The artificial intelligence It’s 80’s sci-fi set in real life. It is so promising that it seems impossible that there could be something even better. However, a team of scientists from the Johns Hopkins University has just presented a proof of concept with which it demonstrates that the biocomputers controlled by human brain cells they could make the AI we are used to completely obsolete.
It would not be artificial intelligence, but organoid intelligence. It is called this way because what it uses to feed to computers are precisely organoids. That is, cell cultures exposed to three dimensional shapewhich share some properties with the organ to which the cells belong.
In this case they are not exactly mini brainsbut they do have the advantages that the multiple connections between neurons. It is something beneficial; Well, although it may not seem like it, the artificial intelligence we are used to is not as good at learning as human beings. If we add that own capacity, we would be facing something without equal. An achievement that would leave even the ideas of science fiction of the 80s outdated.
Artificial intelligence is not that good at learning
Thanks to artificial intelligence, images can be created out of nothing, like all those that decorate this article recently published in hypertextual. Essays can be written, diseases diagnosed, transport controlled, conversations fully coherent with machines… You can even generate robots capable of win a game of chess to the best players of all time.
To do this, the algorithms are trained with huge databases, in which they search for the coincidences and divergences that finally lead them to obtain a result. All this is what is known as machine learning. But no, artificial intelligence is not as good at learning as human beings. It simply is much faster.
The authors of this study explain it in a statement with a simple example. “AlphaGothe AI that beat the world’s number one Go player in 2017, was trained on data from 160,000 games. A person would have to play five hours a day for more than 175 years to experience so many games.”
That does not mean that a human learns worse, but that he cannot do it so quickly. If we had the ability to handle data at the speed that a computer does, we would not need artificial intelligence. And it is precisely in that midpoint in which organoid intelligence comes into play.
mini brains for teach to a computer
In this proof of concept, the researchers have shown what it would be like to use an organoid based on human brain cells to train a computer.
There would be many advantages over artificial intelligence. The first is precisely what was mentioned before. humans have a better learning ability and that is precisely because how our brain and neural connections work what’s in it If that is implemented in a computer, artificial intelligence can be greatly improved.
On the other hand, human brains are much more energy efficient. This can be seen again with Alpha Go, as the energy that went into training the algorithm would be enough to keep an adult active for a decade.
These are the great advantages of organoid intelligence, but it is impossible to think of a disadvantage. Where do we get so many human brain cells? It is not easy and, according to these scientists, although cultures of 50,000 cellsthe ideal would be to reach the 10 million. But it’s a solvable problem, thanks to something known as induced pluripotency stem (iPS) cells. These are adult cells that are reprogrammed back to how they were in the embryonic stages, when they had not yet differentiated. They then differentiate into another cell type of interest. That is, you can take, for example, skin cells and convert them into brain cells. Thus the problem would be solved and a new advantage would be obtained.
Study of brain diseases
These scientists believe that organoid intelligence could be useful for the study of neurodegenerative diseases such as Alzheimer’s. To do this, skin cells would be taken from healthy and sick patients. Later, through the iPS technique, they would be dedifferentiated and differentiated again to transform them into brain cells.
Thus, they could build two organoids, one healthy and one sick, and compare the differences in the learning of both when connected to computers. It would be a very interesting new approach, although there is still a long way to go to get to that point.
In general, there is a long way to go before this organoid intelligence thing is more than a proof of concept. would have to overcome certain technical pitfalls, such as the increase in the number of connected cells that we have already discussed. But also ethical pitfalls. For example, what rights would the people who donated the cells have over the organoid? Could they ever develop consciousness, even a very rudimentary one?
This and much more is what should be studied from now on. Until then, we have artificial intelligence, which is not little. One only has to see the great advances that it is having in recent years.