December 07 2024 20:38:31
News Photos Forum Search Contact History Linkbox Calendar
 
View Thread
Gongumenn | General | General Discussion
Page 4 of 5 < 1 2 3 4 5 >
85
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 02-03-2024 09:53
Vuzman wrote:
Well, I hope my pessimism turns out to be right, because I really don't want these clowns to be in control of any AGI


Well, if you turn out to be wrong, Elon has your back:




You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 06-03-2024 13:23
OpenAI responds to the lawsuit by making some Elon<->openAI correspondence public, which clearly shows that Elon didn't want to share power with anyone.



You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 13-03-2024 00:22
Meet Devin, your replacement. EDIT: this dude has some very clickbaity sensational youtube splash screens, but the content is usually pretty balanced and factual.



You want to tempt the wrath of the whatever from high atop the thing?

Edited by Grizlas on 13-03-2024 11:34
Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 13-03-2024 19:29
The humanoid robot race is picking up speed. Here's the newest from Figure (backed by Open AI):




You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 15-03-2024 18:19
Interesting talk with another AI professor - one from the skeptics camp:




You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 22-05-2024 17:05
This guy might not have a Nobel prize, but he sure has some of that Feynman charisma. Great talk!



You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
OKJones
RE: AI discussion

User Avatar

Commander

Group: Klikan
Location: Argir
Joined: 12.06.06
Posted on 23-05-2024 20:54


This is hilarious. Verstappen is not losing his job anytime soon.


Why would I want to end every post the same way?

Edited by OKJones on 23-05-2024 20:54
Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 24-05-2024 15:52
what a huge mess ! smiley

Apparently this was done by some university AI teams with two months to prepare?
would be interesting to see Waymo and tesla compete.


You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
OKJones
RE: AI discussion

User Avatar

Commander

Group: Klikan
Location: Argir
Joined: 12.06.06
Posted on 24-05-2024 23:27
Ok, thats not a lot of time. But I just assumed this wouldn't be that hard to get right, but what do I know smiley


Why would I want to end every post the same way?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 25-05-2024 09:48
Civil discussion on the dangers of AI with top scientists from both camps:




You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 13-06-2024 14:11
AI is getting pretty sick smiley

Remember my quiz from 2. jólakvøøld?

this "Millum Mussarnar" picture:



and here is the video:

pompt: make the people dance

https://storage.cdn-luma.com/lit_lite_inference_im2vid_v1.0/b66e421e-acb1-4453-8ac2-16bd55ee3479/watermarked_video0534433cbecd24bd4a2fbb3fb91b04c96.mp4

and here, pompt: make the people run

https://storage.cdn-luma.com/lit_lite_inference_im2vid_v1.0/f51b498a-1114-40ea-abab-fcf5a9279c13/watermarked_video03535d5be85fc4161892c01c4550cf050.mp4


You should check out Luma AI which is basically free Sora
you get 30 free prompts a month
the internet is flooding this thing right now, so you have to be patient with the generation


You want to tempt the wrath of the whatever from high atop the thing?

Edited by Grizlas on 13-06-2024 14:24
Send Private Message
Norlander
RE: AI discussion

User Avatar

Field Marshal

Group: Administrator, Klikan, Regulars, Outsiders
Location: Copenhagen
Joined: 09.06.06
Posted on 13-06-2024 15:25
Awesome, will check Luma out. And yeah making graphics for applications is much easier now that we don't have to wait on some graphic artist smiley


The conventional view serves to protect us from the painful job of thinking.
- John Kenneth Galbraith

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 15-06-2024 23:11
A decent AI movie:




You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 16-06-2024 07:25
Elon Musk is throwing a massive fit about Apple partnering with OpenAI, with plans to integrate ChatGPT at the OS level (i.e. Siri)



and now they havea appointed the last NSA chief to the OpenAI board.
Edward Snowden, for one, thinks this is no bueno:




You want to tempt the wrath of the whatever from high atop the thing?

Edited by Grizlas on 16-06-2024 07:29
Send Private Message
Norlander
RE: AI discussion

User Avatar

Field Marshal

Group: Administrator, Klikan, Regulars, Outsiders
Location: Copenhagen
Joined: 09.06.06
Posted on 16-06-2024 11:22
Elon haters having a hard time on how to react to that one smileysmiley

On the one hand it is Elon warning them, so they'll naturally be all for doing whatever he is warning against, but on the other hand he and Peter Thiel (which they also hate) are two of the co-founders of OpenAI, and they think all his companies (current and former) are horrible, so maybe Apple teaming up with them is shit.


The conventional view serves to protect us from the painful job of thinking.
- John Kenneth Galbraith

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 20-07-2024 12:29
Some more doomerism - Narrated by Stephen Fry (or some AI mimicking his voice)




You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 25-07-2024 22:45
Having beaten humans at chess, go and protein folding with AlphaGo, AlphaZero and AlphaFold, Google DeepMind is now going after math. Their new AI system: AlphaProof (and AlphaGeometry) uses a natural language model combined with the reinforcement learning of AlphaZero to teach itself how to to prove math problems in a formal language called Lean.

They tested AlphaProof in some super prestigious math competition and got a 28 points. There were 609 contestants and 58 of them got 29 points or more.

AlfpProof did need some help to make it this far:
1. The contestants had two 4.5 hour sessions, while AlphaProof took 3 days to complete the 4 problems it managed to solve.
2. The math problems were translated from text to Lean manually - even though they claim that they are almost able to do this using the Gemini LLM.

full story:
https://deepmind.google/discover/blog/ai-solves-imo-problems-at-silver-medal-level/

I think this is very interesting. Let's see how it does next year!


You want to tempt the wrath of the whatever from high atop the thing?

Edited by Grizlas on 25-07-2024 23:22
Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 12-08-2024 10:04
Google Deepmind is apparently also working on table tennis:

Whilst the robot lost all matches vs. the most advanced players it won 100% matches vs. beginners and 55% matches vs. intermediate players, demonstrating solidly amateur human-level performance.




I looked through their paper. The robot uses two cameras for tracking the ball, (realistic for a humanoid robot) and another 20 cameras around the area to track the position of the opponents' paddle (not realistic for a humanoid robot). Other than that, I found their "limitations" interesting:

Our research demonstrates progress in training robots to play table tennis against human opponents. However, several
limitations warrant further investigation: (1) Fast Balls: The robot struggles to react effectively to very fast balls due
to multiple factors; lack of data, high system latency, and between-shot reset. Even with a high control frequency
(50hz), the inherent latency in the system (∼100ms) significantly restricts the number of actionable decisions the
robot can make within a given timeframe. This problem was exacerbated by the fact that the robot was reset to
the same initial pose — close to the table, paddle facing forward — in between each ball hit to account for the single
ball training. While increasing the control frequency could mitigate this issue, it reduces the available time per action for
computation. (2) High Balls: The robot is unable to handle balls higher than ∼6f t above the table, as it goes outside the
camera FOV. In practice, it was seldom a problem. (3) Low Balls: The robot’s inability to consistently handle very low
balls is primarily attributed to collision avoidance protocols built into the system. These protocols, while essential for
protecting the robot’s paddle, pose a significant challenge in handling balls that stay close to the table. This limitation was
frequently exploited by skilled players by sending underspin balls that stay very low. (4) Short Pips Paddle: In this work,
we chose a paddle that can impart spin, while still being possible to model within the simulator. The paddle doesn’t
have the smooth rubber found on advanced table tennis paddles [17]. We found accurately modeling such advanced
paddles extremely challenging given the highly nonlinear and multimodal nature of the paddle. This limitation was frequently noted by human participants and highlights the need
for more advanced paddle modeling techniques. (5) Spin Detection The policy is unable to accurately read extreme spins on an incoming ball. This limited the amount of spin that
could be effectively handled, a fact that was exploited by a number of opponents. It was possible to train specialist LLC
policies to handle different types of spin, so improvements in spin detection would likely directly translate to improved
overall performance. (6) No Multi-ball Strategic Game Play: The robot’s gameplay is predominantly ”one ball at a
time”, lacking the multi-ball strategic play observed in skilled human players. This limitation is attributed to the single-ball
training approach. (7) Motion Capture Unreliability: The reliance on motion capture technology introduces a degree of
unreliability, as occasional glitches or inaccuracies in motion tracking can negatively impact the robot’s performance. (8)
Backhand The policy had weaker backhand play compared with the forehand. (9) Predictability The policy did vary
its placement however this could be improved further. Being less predictable in placement of certain balls was one of the
top three areas for improvement as identified by the match referee (in addition to reading spin better and improving the backhand).


Many of these limitations can likely be reduced, but if those "motion capture inaccuracies" are in any way related to their 20 camera capture system, then there is some ways to go before we will see unassisted humanoid robots playing table tennis.

Very interesting nonetheless!


You want to tempt the wrath of the whatever from high atop the thing?

Edited by Grizlas on 12-08-2024 10:47
Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 29-08-2024 16:46
AI trained to "play" doom:

Hyped version:
https://www.instagram.com/reel/C_N0zXlJhxv/?utm_source=ig_web_button_share_sheet&igsh=MzRlODBiNWFlZA==

the underhyped version:



You want to tempt the wrath of the whatever from high atop the thing?

Edited by Grizlas on 29-08-2024 16:47
Send Private Message
Grizlas
RE: AI discussion

User Avatar

General

Group: Administrator, Klikan, Regulars, Outsiders
Location: Denmark
Joined: 08.06.06
Posted on 11-09-2024 16:49
These LLMS are not that great at planning. This shortcoming is often obscured in the popular benchmark charts like chatarena:



Enter PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Change



They use planning problems from their "Blocksworld" problem generating engine which looks something like this:


As you can see, the chatbots do pretty badly with results from 0.5% to 62%. but then they have a Mystery Blocksworld test, where they swap in more obscure operating words:




this completely fucks up the chatbots smiley


You want to tempt the wrath of the whatever from high atop the thing?

Send Private Message
Page 4 of 5 < 1 2 3 4 5 >
Jump to Forum:
Back to front page