saim @saimcodes

The TurboQuant paper (ICLR 2026) contains serious issues in how it describes RaBitQ, including incorrect technical claims and misleading theory/experiment comparisons. We flagged these issues to the authors before submission. They acknowledged them, but chose not to fix them. The paper was later accepted and widely promoted by Google, reaching tens of millions of views. We’re speaking up now because once a misleading narrative spreads, it becomes much harder to correct. We’ve written a public comment on openreview (openreview.net/forum?id=tO3AS…). We would greatly appreciate your attention and help in sharing it.

Google Research @GoogleResearch

3 months ago

Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI

1K 6K 39K 19.4M 22K

on a side note, the founders of inception labs have designed a comprehensive course on deep generative models with a curated combination of videos + notes/slides + related papers + projects. def worth to checkout.

Inception @_inception_ai

4 months ago

Mercury 2 is live. The world's first reasoning diffusion LLM – 5x faster than leading speed-optimized autoregressive models. Built for production: multi-step agents without delays, voice AI with tight latency budgets, instant coding feedback. Diffusion-based generation enables

31 59 430 197K 202

You either die a state of the art or live long enough to see yourself become the baseline

@novasarc01 thankss

@hallerite ludic looks clean. excited for the rlm part

@mfranz_on awesome work dude!!

@zhaisf love this duality: treating low-rank gradients as feature (LoRA) vs bug (Muon) perfectly captures the trade-off in modern training

@kevinlu625 would love to test it outtt

Self-Improving VLM Judges Without Human Annotations

@asmah2107 solid list bruv🔥

@budapp would love to test it out 😁

@yuviecodes @elonmusk @xai @krishgarg @lohanipravin nah bro keep your evil ad-injecting ai shit away from my shows, we never asked for this dystopian garbage 💀

@willccbb yeah this hits hard, basically hedge on steroids. still wondering if any of the new process-rl stuff actually escapes the base support or just finds fancier ways to collapse it 😂

@kylejeong grandparents moving different now 💀

@rryssf Research

Both Perceptual Control Theory (PCT) and the Bayesian/Predictive Processing (PP) framework aim to explain how organisms achieve adaptive behavior in uncertain environments. Although they appear opposed, they actually capture different levels of description. Understanding their relationship requires separating mechanistic implementations, functional descriptions, and computational metaphors. 1. What PCT gets right. PCT highlights a core fact about control systems, biological or engineered, organisms act to control their sensory input, not their motor output. This reframes behavior as the ongoing reduction of deviation from an internally maintained “reference value.” PCT’s successes include catch a flyball, constant optical velocity is easier and more robust than trajectory prediction. Cursor disturbance experiment, people compensate for structured forces without consciously modeling them. Robotics parallels, balance and locomotion often work best with simple reactive controllers, not predictive models. Strength: PCT captures real-world robustness by emphasizing closed-loop control and error correction happening moment-to-moment. 2. What predictive processing gets right. Predictive Processing argues that the nervous system is constantly generating top-down priors about expected sensory states. This isn’t necessarily “future forecasting” in the lay sense. It’s a statistical description of how neural systems reduce surprise and maintain stability. Predictive models are supported by cortical laminar structure (feedback-dominated loops), mismatch negativity electrophysiology, sensorimotor control in fast movements (e.g., saccades), motor planning and anticipatory postural adjustments, cerebellar timing and internal forward models Strength? PP captures how neural systems exploit statistical regularities across time, not just reaction, to stabilize perception and coordination. 3. Why they Ssem to conflict (but don’t)? The conflict arises from conflating two questions, a) Does the organism need an explicit internal model? PCT says no. Behavior can be generated through immediate error correction. PP says yes. Hierarchical priors implicitly encode expectations. b) At what level is “prediction” occurring? PCT focuses on active control loops at the phenomenological and behavioral engineering levels. PP focuses on neural coding strategies, cortical dynamics, and probabilistic inference. Key point. Reactive control and predictive coding can coexist as different layers of the same system. Spinal reflex arcs. PCT-style fast reactive loops. Cerebellar-cortical interactions. PP-style forward models and error minimization. No organism is purely predictive or purely reactive. 4. What the “Hello” experiment actually shows. The experiment demonstrates people can compensate for structured forces without conscious awareness of the structure. Control can succeed without forming an explicit internal model of the disturbance. This supports PCT, but does not refute PP because, sensory prediction errors are still required for control. The controllers could still operate on internal priors about cursor motion or proprioceptive state. The absence of conscious modeling does not imply the absence of implicit modeling in the nervous system. The experiment refutes overly intellectualized, symbolic predictive models, not predictive processing itself. 5. A more integrated scientific view? PCT explains the real-time control architecture. Fast, low-level, negative feedback loops maintaining stability under perturbations, automatically correcting errors without explicit modeling. PP explains the representational and statistical architecture., how the system organizes sensory information, how expectations modulate perception, how the brain compresses and generalizes from past experience, how control targets themselves are selected or updated. An integration hypothesis? The brain uses predictive coding to set reference signals and expected sensory states. PCT-like reactive loops then implement these states in real time. This hybrid model solves robustness (from PCT), generalization and learning (from PP), multiscale control (from integrating both). 6. Why this matters for neuroscience? The controversy dissolves if we recognize that prediction and control are not mutually exclusive. Prediction shapes desired perception. Control maintains that perception against disturbances. For robotics? Robust agents will require both predictive layers for planning, anticipation, and environmental modeling. Reactive control loops for stability and disturbance rejection. For theories of agency? Agency emerges not from prediction or reaction alone, but from maintaining internally specified reference values, updating those values through experience, and stabilizing perception through continuous closed-loop action This aligns beautifully with both the PCT perspective and the deeper generative principles you’re developing.🤔

We've become obsessed with the idea that the brain is a "Prediction Machine." The dominant theory in neuroscience says we're constantly simulating the future, calculating probabilities to guess what happens next. A new paper argues this is a complete illusion. The reality is simpler, and strangely, much more powerful. Here is the argument for Perceptual Control: The "Prediction Illusion" starts with a mistake in observation. When we see someone successfully handle a chaotic environment (like catching a flyball), it *looks* like they predicted the future trajectory of the ball. But observing prediction isn't the same as implementing it. The authors use the perfect analogy: The Watt’s Steam Governor. In the 19th century, this device kept steam engines running at a constant speed. If pressure surged, it slowed the engine. If load increased, it sped up. To an observer, it looked like the machine was "predicting" pressure surges and pre-empting them. But the Governor has no brain. It has no model of the future. It’s a mechanical negative feedback loop. [cite_start]It measures the *current* speed, compares it to the *desired* speed, and adjusts the valve immediately[cite: 80]. It doesn't predict; it controls. This brings us to the "Hello" experiment, which broke my brain a little. Researchers asked people to keep a computer cursor on a target. The computer applied a "disturbance" (forces pushing the cursor away) that the person had to fight against with their mouse. Here's the twist: The disturbance wasn't random. [cite_start]It was an invisible force field shaped like the word "hello" (written upside down and mirrored)[cite: 166]. The participants fought the force, keeping the cursor steady. When researchers looked at the participants' hand movements, they had perfectly written the word "hello". Crucially, the participants had NO idea they were writing words. If the brain were a "prediction machine," it would have needed to model the force to predict the hand movement. But the participants wrote a legible word purely by reacting to immediate error signals—instantaneously correcting the cursor's position. This is **Perceptual Control Theory (PCT)**. The theory suggests the nervous system isn't a linear pipeline (Input → Compute → Output). It’s a closed loop. We act to keep our *perception* of the world matching our internal *reference value*. [Image of Perceptual Control Theory negative feedback loop diagram] Think about catching a baseball. If you were a "prediction machine," you’d calculate the ball's trajectory, wind speed, and gravity, then run to where the ball *will* be. But that’s computationally expensive and error-prone. In reality, fielders just run in a way that keeps the "optical velocity" of the ball constant in their vision. If the ball looks like it's rising too fast, they move back. Dropping? They move forward. No physics calculus required. Just maintaining a visual constant. This solves the "Noise" problem. In predictive models, small jitters in your movement are considered "noise" or errors to be filtered out. It’s the system "feeling out" the environment to maintain control. This has huge implications for AI and robotics. We are currently building robots with massive compute power to "predict" stability. But robots built on PCT principles—like inverted pendulums that just react to maintain verticality—are often more robust and stable than the predictive ones. Why does this matter for you? It changes how we view "agency." We often think we need to predict the outcome of our actions to be effective. [cite_start]But the most efficient systems don't predict the outcome—they specify the goal and let the feedback loop handle the rest[cite: 39]. The "Prediction Illusion" suggests we aren't prophets simulating the future. We are controllers, surfing the present. We don't need to know what the wave will do in 10 seconds. We just need to keep the board steady right now. If you want to dig into the paper, it’s "The prediction illusion: perceptual control mechanisms that fool the observer" by Mansell, Gulrez, and Landman (2025). It’s a dense read, but it completely reframes the "Bayesian Brain" debate. One final thought: Next time you're doing something skilled—driving, typing, sports—notice the difference. Are you calculating what comes next? Or are you just managing the gap between *what you see* and *what you want*? You might find you're doing a lot less "thinking" than you assumed.

keyboard_arrow_left Previous keyboard_arrow_right Next

@_avichawla very insightful

@thekitze 100%

me watching the llm do the job i used to love