Oliver Obst

Assoc Prof in Robotics @ UNSW. Exploring physical & collective intelligence in robotics and real world systems: smart materials, learning, team behaviour

[Oliver Obst @oliverobst]

@aram_cre There is no “semantic” information, like pass or intended targets. The information has ground truth poses, and noisy, incomplete percepts of each player. There’s also a file with the actions received by the server (these are low level actions, kicks, turns and accelerations)

[Susie Dent @susie_dent]

Word of the day is ‘recrudescence’ (17th century): the return of something terrible after a time of reprieve.

[Oliver Obst @oliverobst]

@Michael_J_Black True but some clearly non AI generated reviews are so shallow, unconstructive, or meaningless that I’m sometimes not sure which one to prefer

[Oliver Obst @oliverobst]

Lol

[Ilya Sutskever @ilyasut]

I deeply regret my participation in the board's actions. I never intended to harm OpenAI. I love everything we've built together and I will do everything I can to reunite the company.

[Oliver Obst @oliverobst]

@asurobot @ASURobotics Hey you guys are too quick :) congratulations

[Daniel Jeffries @Dan_Jeffries1]

If you work in AI and you are against open source you should be barred from using any open source tools. Good luck building your models then. No Pytorch for you. No Linux. No Kubernetes. No distributed training frameworks. Nothing.

I never thought we'd be fighting the open source battle twice in my lifetime but here we are.

Don't believe anyone, ever, who tells you that only them and the other "trusted" people can protect you. Kings and Queens and the old church have been running that scam since the Dark Ages.

It was bullshit then and it's bullshit now.

Nobody should be the overseer of AI for the rest of us.

[Oliver Obst @oliverobst]

@zacharylipton @anshulkundaje And while we’re at it, also don’t ask me if a paper is in the top 50/25/10 % of all submissions that haven’t been reviewed yet.

[Francesco Orabona @bremen79]

New blog post: Yet Another ICML Award Fiasco

The story of the @icmlconf 2023 Outstanding Paper Award to the D-Adaptation paper with worse results that the ones from 9 years ago

Please share it to start a needed conversation on mistakenly granted awards

https://t.co/pIIl7BDBlX

[Laura Edelson @LauraEdelson2]

Tech companies are threatening to sue researchers who study them, and I have some first-hand experience surviving this! But researchers shouldn't shy away from this research, and at least this attack vector is less powerful than it appears. Quick 🧵/11 https://t.co/PsLgSZe4Ff

[Oliver Obst @oliverobst]

@DrJimFan @AjdDavison At some point y'all have to explain to me what robotics or representation "will be solved" is supposed to mean

[Juan Álvaro Gallego @JAlGallego]

"The manifold costs of being a non-native English speaker in science" #Science

With a very striking visual summary

https://t.co/RtSByPRPeF

[Oliver Obst @oliverobst]

The Australian Government spends just $12 billion annually on research and science, and you want to spend $7 billion on one sports event? Surely not #qanda

[Oliver Obst @oliverobst]

@sama this is like hoping The Godfather inspires a generation of kids to embrace family values

[Oliver Obst @oliverobst]

Similar arguments have been made about deduction, logic (and AI, 20 years ago), but the argument is too simple and relies on an ambiguous definition of what constitutes new knowledge. Eg, new mathematical proofs have been found by deduction, clearly creating new knowledge.

[Chomba Bupe @ChombaBupe]

Non of these GPT based AI models are building new knowledge, they are just a big associative map between different variables.

[Oliver Obst @oliverobst]

@danielbowen #shuntedAU

[Oliver Obst @oliverobst]

@aryehazan @ccanonne_ I’ve heard it’s expected to draw some large numbers

[Oliver Obst @oliverobst]

@ieee_ras_icra I'm probably just not looking at the right place but wouldn't it be nice to have a list of all accepted papers somewhere on the web page?

[Oliver Obst @oliverobst]

In 2003, 20 years ago, I proposed a 3d physical simulation as a new league for #RoboCup, the largest annual robotics competition. We implemented SimSpark that led to a new competition from 2004 - and it is still being used today. You can read more here:

https://t.co/Y8sVl7JgGR

[Oliver Obst @oliverobst]

It may be that today’s versions of gzip are slightly conscious

[Stephen Diehl @smdiehl]

A 14-line Python script using gzip outperforming a 345m parameter transformer model is probably the most hilarious result I've seen all year.
https://t.co/E7o9gFNA6u

[Oliver Obst @oliverobst]

@maxplanckpress @WeizmannScience The application process looks like it needs clarification

“Each application requires at least three letters of recommendation (one of which should come from the doctoral supervisor or supervisor's mother, if possible).”