In June 2021, scientists at the AI lab DeepMind made a controversial claim.
They titled their paper on the subject: Reward is Enough.
The team argued that AGI could emerge through an incentive mechanism known as a reward function.
Their claims have been dismissed by some scientists, but they nonetheless shine a spotlight on a powerful technique.
What is reinforcement learning?
In reinforcement learning (RL), a software agent learns through trial and error.
When it takes a desired action, the model receives a reward.
40% off TNW Conference!
The technique can be applied to a vast array of tasks, fromcontrolling autonomous vehiclestoimproving energy efficiency.
But its most celebrated achievements have come in the world of games.
In March 2016, the technique had a landmark moment.
The victorywas reportedly watchedby over 200 million people.
During the match, the AI played unconventional moves that baffled its opponent.
The final version of AlphaGo does not use any rules, said Demis Hassabis, DeepMind co-founder and CEO.
This means it is free to learn the game for itself, unconstrained by orthodox thinking.
These constraints were replaced by reward maximization.
How a reward function works
Rewards are common learning incentives for animals.
A squirrel, for instance, develops intellectual abilities in its search for nuts.
A child, meanwhile, may get a chocolate for tidying their room or a spank for bad behavior.
(Dont worry, I dont have kids).
In AI systems, the rewards and punishments are calculated mathematically.
These signals allow the agent to evaluate its performance.
Precups colleagues are now developing multi-purpose RL agents.
Eventually, the lab believes such agents could solve multiple problems in the real world.
There are still major challenges to overcome.
RL agents struggle to maximize rewards in complex environments and assess the long-term repercussions of their actions.
Nonetheless, the reward-is-enough proponents believe the algorithms adaptability could pave a path to AGI.
Story byThomas Macaulay
Thomas is the managing editor of TNW.
He leads our coverage of European tech and oversees our talented team of writers.
Away from work, he e(show all)Thomas is the managing editor of TNW.
He leads our coverage of European tech and oversees our talented team of writers.
Away from work, he enjoys playing chess (badly) and the guitar (even worse).
