minority segment of your overall automated test suite I'd absolutely take a project with integration and no unit tests over one with unit tests but no integration tests

130

121

1K

@simonw

Simon Willison

1 year

I see people being deceived by this again and again: ChatGPT can NOT read content from URLs that you give it, but will convincingly pretend that it can Crucial to spread this message any time you see anyone falling into this trap

Tweet card media

ChatGPT can’t access the internet, even though it really looks like it can

A really common misconception about ChatGPT is that it can access URLs. I’ve seen many different examples of people pasting in a URL and asking for a summary, or asking …

simonwillison.net

78

245

1K

@simonw

Simon Willison

2 years

"GitLab plans to automatically delete projects if they've been inactive for a year and are owned by users of its free tier." Absolutely shocking decision from @gitlab , I very much hope they reconsider this

35

216

1K

@simonw

Simon Willison

3 years

My favourite reactions to this come from people who work in civic tech, because unlike regular corporate gig programmers they're not allowed to just ignore weird edge-cases like this

@cydharrell

Cyd Harrell

3 years

oh no

7

23

258

16

56

1K

@simonw

Simon Willison

1 year

@emnode @tobyordoxford I don't want my search engine to be vengeful

20

13

1K

@simonw

Simon Willison

2 years

Love this idea that the reason voice assistants don't seem to stick for most people is that they're actually command line interfaces, but even less discoverable because they don't provide any visible feedback at all

@edent

Terence Eden is on Mastodon

2 years

@daviddlow @charlesarthur @benedictevans I've droned on endlessly about how you can't expect normal people to use the command line. That's what Alexa is. If you don't say the *precise* invocation correctly, you get an error. And because there's no display, you have to remember dozens of different commands. It's too hard

10

30

306

33

249

1K

@simonw

Simon Willison

5 months

Love that we live in a time where "your software got lazier" is a legit piece of feedback

@ChatGPTapp

ChatGPT

5 months

we've heard all your feedback about GPT4 getting lazier! we haven't updated the model since Nov 11th, and this certainly isn't intentional. model behavior can be unpredictable, and we're looking into fixing it 🫡

894

665

9K

20

82

1K

@simonw

Simon Willison

1 year

If you're a programmer and you're still thinking that all of this ChatGPT stuff is a waste of your time, I strongly suggest reviewing this example It's over-hyped, sure - but it's not something anyone in our profession should continue to ignore

@simonw

Simon Willison

1 year

This entire benchmarking project took just three prompts

Tweet media one

Tweet media two

Tweet media three

10

32

352

37

127

1K

@simonw

Simon Willison

5 years

Sentences with the word "just" in them always work better if you drop that word. "Why don't you just add caching?" - that one word implies "I don't value your expertise or expect you to have thought this through" "Why don't you add caching?" - now we can have a conversation.

47

226

1K

@simonw

Simon Willison

2 months

Just got ChatGPT Code Interpreter to write me a SQLite extension in C from scratch, then compile it with GCC, then load it into Python and test it out, then iterate on it to fix the bugs All on my phone while pottering around the house

29

72

1K

@simonw

Simon Willison

4 years

Made myself a self-updating GitHub personal README! It uses a GitHub Action to update itself with my latest GitHub releases, blog entries and TILs

Tweet media one

22

134

1K

@simonw

Simon Willison

6 months

Looks like the reason that letter only had 500 signatures out of 770 might be that the rest of the company were asleep

@lilianweng

Lilian Weng

6 months

About 650 / 770 signed at this moment. As people start waking up, more will come. All the efforts started after 1:30 AM, 500+ within two hours and all of this after 2 crazy days with very little sleep.

161

605

5K

16

54

983

@simonw

Simon Willison

6 months

Sigh. Tip for if you're planning on suing an AI company: asking a model if something is included in its training data is not a reliable way method for telling what is in its training data

@minimaxir

Max Woolf

6 months

ಠ_ಠ

Tweet media one

39

14

287

27

49

974

@simonw

Simon Willison

2 months

TIL Google Chrome has a --headless option you can use to take a screenshot from the CLI that's built into the default installation

Tweet media one

21

94

965

@simonw

Simon Willison

6 years

I just released datasette - a new tool for turning any SQLite database into a web interface and JSON API:

Tweet media one

22

308

946

@simonw

Simon Willison

6 years

Maybe the solution to the Fermi paradox is that significantly advanced civilizations discover crypto currencies and then furiously burn through all available energy sources until they go extinct

@EricHolthaus

Eric Holthaus

6 years

Uhhh... about bitcoin... it's actually ruining the planet. The bitcoin computer network currently uses as much electricity as Denmark. In 18 months, it will use as much as the entire United States. Something's gotta give. This simply can’t continue.

732

11K

16K

29

484

935

@simonw

Simon Willison

10 months

Has anyone managed to run Llama 2 GPU accelerated on an M1/M2 Mac yet? Bonus points if you can provide extremely meticulous step-by-step instructions for replicating what you did!

26

55

955

@simonw

Simon Willison

2 years

@moorehn @dancow @sewellchan @guardian I'm fascinated by their use of the term "build" - they talk about building a lot, took me a while to realize that their version of building is funneling money into speculative investments and convincing others to do the same

20

28

930

@simonw

Simon Willison

24 days

What's the cheapest option right now for me to spin up a Linux server somewhere for an hour with enough GPU to run the latest Mixtral model?

111

63

952

@simonw

Simon Willison

3 years

One of the most obvious flaws in using blockchains for anything involving regular human beings is one I've not seen much discussion of: Regular human beings cannot protect their passwords, credentials or private keys. They just can't.

52

155

921

@simonw

Simon Willison

2 years

Weird, GitHub have deprecated one of their GraphQL APIs in favor of a REST one

Tweet media one

32

129

905

@simonw

Simon Willison

7 months

Embeddings: What they are and why they matter I took my recent PyBay talk and turned it into the most comprehensive answer I could possibly provide to the question "What are embeddings?"

Tweet card media

Embeddings: What they are and why they matter

Embeddings are a really neat trick that often come wrapped in a pile of intimidating jargon. If you can make it through that jargon, they unlock powerful and exciting techniques …

simonwillison.net

21

133

894

@simonw

Simon Willison

2 years

I built a new tool: s3-ocr, a utility for running OCR (with Amazon Textract) against every PDF file in a S3 bucket and getting the results back as a searchable SQLite database

Tweet card media

s3-ocr: Extract text from PDF files stored in an S3 bucket

I’ve released s3-ocr, a new tool that runs Amazon’s Textract OCR text extraction against PDF files in an S3 bucket, then writes the resulting text out to a SQLite database …

simonwillison.net

22

113

848

@simonw

Simon Willison

2 years

My own answer: I either open the CSV directly in the Datasette Desktop Mac application () or I do this: sqlite-utils insert /tmp/data.db rows big.csv --csv datasette /tmp/data.db That gives me a table called "rows" in a fresh SQLite database

13

41

849

@simonw

Simon Willison

4 years

This video is a great example of why I just can't agree with historians who are pushing back at efforts to use AI to colorize and up-rez historic videos - it's honestly like a magic portal window back to 124 years ago

Tweet card media

YouTubers are upscaling the past to 4K. Historians want them to stop

YouTubers are using AI to bring history to life. But historians argue the process is nonsense

14

144

829

@simonw

Simon Willison

1 year

Notes on how I turned an hour-long video of a digital thermometer into temperature readings over time by using ffmpeg to split the video into a frame every 10s and Google Cloud Vision to run the OCR I got ChatGPT/GPT-4 to show me how to use both of those:

Tweet media one

27

87

840

@simonw

Simon Willison

1 year

This is the thing I find most interesting about ChatGPT as a learning assistant: it turns out having a teacher that's mostly right but occasionally very wrong makes me think much more critically about what I'm learning, which I think is helping me build more robust mental models

@random_walker

Arvind Narayanan

1 year

When ChatGPT came out I thought I wouldn't use it for learning because of its tendency to slip in some BS among 10 helpful explanations. Then I tried it, and found that it forces me to think critically about every sentence, which is the most effective mindset for learning.

40

189

2K

43

100

825

@simonw

Simon Willison

6 years

Couldn't agree with this more: I Google the most trivial code things in languages that I've used for over a decade dozens of times a day. The amount of detail you actually need to commit to memory in 2018 keeps getting smaller - remember what CAN be done, not exactly how to do it

@patio11

Patrick McKenzie

6 years

Periodic observation for the benefit of junior developers: You do not have to be embarrassed about not knowing a particular bit of syntax or API. Googling things efficiently is a core job skill. ~15 years in I'll still look up "append to array javascript."

70

766

3K

20

303

817

@simonw

Simon Willison

3 years

Amusingly I first learned this when someone sent us an encrypted PDF where the password was our zip code, and what I thought was the zip code didn't work so I ran a brute force password cracker against it

5

25

818

@simonw

Simon Willison

2 years

This is a crazy cool hack: Crunchy Data have a new PostgreSQL tutorial series which runs a full PostgreSQL server compiled to WebAssembly entirely in your web browser so you can try things out!

Tweet media one

@craigkerstiens

Craig Kerstiens

@craigkerstiens

2 years

And here it is, come join us @crunchydata and learn some Postgres at our playground -

19

75

293

10

192

822

@simonw

Simon Willison

18 days

Realized I've reached the point now with GitHub Copilot autocomplete where I can often guess exactly what it's going to suggest, pause for a moment to wait for it to make the suggestion and accept it and move on It's majority a typing assistant now and I really like it for that

42

30

822

@simonw

Simon Willison

1 year

Good thing my blog is behind Cloudflare

@elonmusk

Elon Musk

1 year

Might need a bit more polish …

4K

4K

40K

34

16

804

@simonw

Simon Willison

6 months

One thing I missed from yesterday: the cost of storing data in an OpenAI assistant for retrieval question answering etc is VERY steep: $0.20/GB/assistant/day Compare to S3 standard pricing which is about $0.023 per GB per month, so I think ChatGPT assistant storage is 260x more!

39

62

809

@simonw

Simon Willison

4 months

What are some LLM-driven products that you use at least once a week that aren't ChatGPT, Bard, Bing or GitHub Copilot - things that are built on LLMs but aren't direct chat interfaces to LLMs

206

47

793

@simonw

Simon Willison

2 years

A tip for writing more: expand your definition of completing a project (any project, no matter how small) to include writing a blog post (or README or similar) that explains that project

23

115

784

@simonw

Simon Willison

3 years

This is fascinating: as-of 2017 university instructors have been increasingly encountering students who have absolutely no idea how files and folders on a computer work

@dancow

Dan Nguyen

3 years

omg turns out I wasn't completely bullshitting all the times when I fussed to students that "file not found" would be their most common and soul-destroying "bug" when trying to learn programming

Tweet media one

26

138

587

43

247

785

@simonw

Simon Willison

3 years

Also fun: Instacart refuse to deliver to one of the zip codes, but we tried the other one and successfully placed an order

8

9

780

@simonw

Simon Willison

2 years

A public librarian on why two-factor authentication is a huge barrier to people with limited technology access

@heyakwan

Aydin Kwan

2 years

The assumption that people will have consistent access to the same mobile number simply isn’t true for a lot of people. Phones cost money. So do phone plans. Phones break or get stolen. We all know this. Yet people who don’t have phones still need access to their email accounts.

20

269

2K

10

341

777

@simonw

Simon Willison

7 months

New release of llm-gpt4all - it's now really easy to run models like mistral-7b-instruct-v0 on your own machine: pip install llm llm install llm-gpt4all llm -m mistral-7b-instruct-v0 "ten facts about pelicans"

Tweet card media

Release 0.2 · simonw/llm-gpt4all

Upgrade for compatibility with the latest gpt4all. #16 The new version of gpt4all uses models in the gguf format. You may have existing ggml models - these will no longer be used by the plugin, so...

22

110

782

@simonw

Simon Willison

2 months

@AdamRackis Engineers who understand web fundamentals are vastly more useful and productive than engineers who can only work with whatever the latest popular framework is

27

15

780

@simonw

Simon Willison

6 months

I had not realized quite how easy voice cloning is now. I used the free demo on just now (also used for this Attenborough demo) and got a passable imitation of my own voice on first try, from a poor quality 10m audio sample I had lying around already

Tweet card media

AI Voice Generator & Text to Speech

Rated the best text to speech (TTS) software online. Create premium AI voices for free and generate text-to-speech voiceovers in minutes with our character AI voice generator. Use free text to speech...

@charliebholtz

Charlie Holtz

6 months

David Attenborough is now narrating my life Here's a GPT-4-vision + @elevenlabsio python script so you can star in your own Planet Earth:

728

5K

27K

18

57

777

@simonw

Simon Willison

9 months

The video for my North Bay Python talk is out, and I've put together an accompanying edited transcript with annotated slides and links If you haven't been completely immersed in this world for the last year, my hope is this can help catch you up!

Tweet card media

Catching up on the weird world of LLMs

I gave a talk on Sunday at North Bay Python where I attempted to summarize the last few years of development in the space of LLMs—Large Language Models, the technology …

simonwillison.net

20

154

772

@simonw

Simon Willison

6 months

This is really worth your time - a very solid technical introduction to LLMs, great if you've not been paying close attention but I picked up quite a few useful details from it too

@karpathy

Andrej Karpathy

6 months

New YouTube video: 1hr general-audience introduction to Large Language Models Based on a 30min talk I gave recently; It tries to be non-technical intro, covers mental models for LLM inference, training, finetuning, the emerging LLM OS and LLM Security.

Tweet media one

585

3K

18K

16

62

767

@simonw

Simon Willison

5 years

@PitHoyaFan88 @drvox Absolutely this - the mythology of Person of Interest kept getting deeper and more interesting and they nailed the ending. Recent developments in AI have made it even more relevant - the tech it imagined feels more and more realistic every year

7

39

732

@simonw

Simon Willison

7 months

Frustrating limitation from Claude by @AnthropicAI - I uploaded a 30 page PDF and asked it "What are the key findings in this paper? Illustrate with quotes" And it refused! "I cannot provide lengthy excerpts or summaries from the copyrighted material you shared."

Tweet media one

62

52

754

@simonw

Simon Willison

1 year

Some wild speculation here, but I think it might be possible to train a LLaMA 7B sized model for $85,000 now, and maybe run that model directly in your web browser - with more capabilities than ChatGPT, through hooking up extra tools to it (like Bing)

Tweet card media

Could you train a ChatGPT-beating model for $85,000 and run it in a browser?

I think it’s now possible to train a large language model with similar functionality to GPT-3 for $85,000. And I think we might soon be able to run the resulting …

simonwillison.net

25

101

751

@simonw

Simon Willison

3 years

"Cryptocurrency is one of the worst inventions of the 21st century. I am ashamed to share an industry with this exploitative grift. It has failed to be a useful currency, invented a new class of internet abuse, further enriched the rich, wasted staggering amounts of electricity,

@Rich_Harris

Rich Harris

3 years

"If not for cryptocurrency, these services would still be available."

29

336

911

12

311

735

@simonw

Simon Willison

27 days

Have any of the large scale LLM training organizations - Anthropic, OpenAI, Gemini, Mistral, the Llama team - published anything notable about this idea of "model collapse" yet - the worry that LLM quality will drop as their training data becomes pollinated by model output?

68

62

744

@simonw

Simon Willison

2 years

DALL-E mini has renamed itself to Craiyon, in response to a request from OpenAI. You can use it for free (with no waiting list) at

Tweet card media

Craiyon - Your FREE AI image generator tool: Create AI art!

Craiyon is an AI model that can draw images from any text prompt!

www.craiyon.com

15

166

718

@simonw

Simon Willison

2 months

Well this @OpenAI refusal is new and I don't like it at all

Tweet media one

86

36

711

@simonw

Simon Willison

2 years

Want to know the secret to blogging more often? Lower your standards! A post which you don't think is ready yet is a LOT better than a giant folder full of drafts that no-one ever gets to see (Your readers won't ever know how good the thing you wanted to write would have been)

33

94

703

@simonw

Simon Willison

2 years

I'm finding Mastodon a lot easier to understand having I've realized that it's just blogs. Everyone gets a little blog, on their own server or someone else's. Following someone is pretty much subscribing to their feed. You can even roll your own implementation from scratch.

17

144

697

@simonw

Simon Willison

1 year

Made some quick notes on how to use the OpenAI Python library to make ChatGPT API calls and stream out the response tokens as they arrive

10

65

688

@simonw

Simon Willison

1 year

Got access to Google Bard! I'm pleased to report that it is ethically opposed to necromancy

Tweet media one

37

76

684

@simonw

Simon Willison

9 months

This talk is the most complete version of my thinking around Large Language Models to date, including thoughts on personal AI ethics, practical applications and the enormous impact we are already feeling from Llama 2 Video, slides, transcript & links:

Tweet card media

Making Large Language Models work for you

I gave an invited keynote at WordCamp 2023 in National Harbor, Maryland on Friday. I was invited to provide a practical take on Large Language Models: what they are, how …

simonwillison.net

10

115

685

@simonw

Simon Willison

2 months

The GPT-4 barrier has finally been smashed

The GPT-4 barrier has finally been broken

Four weeks ago, GPT-4 remained the undisputed champion: consistently at the top of every key benchmark, but more importantly the clear winner in terms of “vibes”. Almost everyone investing serious …

simonwillison.net

29

96

682

@simonw

Simon Willison

7 years

Accidentally deleted my Python source code but it was still resident in a running process - here's how I got it back

Tweet card media

How to recover lost Python source code if it's still resident in-memory

How to recover lost Python source code if it's still resident in-memory - recover_source_code.md

gist.github.com

13

415

668

@simonw

Simon Willison

4 years

Anyone else getting tired of meticulously erasing ?s=21 several times a day?

20

31

675

@simonw

Simon Willison

5 months

This Q story is deeply concerning - if it's true that Q has access to private data like the location of AWS data centers that would suggest the team working on it have not been taking things like prompt injection attacks seriously at all

@emollick

Ethan Mollick

5 months

I know I say it a lot, but using LLMs to build customer service bots with RAG access to your data is not the low-hanging fruit it seems to be. It is, in fact, right in the weak spot of current LLMs - you risk both hallucinations & data exfiltration.

47

250

2K

21

77

680

@simonw

Simon Willison

1 year

Web LLM runs the vicuna-7b Large Language Model entirely in your browser, and it’s very impressive

Tweet card media

Web LLM runs the vicuna-7b Large Language Model entirely in your browser, and it’s very impressive

A month ago I asked Could you train a ChatGPT-beating model for $85,000 and run it in a browser?. $85,000 was a hypothetical training cost for LLaMA 7B plus Stanford …

simonwillison.net

14

126

672

@simonw

Simon Willison

4 months

To round off the year, I pulled together some notes on "Stuff we figured out about AI in 2023" We figured out a lot!

Tweet card media

Stuff we figured out about AI in 2023

2023 was the breakthrough year for Large Language Models (LLMs). I think it’s OK to call these AI—they’re the latest and (currently) most interesting development in the academic field of …

simonwillison.net

18

128

663