[ML News] Facebook AI adapting robots | Baidu autonomous excavators | Happy Birthday EleutherAI

preview_player
Показать описание
A look into the happenings of the Machine Learning world.

OUTLINE:
0:00 - Intro
0:25 - Facebook AI trains rapidly adapting robots
3:05 - Baidu presents autonomous excavator system
4:45 - EleutherAI turns 1
6:05 - Elon Musk says FSD harder than expected
8:10 - AI interview tools still fall short
11:10 - RunwayML AI-powered cloud video editor
11:55 - MineRL BASALT competition to learn from human feedback
13:15 - The Myth of the Expert Reviewer
15:55 - NVIDIA unveils Cambridge-1 supercomputer
17:10 - CLIP art sees rapid improvements
19:00 - AI demystifies boiling
21:20 - AI avatars for easier language learning
23:20 - Outro

References:
Facebook AI trains rapidly adapting robots

Baidu presents autonomous excavator system

EleutherAI turns 1

Elon Musk says FSD is harder than expected

AI interview tools still fall short

RunwayML AI-powered cloud video editor

MineRL BASALT competition to learn from human feedback

The Myth of the Expert Reviewer

NVIDIA unveils Cambridge-1 supercomputer

CLIP art sees rapid improvements

AI demystifies boiling

AI avatars for easier language learning

Links:

If you want to support me, the best thing to do is to share out the content :)

If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this):
Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq
Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2
Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m
Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
Рекомендации по теме
Комментарии
Автор

No, I'm not "back", this was just in the backlog 🙃

YannicKilcher
Автор

Hey, I'm the lead organizer for the MineRL BASALT competition. Thanks for including us! I just wanted to note that we're aware of the problems with an economy-based human evaluation. For those reasons, this will only be used to provide intermediate feedback to participants, where the stakes are lower. For the final scores that determine who wins the prizes, we will be hiring contractors to provide the evaluations, and doing quality control to ensure the comparisons are good (and in particular not just clicking randomly as fast as possible).

rohinshah
Автор

Our beloved ML News has been upgraded to “moderately regular” !!!

PeterOtt
Автор

No, he is not back. He is still on a break. He just surpassed the speed of light this time and went back in time upload the video and returned to his break.

onuryes
Автор

Yannick : "I go on a break."

Yannick literally 2 days later : "ML NEWS!"

AvastarBin
Автор

"AI demystifies boiling"
Finally!
"What color were the eggs?!" Lmao

maherhayek
Автор

The "boiling" section was incredibly funny 😂

CristianGarcia
Автор

4:00 Not sure about AI generated voice. It sounds human to me. Then perhaps my "discriminator" isnt trained enough.
.
BOILING IS A SOCIAL CONSTRUCT XD

herp_derpingson
Автор

3 days after the video about taking a break. ML News a *regular* update on ML world confirmed.

audrius
Автор

17:04 silicon valley reference, SILICON VALLEY REFERENCE

theshishirjoshi
Автор

Thanks for pointing out that Verge Url!

alexijohansen
Автор

love these ML news so much!!! great work!!!

KalunHo
Автор

I am trying to build a small robot that patrols my yard looking out for critters like scorpions, wasps, etc. I love that we live in a world where even someone like me without lots of money can do that, where the hardware is cheap enough and lots of ML is freely available.

jameshughes
Автор

The point about the "perfectly friendly and helpful" human in the avatars and chatbots etc. is the problem for me too, and was a great point I think. The avatars should be just a little stinky and have their own opinions etc., that would make them less creepy and more likeable to be around. :)

ImaplanetJupiteeeerr
Автор

Can you discuss the alphafold2 paper when you are back? Would love to hear your explanation on it!

vcp_rocketeers_hk
Автор

Self driving cars already are much safer than human drivers. What do we need the hard problem to be solved for? So that cars could regret hitting pedestrians?

XOPOIIIO
Автор

On the reviewers. The problems is not the reviewers. The problem is that the scientific research is not funded appropriately by the government. The university professors now have to compete against each other for grants. 50 years ago the funding was much easier to get. The scientists were much more inclined to promote good ideas, because it did not affect their lives as much. Today, if you spend time reading papers, it is time you are not writing yours, or revising yours to fit the whims of publishers and reviewers. It turns the scientific community into tightly woven circles of "back scratchers". Not to mention the horribly backwards system of scientific publishers being a business.

In this environment it is not the best articles that are being published but the most entertaining. With the obvious effects on the quality.

evgenysavelev
Автор

The Verge lost all credibility after their screw-up with the PC built and how they responded to their criticisms, their articles are a joke to anyone competent

AkashMishra
Автор

A 6 in Germany is a 1 on normal grading scales

joedalton
Автор

Wait I thought you are taking a break??

seetj
visit shbcf.ru