Seems promising

AI benchmarking should *always* be presented as log percent error rate.

It is clearly not the standard, and it causes people to draw these inane conclusions.

For example:

Humans are 99.999% perfect drivers as measured by accident rate. On a 1 to 100 scale of course it would

a first 5 minutes review of the rabbit r1

just being able to generate mid journey images on the fly

(rather than navigating discord's laborious interface)

gives me instant joy

Apple shipping 2019-era AI capability in 2024

Appropriately captures the company's apparent lack of AI urgency

as the revolution in language-based UI portends a new hardware-software era

Apple released a set of AI models
huggingface.co/apple/OpenELM
Performance no better than random guessing on the MMLU benchmark

On PIQA and Winogrande their 3B parameter models do as well as Roberta with 1/10th the parameters did in 2019

Not a surprise they're looking to partner…

if you think about
robotics operating speed could become a key performance metric
not just for meeting end-user requirements/objectives
but also to maximize training utility of the robotic asset-base
A robot that can operate at 2x the speed can learn a new task twice as fast.

whoa
(conditioned with a fair amount of skepticism)
(given how wildly performant these actuators appear to be compared to anything else that has been demo'd--jerk, speed and precision at speed--all seem a stepchange better)
(and while not teloperated almost certainly scripted)

Tesla's full self drive, now the most powerful publicly available AI system in the world.

More compute than Llama 3 or GPT-4 for each incremental build.

(Under reasonable assumptions.)

Who will buy a car when you can ride in a true automobile?

On why the focus on autonomous is consistent with Tesla’s mission

Why do we believe the ARK Venture Fund, $ARKVX, provides better access to private companies than exchange-traded closed-end funds like The Destiny Tech 100, $DXYZ?

Learn more here.

References to other securities are not an offer to buy, sell or hold. ark-ventures.com/commentary/why…

Huge thank you to Brett Winton & everyone else for tuning in!

This is a must listen recording, check it out down below 👇

Good Morning!

We'll be diving into the future of markets & investing today at 4:00 PM ET with Brett Winton, Chief Futurist at Ark!

Hit us now in the replies with your questions about AI, robotics, or space and we'll make sure to cover everything!

RSVP: twitter.com/i/spaces/1jMJg…

1/ Here is why the Destiny Tech 100 Fund might be too good to be true 🧵

Tomorrow, listen as ARK Chief Futurist Brett Winton joins the Stocktwits team live on X Spaces. Join us at 4 PM ET for a conversation about investing in the future, including AI, robotics, space exploration, and more!

RSVP: twitter.com/i/spaces/1OwGW…

Updated.

A major brand safety verification platform was showing X’s brand safety score varying from 70% to the low 90s for four and a half months when it was actually 99.99% for the entire time period.

🤯

simply put, writing
simply is the worthy work
of writing well