Daphne Shen

learning new things. chief of staff @airtreevc; ex operator @ Lyka+Baraja; ex founder @ V-KAIWA. passionate about tech with impact and 🐕.

[Daphne Shen @daphshen]

OpenAI’s new SHAP-E generative models have been really fun to play with, although still in its infancy (eg: my ‘futuristic abstract robot’, and ‘rainbow corgi’ 😅)

I wrote up how I setup and a few results of my prompts here: https://t.co/fJI6xcDcqo

[Daphne Shen @daphshen]

@j_ennychu he is v distinguished 😤

[Daphne Shen @daphshen]

This game is so so much fun!!!

[jenny ✨ chu @j_ennychu]

I never thought there would be a day that I would pay for Apple Arcade AND play a Hello Kitty / Sanrio game.

But here I am.

Anyone else playing Hello kitty Island Adventure? We can explore the island and take photos with Gudetama together. 🥹

[Daphne Shen @daphshen]

@ArtoftheScore @sydsymph It was great, and we’ve been listening to the podcast in the car since!

[Daphne Shen @daphshen]

Hans Zimmer by the @sydsymph this weekend was amazing 😍 + the commentary by @ArtoftheScore was so informative! It felt like listening to a live podcast and the info on how the sounds ‘worked’ was super interesting, and that’s coming from someone who sucks at anything music 😅

[Marc Andreessen 🇺🇸 @pmarca]

Why AI Will Save The World
By Marc Andreessen

The era of Artificial Intelligence is here, and boy are people freaking out.

Fortunately, I am here to bring the good news: AI will not destroy the world, and in fact may save it.
🧵

[Daphne Shen @daphshen]

$3,499 😩

[TechCrunch @TechCrunch]

Breaking: Apple's Vision Pro headset will launch for $3,499 early next year. https://t.co/NCsxLp0GGz #WWDC23

[Daphne Shen @daphshen]

I'm really liking @eeselapp, their categorisation and reccos are great - I almost wish they were **more** annoying with their notifs, though, because while it saves me time, I'm still in the habit of calling it up and using it. 😅

[Daphne Shen @daphshen]

I also wrote about the journey here, which walks thru learnings & challenges! I also now have 300+ photos of cardigan & pembrokes on my laptop... what to do with them...🤣

https://t.co/aGChymcg4m

[Daphne Shen @daphshen]

Of course I had to make a corgi classifier for my very first image classifier project... and it actually performs pretty well 😳

@huggingface has been so beginner friendly, even for someone without web dev experience!

You can play around with it here 💕
https://t.co/UjvHQS13X3

[Daphne Shen @daphshen]

After like 5 years of coding here and there, I finally understand how to deal with API keys / .gitignores. 🤣 It was fun to access Bard via API/python!

https://t.co/NFZqxTAHsF

[Oliver Cameron @olivercameron]

I recently left @Cruise after 2 amazing years.

In that time, we launched & scaled cars with no driver to tens of thousands of happy customers. I'm super proud to have worked at such a special place.

What's next after 7 years in self-driving cars & AI, and what did I learn? 👇

[Daphne Shen @daphshen]

@batkomichael Buy a new copy, then return the old one!! It's worth the hassle

[Daphne Shen @daphshen]

It's so hard to believe this isn't a real person. 🤯 They even made the dips in his back uneven and it's totally gone past the uncanny valley!

[Rowan Cheung @rowancheung]

3. Epic Games Introduces Unreal Engine 5.2

Epic Games unveiled its Machine Learning Deformer on Unreal Engine 5.2.

The sample showcases Unreal Engine's ML technology crafting insanely high-quality, next-gen characters complete with muscle, flesh, AND cloth simulations.

[Michael Girdley @girdley]

PSA: You need a Chief of Staff.

I hired one last year:

My productivity doubled – and I’m having a ton more fun.

I wish I’d done it years ago.

Here's how I did it:

[Daphne Shen @daphshen]

This was us this week 😌

[delian @zebulgar]

What people DON'T REALIZE is how IMPORTANT it is to set yourself up for an ULTRA PRODUCTIVE WEEK on Sundays with these SUNDAY HABITS

Below are a few ways I get myself setup for an ULTRA PRODUCTIVE WEEK:

🧵🧵🧵
👇👇👇

[Daphne Shen @daphshen]

First @huggingface app - so cool to see something deployed & live, even without web development experience 🥳🥺 next up is my own... hehe

https://t.co/t3OILet8eL

[Daphne Shen @daphshen]

Love this - we used to talk to lots of people in academia, with amazing science ideas but no idea how to commercialise into a product or business. Such a practical spin on the traditional business model canvas given the time frames on some of their research/tech!

[Main Sequence @MSeqVC]

Cultivating the next generation of #DeepTech requires a unique approach — one that we’ve coined as Venture Science®. Combining VC tactics & #science, this process helps us uncover future innovators🚀
Starting in Deep Tech can be tricky, so we've created a unique toolkit to help👇

[Jerry Liu @jerryjliu0]

The new 100k token model from @AnthropicAI is awesome: dump in giant docs/books into the prompt, do LLM tasks! 📚🛠️

It got me thinking about the relationship to fine-tuning and in-context learning - is it better than both, worse than both, or used in niche cases? 🧵

In the extreme, if context windows are infinite, then putting all the data into the prompt seems similar to “throw-away” fine-tuning 🤔

Pros ✅:
- Similar to fine-tuning, you get benefits of explicitly giving this black-box model access to all your knowledge (just in the inputs instead of in the weights)
- Less hand-engineered than retrieval-augmented generation (RAG)
- You can more easily feed in new data than actually fine-tuning (which seems hard to use)

Cons ❌:
- You have to feed in this data for every inference call
- As a result, marginal cost/latency go way up. 💵

Retrieval augmented generation is more limited in functionality (because inherently requires some hand-engineering and data pipelining), but on the other hand can reduce cost/latency. So is this approach of feeding everything into the input prompt a happy middle or worst of both worlds?

Of course, going back to the context window of 100k: 100k tokens is a lot, but if you have gigabytes or terabytes of data, 100k tokens can’t fit everything (it can’t actually fit UBER SEC filings). So either way you will need to do some retrieval from your data, in the absence of fine-tuning. And then the question becomes whether you'll always want to maximize the context window, or you're ok with retrieving smaller chunks.

Thoughts? Added some diagrams below to help clarify my thinking 🖼️

[Daphne Shen @daphshen]

when game studios put in effort in the small details 💓 - the animation of loading Link with all these items is so cute and emotive! And he hums to himself while he cooks! 😭

[Daphne Shen @daphshen]

@unixpickle @OpenAI Thanks so much - it was really easy to get set up! That’s really interesting and I’m glad I’m not the only one that was confused about the full robot… 😅

[Daphne Shen @daphshen]

OpenAI’s new SHAP-E generative models have been really fun to play with, although still in its infancy (eg: my ‘futuristic abstract robot’, and ‘rainbow corgi’ 😅)

I wrote up how I setup and a few results of my prompts here: https://t.co/fJI6xcDcqo