Yeah the problem isn't that value won't be generated by AI, I just personally think that the quantity of investment (without diligence and based on nonsensical false promises) isn't commensurate with the value that will come from the tech.

Self-driving already went through this exact cycle imo.

We're doing cutting edge work on dynamic reconstructions and deploying this technology as the basis for our data pipeline at Stack. If you want to do amazing research that has real world impact, on a team with insanely talented people then you should join us!

We must hit some balls while we talk about RL next time I’m in NYC!

Not to toot my own horn or anything but I’d like to think I did this for multiple papers at NeurIPS this cycle 🫡

Yeah, isn’t asking for experiments explicitly not allowed though? I usually try to limit my requests to different kinds of metrics or analysis of existing experiments/results.

Now you’ll get to enjoy the feeling of wanting to play when you’re away and don’t have a piano, while not playing as much as you should when you’re at home and do have a piano.

Ask me how I know 🫣

Yup! I’ve already written a comment to the ACs to this effect :)

If the reviewers and ACs are engaged during the discussion phase, that’s the best mitigation for these kinds of issues in the review process!

I am fuming on behalf of these authors. This review is complete and utter trash. Whomever wrote it should never be allowed to review again at an ML conference.

It's okay to not be familiar with a field. It's not okay to pretend like you are while spewing nonsense.

We also support pydantic base classes and dataclasses so you can use those for strong typing & strong shape checks.

Check out the source code here: github.com/stackav-oss/...

Feedback & comments are welcome. Happy shape-checking! :)

Yeah I'll be bouncing between the two autonomous driving workshops (WAD and DDADS), so I'll try to catch you during the day!

Woah awesome, we should meet up!

Also @hf.co exists if you want a nice UI!

Even with gdown (which is unofficial) there are still download rate limits for files!

Our goal with the scenario mining challenge is to make it easier to identify scenarios and actors in safety critical situations in order to build out a set of scenarios for validating the performance of AV systems.

Check out the challenge website here: www.argoverse.org/tasks.html

Drago once said that "There are two main challenges in autonomous driving - one is to build a system that can handle real world complexity and edge case complexity, and there's a second challenge which is to evaluate and validate the performance of this system..." - youtu.be/dtMMW0kTXIE?...

Given a natural language text prompt Ex: "vehicles slowing down for a jaywalker" the goal is to identify, detect & track the relevant actors in the scene. All sensor & map data can be used, and we encourage teams to use pre-trained models as well.