zach s

senior program manager @disneyaccel. prev @blackstone. alum @harvard. ps, ds, nyg, la, & spongebob supremacy. views my own

[zach s @ShalcyonPrime]

I’m obsessed with animation. Especially in gaming

Forget “realistic graphics” - I value animation that conveys real emotional and plot detail. Form as content

Pixar, yeah. But also games like @Guacamelee

@IGN and @GameSpot don’t talk about this “detail” enough

[Tim Sweeney @TimSweeneyEpic]

Looking at Stable Diffusion prompt sites and seeing people put “rendered in Unreal Engine 5” in photographic prompts is a sign of our times!

[Soren Iverson @soren_iverson]

iMessage shows you when someone takes a screenshot

[Vitalist Doxometrist @doxometrist]

To those considering using cattle to plough the earth: PLEASE DON'T.

Not only are you hurting the industry of manual ploughers but (esp if you are not being upfront) your reputation is going to take a hit.

There is no magic shortcut. You need to do the work.

[zach s @ShalcyonPrime]

A long term project while I was on our corporate strategy team. A solid win for the House of Mouse imo

Very proud - So much internal alignment needs to happen for an announcement like this. And amidst leadership and market shifts, they landed the plane

Late but great! Go team

[ESPN PR @ESPNPR]

ESPN has entered an agreement with @PENNEntertain to launch ESPN BET, a branded sportsbook in the U.S.

ESPN BET will be available for fans this fall

Details: https://t.co/evaUbfn5Nh

[zach s @ShalcyonPrime]

Early stud participant in the Disney Accelerator!

[Caleb @caleb_friesen]

Did a deep dive into the very early years of BYJU'S recently and made some interesting discoveries 🧵

[zach s @ShalcyonPrime]

Genius.

[Trung Phan @TrungTPhan]

More details on Amazon’s AI strategy.

🔗 https://t.co/BMuiIWbM3p

[Trung Phan @TrungTPhan]

The pace of technology adoption for large organizations is soooooo slow.

[Jim Fan @DrJimFan]

You'll soon see lots of "Llama just dethroned ChatGPT" or "OpenAI is so done" posts on Twitter. Before your timeline gets flooded, I'll share my notes:

▸ Llama-2 likely costs $20M+ to train. Meta has done an incredible service to the community by releasing the model with a commercially-friendly license. AI researchers from big companies were wary of Llama-1 due to licensing issues, but now I think many of them will jump on the ship and contribute their firepower.

▸ Meta's team did a human study on 4K prompts to evaluate Llama-2's helpfulness. They use "win rate" as a metric to compare models, in similar spirit as the Vicuna benchmark. 70B model roughly ties with GPT-3.5-0301, and performs noticeably stronger than Falcon, MPT, and Vicuna.
I trust these real human ratings more than academic benchmarks, because they typically capture the "in-the-wild vibe" better.

▸ Llama-2 is NOT yet at GPT-3.5 level, mainly because of its weak coding abilities. On "HumanEval" (standard coding benchmark), it isn't nearly as good as StarCoder or many other models specifically designed for coding. That being said, I have little doubt that Llama-2 will improve significantly thanks to its open weights.

▸ Meta's team goes above and beyond on AI safety issues. In fact, almost half of the paper is talking about safety guardrails, red-teaming, and evaluations. A round of applause for such responsible efforts!

In prior works, there's a thorny tradeoff between helpfulness and safety. Meta mitigates this by training 2 separate reward models. They aren't open-source yet, but would be extremely valuable to the community.

▸ I think Llama-2 will dramatically boost multimodal AI and robotics research. These fields need more than just blackbox access to an API.
So far, we have to convert the complex sensory signals (video, audio, 3D perception) to text description and then feed to an LLM, which is awkward and leads to huge information loss. It'd be much more effective to graft sensory modules directly on a strong LLM backbone.

▸ The whitepaper itself is a masterpiece. Unlike GPT-4's paper that shared very little info, Llama-2 spelled out the entire recipe, including model details, training stages, hardware, data pipeline, and annotation process. For example, there's a systematic analysis on the effect of RLHF with nice visualizations.

Quote sec 5.1: "We posit that the superior writing abilities of LLMs, as manifested in surpassing human annotators in certain tasks, are fundamentally driven by RLHF."

Congrats to the team again 🥂! Today is another delightful day in OSS AI.

[VENOMHOLOGY @venmohology]

Venom arrives at the Barbie World Premiere #Barbie @VenomMovie

[Andrew @andrewmichaelio]

Mark Zuckerberg talking about Facebook users

[zach s @ShalcyonPrime]

BREAKING: Local mutant rat sees tsunami outside window, attempts to divert tsunami by acting casual

[Elon Musk @elonmusk]

Must confess to a deep love of fashion & architecture

[zach s @ShalcyonPrime]

I guess you’re Zuck brother

[Austin Rief ☕️ @austin_rief]

There’s never been a better time to compete with twitter.

If I’m Zuck, I’m launching my Twitter competitor today.

[zach s @ShalcyonPrime]

Toilet and traffic time share hitting astronomically different rn

[Tom Osman 🐦‍⬛ @tomosman]

Pay attention to your time.

1 hour per day = 4% of your life.

[zach s @ShalcyonPrime]

Would love to hear more sleepers!

Let me know what comes to mind - favorite animated content that uses its graphical style as storytelling currency

[zach s @ShalcyonPrime]

I’m obsessed with animation. Especially in gaming

Forget “realistic graphics” - I value animation that conveys real emotional and plot detail. Form as content

Pixar, yeah. But also games like @Guacamelee

@IGN and @GameSpot don’t talk about this “detail” enough

[zach s @ShalcyonPrime]

Less slept on examples:

- ANYTHING Studio Ghibli
- SpongeBob (the true goat)
- Shrek
- Regular Show

[Luciano Arango @ai_luce]

🗽💳Proud to announce the launch of Maza!🚀 As a child of two undocumented immigrant parents, I witnessed the struggles they faced firsthand. Today, I'm fulfilling my lifelong dream by founding Maza, a company that provides a path to the American Dream for immigrants without SSNs

[zach s @ShalcyonPrime]

Idea: 2 way opt-in for iMessage users to share their current time zone

Can tie to location sharing/Find My sharing, or just opt to share/receive timezone on iMessage

Would improve comms/expectations on responsiveness. Can add timestamp metadata too

@TrungTPhan @anothercohen

[Alex Heath @alexeheath]

I’ve confirmed that Mark Zuckerberg is serious about fighting @elonmusk and is now waiting on the details (if Musk decides to follow through)

“The story speaks for itself,” a Meta spokesperson says re: Zuck’s IG post saying “send me location”

https://t.co/4g1IkqOl47