正在加载视频...

视频加载失败

Neuralink just released a major new update on how it is training brain-computer interfaces Participants have now generated more than 50,000 HOURS of unlabeled neural data Neuralink is now pretraining neural encoders on thousands of hours of each participant’s brain activity before using them for cursor control And the...

18,574 次观看 • 2 天前 •via X (Twitter)

10 条评论

daniel warren 的头像
daniel warren2 天前

On the motor-cortex implants, some participants map imagined finger movements to keys and have reached about 40 words per minute. That is spelled output from movement intent, not overheard inner speech.

FluxGravitas 的头像
FluxGravitas2 天前

Same playbook as LLMs: pretrain on everything, fine tune for the task. Calibration collapsing from 55 minutes to 10 is few shot learning for brains. The mind is becoming just another modality.

Wesley Parker 的头像
Wesley Parker2 天前

Give them LSD and watch the magic

Mukeesh ML 的头像
Mukeesh ML2 天前

The speed is massive

Jory Bicknell 的头像
Jory Bicknell2 天前

I'm excited to test this!

Alp Urungu 的头像
Alp Urungu2 天前

!

Somto 🃏 的头像
Somto 🃏2 天前

This is insane. 50k hours of brain data is like building ChatGPT but for the mind

Gereksiz 的头像
Gereksiz2 天前

Fifty thousand hours of unlabeled data pretrains the encoder on each person’s own signals before cursor control. Cutting calibration from 55 to 10 minutes shows stability. 11.32 bits/s is a new speed record; cross-person generalization is still a goal, not a result.

BraveTom 的头像
BraveTom2 天前

This is very interesting 🤔

Macro Bombastic 的头像
Macro Bombastic2 天前

tbh the calibration drop is the real story here

相关视频

The Mathematics of Moving a Cursor with Neural Signals What might Neuralink Neuralink be doing Mathematically? Consider the task of moving a cursor without touching it. The machine is not looking for a full thought, a sentence, or an image. For this Control problem, the useful object is an intended movement state. sₜ = (pₜ, vₜ) Here, pₜ is the cursor position at time t, and vₜ is the velocity the user is trying to express. The implant records neural activity through many electrode channels, then the decoder tries to estimate vₜ from that activity. Neuralink’s PRIME material describes the N1 Implant as recording and transmitting brain activity with the goal of enabling computer control. For channel i, a simple population model is rᵢ(t) ≈ bᵢ + aᵢ max(0, dᵢ · vₜ) + ηᵢ(t) where rᵢ(t) is the measured activity, bᵢ is baseline activity, aᵢ is channel gain, dᵢ is the channel’s preferred movement direction, and ηᵢ(t) is noise. One channel is not the command. The useful signal is the pattern across many channels: rₜ = (r₁(t), r₂(t), …, rₙ(t)) The decoder subtracts the baseline vector b and applies a learned map W: v̂ₜ = W(rₜ − b) This gives an estimate of the intended velocity. The cursor then updates by pₜ₊₁ = pₜ + Δt v̂ₜ This is the loop shown in the render: neural activity -> decoded velocity -> cursor motion The cortical network and electrode threads show the measurement side. The N1 Implant is described as using 1,024 electrodes distributed across 64 flexible threads, each thinner than a human hair. The decoder panel shows the computational side with activity rₜ, decoded velocity v̂ₜ, and the cursor state pₜ changing over time. A noisy biological pattern becomes a state estimate. That estimate becomes motion on a screen. Therefore, the first lesson is not that Neuralink makes the brain a screen. For cursor control, the Mathematics is more precise: A small piece of intention is represented as a hidden state, measured through neural activity, decoded as a vector, and turned into action. #Neuralink #BrainComputerInterface #NeuralEngineering #Mathematics #StateEstimation #Neuroscience #MachineLearning #BiomedicalEngineering

Mathelirium

14,520 次观看 • 5 个月前