-4

Modern computer systems contain only a few hundred thousand neurons and in rare cases, a few million, which, in terms of intelligence, is comparable to a bee.

Ab Bees & AI

How true is this statement? 🤔

Improve this question
6
  • 8
    I don't think "as intelligent as a bee" is an empirical statement that can be tested. The empirical statement is that modern neural nets contain about as many nodes as a bee has neurons, which is not what is asked.
    – Oddthinking ♦
    Nov 12 at 16:03
  • 10
    "Modern computer systems contain only a few hundred thousand neurons". Hum, no, modern computers contain exactly 0 actual neuron, unless some mad scientist somewhere started putting brains in jars. Jokes aside, the scope of the question is extremely vast and cannot be answered succinctly enough for this site. Why? That would require to cover: intelligence and its definition(s), analog vs digital system, abstraction, current computing levels, philosophy of intelligence and reasoning, the biology of a neuron, and probably some more.
    – Entropy
    Nov 12 at 18:02
  • 2
    But does a bee have access to the 175 billion parameters that ChapGT has?
    – Weather Vane
    Nov 12 at 19:28
  • 3
    We are not interested in the fact that the brain has the consistency of cold porridge. - Alan Turing. Comparison of the number of transistors in a chipset or in that amount of memory allocated for fuzzy logic comparisons to the number of neurons in a brain is an apples-to-oranges comparison. Can you give a correlation of number of neurons to intelligence (whatever that is) that is justifiable?
    – doneal24
    Nov 12 at 20:34
  • 1
    @JeromeViveiros, most marvelous! That would depend on quite a number of variables, quite easily deducible from computer-human interaction ... cogito.
    – Agent Smith
    Nov 13 at 12:14

1 Answer 1

Reset to default
1

Bees have 1 million neurons and ~1 billion connections

This is reflected in the complicated brain structure and anatomy, with around 1 million neurons and an estimated billion synapses in the honey bee brain.

GPT-3 is estimated to have several million neurons. Original research: The lower bound of neurons is ~500k, which assumes a fully connected neuronal network, which GPT-3 is not. You arrive at this value by simply taking the square root of the parameters.

GPT-3 has 175 billion neuron connections (= parameters).

So, without claiming anything about intelligence, we have in fact passed the point where machine learning models are bigger than bee brains.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .