Neural Network Pioneers Win Nobel Prize for Transforming AI

The Nobel Prize in Physics 2024 was awarded to John Hopfield and Geoffrey Hinton for their groundbreaking work in artificial neural networks. Their research, inspired by the human brain, laid the foundation for today's powerful machine learning technologies. Hopfield's associative memory and Hinton's method for finding properties in data have revolutionized how computers learn and process information.

1. Natural and
artificial neurons

The brain's neural
network is built from
living cells, neurons,
with advanced internal
machinery. They can
send signals to each
other through the
synapses. When we
learn things, the
connections between
some neurons gets
stronger, while others
get weaker.

NEURON

Artificial neural
networks are built
from nodes that are
coded with a value.
The nodes are
connected to each
other and, when the
network is trained,
the connections.
between nodes that
are active at the
same time get
stronger, otherwi-
se they get

WEAKER weaker.

SYNAPSE

STRONGER

© Johan Jarnestad/The Royal Swedish Academy of Sciences

2. INPUT PATTERN

Memories are stored Whenthetrainednetworkis 29920000000000
ji fed with a distorted or 882220080888
in a landscape Ineomplete stern, can ee

222322888
be likened to dropping a 2930808800
ball down a slope in this

landscape.

John Hopfield's associative memory stores
information in a manner similar to shaping a
landscape. When the network is trained, it
creates a valley in a virtual energy landscape
for every saved pattern.

fee

Johan Jarnestad/The Royal Swedish Academy of Sciences A

2 The ball rots until it reaches a place

where it is surrounded by uphills. In the
same way, the network makes its way
towards lower energy and finds the
closest saved pattern.

ENERGY LEVEL

SAVED PATTERN

20000000000000
99999909999009
88999999999909
3828880:

3. ‘ferent types of network

HOPFIELD NETWORK BOLTZMANN MACHINE RESTRICTED BOLTZMANN.

© ©) o ©

SO visite 4 Hidden

nodes nodes

John Hopfield's associative Geoffrey Hinton's Boltzmann In a restricted Boltzmann machine,

memory is built so that all the machine is often constructed in there are no connections between

nodes are connected to each two layers, where information is nodes in the same layer. The

other. Information is fed in and fed in and read out using a layer machines are frequently used in a

read out from all the nodes. of visible nodes. They are chain, one after the other. After
connected to hidden nodes, which training the first restricted Boltzmann
affect how the network functions machine, the content of the hidden
in its entirety. nodes is used to train the next

machine, and so on.

© Johan Jarnestad/The Royal Swedish Academy of Sciences.