> Zelnick argues it's really just running a predictive model based on patterns it's already seen
Right conclusion, wrong argument. Brains are also predictive models based on patterns they've already seen. AI isn't human-level yet because brains have several orders of magnitude more weights, constantly train by self-optimising their circuitry, and potentially even use quantum effects to compute. Oh, and they have lifetimes of real training data in a human body. All that on less power than a 20" monitor. Good luck catching up with that.
> Zelnick argues it's really just running a predictive model based on patterns it's already seen
Right conclusion, wrong argument. Brains are also predictive models based on patterns they've already seen. AI isn't human-level yet because brains have several orders of magnitude more weights, constantly train by self-optimising their circuitry, and potentially even use quantum effects to compute. Oh, and they have lifetimes of real training data in a human body. All that on less power than a 20" monitor. Good luck catching up with that.
> Right conclusion, wrong argument. Brains are also predictive models based on patterns they've already seen.
Human brains are more than just predictive models, as far as we can tell.
The brain also performs a wide range of other processes, including:
- Perception
- Attention
- Memory
- Decision-Making and Planning
- Emotion & Motivation
- Language and Communication
- Social Cognition
- Motor Control
- Imagination & Creativity
- Self-Awareness and Metacognition
Prediction is one mechanism that supports perception, action, and learning, but the brain also feels, remembers, plans, imagines, and relates.
Did an AI write this?
That's not how I meant 'predictive model'. But certainly there's more to cognition than inferring state X' from X.