Argmax Wed, 04 Mar 2026 18:10:33 -0500 https://www.argmax.fm en-us © 2026 Argmax yes e7012088-1726-5d8c-bbf8-780ccc4ea547 Vahe Hagopian, Taka Hasegawa, Farrukh Rahman episodic false A show where three machine learning enthusiasts talk about recent papers and developments in machine learning. Watch our video on YouTube https://www.youtube.com/@argmaxfm

]]> Buzzsprout (https://www.buzzsprout.com) Vahe Hagopian, Taka Hasegawa, Farrukh Rahman https://storage.buzzsprout.com/76azp67hw9mw8w2g12c46225evxj?.jpg Argmax Mixture of Experts Mixture of Experts In this episode we talk about the paper "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer" by Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, Jeff Dean.

]]> In this episode we talk about the paper "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer" by Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, Jeff Dean.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-15891978 Tue, 08 Oct 2024 12:00:00 -0700 3286 full false LoRA LoRA We talk about Low Rank Approximation for fine tuning Transformers. We are also on YouTube now! Check out the video here: https://youtu.be/lLzHr0VFi3Y

]]> We talk about Low Rank Approximation for fine tuning Transformers. We are also on YouTube now! Check out the video here: https://youtu.be/lLzHr0VFi3Y

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-13517489 Sat, 02 Sep 2023 14:00:00 -0700 3776 2 1 full false 15: InstructGPT 15: InstructGPT In this episode we discuss the paper "Training language models to follow instructions with human feedback" by Ouyang et al (2022). We discuss the RLHF paradigm and how important RL is to tuning GPT.

]]> In this episode we discuss the paper "Training language models to follow instructions with human feedback" by Ouyang et al (2022). We discuss the RLHF paradigm and how important RL is to tuning GPT.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-12534281 Tue, 28 Mar 2023 09:00:00 -0700 3447 1 15 full false 14: Whisper 14: Whisper This week we talk about Whisper. It is a weakly supervised speech recognition model.



]]> This week we talk about Whisper. It is a weakly supervised speech recognition model.



]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-12457660 Fri, 17 Mar 2023 08:00:00 -0700 2954 1 14 full false 13: AlphaTensor 13: AlphaTensor We talk about AlphaTensor, and how researchers were able to find a new algorithm for matrix multiplication.

]]> We talk about AlphaTensor, and how researchers were able to find a new algorithm for matrix multiplication.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-12417434 Fri, 10 Mar 2023 19:00:00 -0800 2945 1 13 full false 12: SIRENs 12: SIRENs In this episode we talked about "Implicit Neural Representations with Periodic Activation Functions" and the strength of periodic non-linearities.

]]> In this episode we talked about "Implicit Neural Representations with Periodic Activation Functions" and the strength of periodic non-linearities.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-11562095 Mon, 24 Oct 2022 18:00:00 -0700 3257 1 12 full true 11: CVPR Workshop on Autonomous Driving Keynote by Ashok Elluswamy, a Tesla engineer 11: CVPR Workshop on Autonomous Driving Keynote by Ashok Elluswamy, a Tesla engineer In this episode we discuss this video: https://youtu.be/jPCV4GKX9Dw

How Tesla approaches collision detection with novel methods.

]]> In this episode we discuss this video: https://youtu.be/jPCV4GKX9Dw

How Tesla approaches collision detection with novel methods.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-11417019 Fri, 30 Sep 2022 16:00:00 -0700 2931 1 11 full false 10: Outracing champion Gran Turismo drivers with deep reinforcement learning 10: Outracing champion Gran Turismo drivers with deep reinforcement learning We discuss Sony AI's accomplishment of creating a novel AI agent that can beat professional racers in Gran Turismo. Some topics include:
- The crafting of rewards to make the agent behave nicely
- What is QR-SAC?
- How to deal with "rare" experiences in the replay buffer

Link to paper: https://www.nature.com/articles/s41586-021-04357-7

]]> We discuss Sony AI's accomplishment of creating a novel AI agent that can beat professional racers in Gran Turismo. Some topics include:
- The crafting of rewards to make the agent behave nicely
- What is QR-SAC?
- How to deal with "rare" experiences in the replay buffer

Link to paper: https://www.nature.com/articles/s41586-021-04357-7

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-11182813 Mon, 22 Aug 2022 17:00:00 -0700 3290 1 10 full false 9: Heads-Up Limit Hold'em Poker Is Solved 9: Heads-Up Limit Hold'em Poker Is Solved Today we talk about recent AI advances in Poker; specifically the use of counterfactual regret minimization to solve the game of 2-player Limit Texas Hold'em.

]]> Today we talk about recent AI advances in Poker; specifically the use of counterfactual regret minimization to solve the game of 2-player Limit Texas Hold'em.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-11051132 Fri, 29 Jul 2022 16:00:00 -0700 2875 full false 8: GATO (A Generalist Agent) 8: GATO (A Generalist Agent) Today we talk about GATO, a multi-modal, multi-task, multi-embodiment generalist agent.

]]> Today we talk about GATO, a multi-modal, multi-task, multi-embodiment generalist agent.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-11051122 Fri, 29 Jul 2022 16:00:00 -0700 2691 1 8 full false 7: Deep Unsupervised Learning Using Nonequilibrium Thermodynamics (Diffusion Models) 7: Deep Unsupervised Learning Using Nonequilibrium Thermodynamics (Diffusion Models) We start talking about diffusion models as a technique for generative deep learning.

]]> We start talking about diffusion models as a technique for generative deep learning.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-10788104 Mon, 13 Jun 2022 18:00:00 -0700 1855 1 7 full true 6: Deep Reinforcement Learning at the Edge of the Statistical Precipice 6: Deep Reinforcement Learning at the Edge of the Statistical Precipice We discuss NeurIPS outstanding paper award winning paper, talking about important topics surrounding metrics and reproducibility.

]]> We discuss NeurIPS outstanding paper award winning paper, talking about important topics surrounding metrics and reproducibility.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-10748486 Mon, 06 Jun 2022 16:00:00 -0700 3668 1 6 full false 5: QMIX 5: QMIX We talk about QMIX https://arxiv.org/abs/1803.11485 as an example of Deep Multi-agent RL.

]]> We talk about QMIX https://arxiv.org/abs/1803.11485 as an example of Deep Multi-agent RL.

]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-10504620 Mon, 25 Apr 2022 20:00:00 -0700 2526 1 5 full false 4: Can Neural Nets Learn the Same Model Twice? 4: Can Neural Nets Learn the Same Model Twice? Todays paper: Can Neural Nets Learn the Same Model Twice? Investigating Reproducibility
and Double Descent from the Decision Boundary Perspective (https://arxiv.org/pdf/2203.08124.pdf)

Summary:
A discussion of reproducibility and double descent through visualizations of decision boundaries.

Highlights of the discussion:

  • Relationship between model performance and reproducibility
  • Which models are robust and reproducible
  • How they calculate the various scores



]]> Todays paper: Can Neural Nets Learn the Same Model Twice? Investigating Reproducibility
and Double Descent from the Decision Boundary Perspective (https://arxiv.org/pdf/2203.08124.pdf)

Summary:
A discussion of reproducibility and double descent through visualizations of decision boundaries.

Highlights of the discussion:

  • Relationship between model performance and reproducibility
  • Which models are robust and reproducible
  • How they calculate the various scores



]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-10386978 Tue, 05 Apr 2022 19:00:00 -0700 3323 1 4 full false 3: VICReg 3: VICReg Todays paper: VICReg (https://arxiv.org/abs/2105.04906)

Summary of the paper
VICReg prevents representation collapse using a mixture of variance, invariance and covariance when calculating the loss. It does not require negative samples and achieves great performance on downstream tasks.

Highlights of discussion

  • The VICReg architecture (Figure 1)
  • Sensitivity to hyperparameters (Table 7)
  • Top 5 metric usefulness
]]> Todays paper: VICReg (https://arxiv.org/abs/2105.04906)

Summary of the paper
VICReg prevents representation collapse using a mixture of variance, invariance and covariance when calculating the loss. It does not require negative samples and achieves great performance on downstream tasks.

Highlights of discussion

  • The VICReg architecture (Figure 1)
  • Sensitivity to hyperparameters (Table 7)
  • Top 5 metric usefulness
]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-10292930 Mon, 21 Mar 2022 16:00:00 -0700 2686 1 3 full false 2: data2vec 2: data2vec Todays paper: data2vec (https://arxiv.org/abs/2202.03555)

Summary of the paper
A multimodal SSL algorithm that predicts latent representation of different types of input.

Highlights of discussion

  • What are the motivations of SSL and multimodal
  • How does the student teacher learning work?
  • What are similarities and differences between ViT, BYOL, and Reinforcement Learning algorithms.
]]> Todays paper: data2vec (https://arxiv.org/abs/2202.03555)

Summary of the paper
A multimodal SSL algorithm that predicts latent representation of different types of input.

Highlights of discussion

  • What are the motivations of SSL and multimodal
  • How does the student teacher learning work?
  • What are similarities and differences between ViT, BYOL, and Reinforcement Learning algorithms.
]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-10205015 Mon, 07 Mar 2022 11:00:00 -0800 3203 1 2 full false 1: Reward is Enough 1: Reward is Enough This is the first episode of Argmax! We talk about our motivations for doing a podcast, and what we hope listeners will get out of it.

Todays paper: Reward is Enough

Summary of the paper
The authors present the Reward is Enough hypothesis: Intelligence, and its associated abilities, can be understood as subserving the maximisation of reward by an agent acting in its environment.

Highlights of discussion

  • High level overview of Reinforcement Learning
  • How evolution can be encoded as a reward maximization problem
  • What is the one reward signal we are trying to optimize?
]]> This is the first episode of Argmax! We talk about our motivations for doing a podcast, and what we hope listeners will get out of it.

Todays paper: Reward is Enough

Summary of the paper
The authors present the Reward is Enough hypothesis: Intelligence, and its associated abilities, can be understood as subserving the maximisation of reward by an agent acting in its environment.

Highlights of discussion

  • High level overview of Reinforcement Learning
  • How evolution can be encoded as a reward maximization problem
  • What is the one reward signal we are trying to optimize?
]]> Vahe Hagopian, Taka Hasegawa, Farrukh Rahman Buzzsprout-10116664 Mon, 21 Feb 2022 11:00:00 -0800 3276 1 1 full false