‘AI baby’ is born that can think and learn just like a three-month old human07/12/2022
A new artificial intelligence is able to learn and think just like a human infant, in a major breakthrough for the technology.
The deep learning tool, called PLATO, is able to understand basic rules of reality, such as gravity, in the same way as humans.
It was created to understand 'intuitive physics', or how objects interact in the physical world. For example, infants are able to understand very quickly that if a ball is dropped it will fall to the floor, while AI has previously struggled to grasp these rules.
READ NEXT: You can stop brushing your teeth thanks to nanobots that floss, brush and rinse all at once
PLATO was trained with videos of balls falling to the ground and bouncing off of each other and then was shown 'impossible' scenes with nonsense physics.
The AI was reportedly 'surprised' just like a real baby, and was able to learn in just 28 hours, according to a research paper in the Nature Human Behaviour journal.
While AI models can be easily trained on very specific tasks and perform these effectively, teaching them how to learn in a 'generalised' way has proved tricky for scientists.
For example, teaching an AI to recognise human faces requires training data made up of hundreds of thousands of examples of faces, and even then it makes mistakes.
Or, in this case, training it to understand that a ball will bounce when it lands on the floor would require researchers to feed it thousands of videos of balls.
Ditch the iPhone – metaverse contact lens puts the Internet inside your eyeballs
Children, on the other hand, are able to learn what a face is or whether a ball will bounce extremely quickly, as the human brain is able to generalise about new encounters based on a small amount of prior information.
The race to create a 'general' artificial intelligence is very much on among top Silicon Valley tech companies.
Last month, eccentric Google engineer Blake Lemoine was suspended from the company after publicly claiming that its top-secret LaMDA AI had become sentient, calling it a 'sweet kid'.
- Korean cops to stop crime with 'dystopian' Iron Man police armour and robot dogs
- AI could do better job of managing country than Boris Johnson
- God of War fans blown away by new Ragnarok trailer as hit title gets release date
- 'Creepy' iPhone page shows you which apps are tracking your Internet history
- Chinese government uses mindreading AI to 'test loyalty' of Communist Party members
Source: Read Full Article