Quander

Machine Learning Street Talk check_circle

0 subscribers

1:03:36

Test-Time Adaptation: the key to reasoning with DL

more_vert

Machine Learning Street Talk check_circle

7.1K views 3 days ago

1:23:12

The Gap Between Humans and Machines Is ___

more_vert

Machine Learning Street Talk check_circle

6.2K views 7 days ago

39:30

Panel discussion on ARC Prize 2024 (Zurich)

more_vert

Machine Learning Street Talk check_circle

4.5K views 8 days ago

1:41:19

Tau Language: The Software Synthesis Future

more_vert

Machine Learning Street Talk check_circle

12K views 13 days ago

1:00:55

Transformers Need Glasses!

more_vert

Machine Learning Street Talk check_circle

17K views 2 weeks ago

1:18:11

Neural Networks Are Elastic Origami!

more_vert

Machine Learning Street Talk check_circle

12K views 1 month ago

1:09:05

The ARC Prize 2024 Winning Algorithm

more_vert

Machine Learning Street Talk check_circle

13K views 1 month ago

53:32

ImageNet Moment for Reinforcement Learning?

more_vert

Machine Learning Street Talk check_circle

23K views 1 month ago

51:27

Can Latent Program Networks Solve Abstract Reasoning?

more_vert

Machine Learning Street Talk check_circle

12K views 1 month ago

1:41:54

GODFATHER OF AI: MIGHT THE ROBOTS TAKE OVER?

more_vert

Machine Learning Street Talk check_circle

31K views 2 months ago

1:21:16

Language Models are "Modelling The World"

more_vert

Machine Learning Street Talk check_circle

28K views 2 months ago

3:42:37

NEURAL NETWORKS ARE WEIRD! - Neel Nanda (DeepMind)

more_vert

Machine Learning Street Talk check_circle

100K views 3 months ago

4:17:08

YUDKOWSKY + WOLFRAM ON AI RISK.

more_vert

Machine Learning Street Talk check_circle

103K views 4 months ago

2:42:55

Pattern Recognition vs True Intelligence - Francois Chollet

more_vert

Machine Learning Street Talk check_circle

67K views 4 months ago

1:03:36

Michael Levin - Why Intelligence Isn't Limited To Brains.

more_vert

Machine Learning Street Talk check_circle

83K views 5 months ago

1:52:46

Joscha Bach - Why Your Thoughts Aren't Yours.

more_vert

Machine Learning Street Talk check_circle

114K views 5 months ago

1:39:39

Neural and Non-Neural AI, Reasoning, Transformers, and LSTMs

more_vert

Machine Learning Street Talk check_circle

89K views 6 months ago

00:57:30

CHATGPT + WOLFRAM - THE FUTURE OF AI!

more_vert

Machine Learning Street Talk check_circle

1 views 2 years ago

,Episode #110 HUGE ANNOUNCEMENT, CHATGPT+WOLFRAM! You saw it HERE first! Pod version: https://podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/110-Dr--STEPHEN-WOLFRAM---HUGE-ChatGPTWolfram-announcement-e210qda Support us! https://www.patreon.com/mlst MLST Discord: https://discord.gg/aNPkGUQtc5 Stephen's announcement post: https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/ OpenAI's announcement post: https://openai.com/blog/chatgpt-plugins In an era of technology and innovation, few individuals have left as indelible a mark on the fabric of modern science as our esteemed guest, Dr. Steven Wolfram. Dr. Wolfram is a renowned polymath who has made significant contributions to the fields of physics, computer science, and mathematics. A prodigious young man too, Wolfram earned a Ph.D. in theoretical physics from the California Institute of Technology by the age of 20. He became the youngest recipient of the prestigious MacArthur Fellowship at the age of 21. Wolfram's groundbreaking computational tool, Mathematica, was launched in 1988 and has become a cornerstone for researchers and innovators worldwide. In 2002, he published "A New Kind of Science," a paradigm-shifting work that explores the foundations of science through the lens of computational systems. In 2009, Wolfram created Wolfram Alpha, a computational knowledge engine utilized by millions of users worldwide. His current focus is on the Wolfram Language, a powerful programming language designed to democratize access to cutting-edge technology. Wolfram's numerous accolades include honorary doctorates and fellowships from prestigious institutions. As an influential thinker, Dr. Wolfram has dedicated his life to unraveling the mysteries of the universe and making computation accessible to all. First of all... we have an announcement to make, you heard it FIRST here on MLST! .... [00:00] Intro [02:57] Big announcement! Wolfram + ChatGPT! [05:33] What does it mean to understand? [13:48] Feeding information back into the model [20:09] Semantics and cognitive categories [23:50] Navigating the ruliad [31:39] Computational irreducibility [38:43] Conceivability and interestingness [43:43] Human intelligible sciences

00:00:59

Huge announcement from Stephen Wolfram! Official ChatGPT integration! #gpt4

more_vert

Machine Learning Street Talk check_circle

1 views 2 years ago

3:19:44

#61: Prof. YANN LECUN: Interpolation, Extrapolation and Linearisation (w/ Dr. Randall Balestriero)

more_vert

Machine Learning Street Talk check_circle

167K views 3 years ago

01:51:48

OpenAI GPT-3: Language Models are Few-Shot Learners

more_vert

Machine Learning Street Talk check_circle

1 views 4 years ago

,**ERRATA**: Open AI/GPT-3 DOES NOT USE Microsoft's ZeRO/DeepSpeed for training Discord: https://discord.gg/4H8xxDF In this episode of Machine Learning Street Talk, Tim Scarfe, Yannic Kilcher and Connor Shorten discuss their takeaways from OpenAI’s GPT-3 language model. OpenAI trained a 175 BILLION parameter autoregressive language model. The paper demonstrates how self-supervised language modelling at this scale can perform many downstream tasks without fine-tuning. 00:00:00 Intro 00:00:54 ZeRO1+2 (model + Data parallelism) [GPT-3 DOES *NOT* USE THIS] (Connor) 00:03:17 Recent history of NLP (Tim) 00:06:04 Yannic "Light-speed" Kilcher's brief overview of GPT-3 00:14:25 Reviewing Yannic's YT comments on his GPT-3 video (Tim) 00:20:26 Main show intro 00:23:03 Is GPT-3 reasoning? 00:28:15 Architecture discussion and autoregressive (GPT*) vs denoising autoencoder (BERT) 00:36:18 Utility of GPT-3 in industry 00:43:03 Can GPT-3 do math? (reasoning/system 1/system 2) 00:51:03 Generalisation 00:56:48 Esoterics of language models 00:58:46 Architectural trade-offs 01:07:37 Memorization machines and intepretability 01:17:16 Nearest neighbour probes / watermarks 01:20:03 YouTube comments on GPT-3 video 01:21:50 GPT-3 news article generation issue 01:27:36 Sampling data for language models / bias / fairness / politics 01:51:12 Outro These paradigms of task adaptation are divided into zero, one, and few shot learning. Zero-shot learning is a very extreme case where we expect a language model to perform a task such as sentiment classification or extractive question answering, without any additional supervision. One and Few-shot learning provide some examples to the model. However, GPT-3s definition of this diverges a bit from the conventional literature. GPT-3 provides one and few-shot examples in the form of “In-Context Learning”. Instead of fine-tuning the model on a few examples, the model has to use the input to infer the downstream task. For example, the GPT-3 transformer has an input sequence of 2048 tokens, so demonstrations of a task such as yelp sentiment reviews, would have to fit in this input sequence as well as the new review. **ERRATA-continued** It has come to our attention that there was a serious factual error in our video -- GPT-3 DOES NOT USE Microsoft's ZeRO/ZeRO2 or DeepSpeed for training and there is no reference to this in either their blog post or paper. We are really sorry about this mistake and will be more careful to fact-check in future. Thanks for watching! Please Subscribe! Paper Links: GPT-3: https://arxiv.org/abs/2005.14165 #machinelearning #naturallanguageprocessing #deeplearning #gpt3