Speech recognition has been successful and can be Xi unsupervised
Abundant color from the temple of Wafei
Qubits | Official account QbitAI
An AI system composed of a "mini brain" and microelectrodes built from real human brain cells has been able to perform speech recognition.
The kind that pinpoints the voice of a particular person from hundreds of sound clips.
Recently, a cutting-edge brain-like study appeared in the journal Nature.
This particular AI system can even perform unsupervised Xi:
The researchers simply played the audio clips over and over again, providing no form of feedback to tell the system whether the answer was right or wrong.
Eventually, after two days of training, the system's accuracy rate went from 51% to 78%.
How is this achieved?
Here comes the organoid neural network
The main purpose of inventing this system is to solve the problem of high energy consumption of silicon chips.
Generally speaking, the solution to this problem relies on brain-like calculations.
However, most of the "traditional" brain-like chips designed under this idea are directly based on the principle of digital electronics, and their ability to completely imitate brain functions is really limited.
Here, the study directly uses something called "organoids":
It refers to miniature organs that can be cultured in the laboratory using human stem cells, containing some of the key properties of the organ they represent.
Specifically, the researchers connected brain organoids (shaped like small balls) made up of living brain cells and high-density microelectrode arrays to build a system called "Brainoware".
The role of microelectrodes in Brainoware is to send electrical signals to organoids to deliver information to the "brain", and to detect the firing response of nerve cells in the brain, which is then handed over to an external device for reading and interpretation.
Such a system can exhibit neural network-like functionality and allow for unsupervised Xi.
Connect it to specific hardware and it can be trained on speech recognition.
In this task, the researchers converted 240 audio clips of eight people speaking Japanese vowels into signal sequences and sent them to the system to recognize a person's voice.
In the beginning, Brainoware was only 30%-40% accurate.
But after two days of training, it was able to identify specific speakers with 78% accuracy.
The authors emphasize here that the so-called training is just repeating audio clips without giving any feedback, which is the so-called unsupervised Xi.
However, it is important to note that at this time Brainoware can only identify who is speaking, but cannot understand any of the speech.
After that experiment, the researchers tried a drug to block the formation of new connections between nerve cells in brain organoids.
It was found that after this operation, the accuracy of the system did not improve in any way.
The authors explain that this shows that Brainoware's ability to learn Xi depends on neuroplasticity.
Will the computers of the future be made up of brains?
In March, the team was already using the system to try to predict Hénon diagrams, a dynamical system in mathematics that exhibits chaotic behavior.
As a result, after 4 days of unsupervised Xi (each day represents a training cycle), Brainoware was found to be more accurate in predicting than artificial neural networks without long short-term memory units.
The latter, by contrast, has gone through at least 50 training cycles.
A little earlier, an Australian research team tried to teach the "brain in the plate" to play a game of table tennis, and it learned it in 5 minutes, 17 times faster than the AI.
So in the future, will computers be made up of brains?
It's hard to say.
As the authors of this article explain, their research is currently a proof of concept, and there are many questions to be solved:
For example, the performance of the Brainoware system can be improved, but the most important problem is that organoids only survive for one to two months.
And, while the Brainoware itself doesn't require much power consumption, the power consumption levels of the external devices that keep it running are not low.
Such as a series of and so on.
Overall, there are scientists who predict that a truly general-purpose biocomputing system could take decades to create.
But in any case, its research will help us to further understand the mysteries of the human brain Xi.
Reference Links:
[1]https://www.nature.com/articles/s41928-023-01069-w
[2]https://www.newscientist.com/article/2407768-ai-made-from-living-human-brain-cells-performs-speech-recognition/
[3]https://www.genengnews.com/topics/artificial-intelligence/brainoware-organoid-neural-networks-inspire-brain-ai-hardware/