Researchers are trying to cut the genetic code from 20 to 19 amino acids

That said, the results also show the limits of working with current AI models, mainly because, unlike a human, they cannot fully explain how they make decisions. For example, some of the models made very different ideas from each other, which the researchers say that they are looking at different areas of the space of possible events. But we don’t know if that’s the case, or if each model had mathematical reasons to dislike the other’s ideas.

This is one of several things in the paper where the researchers tried to guess in hindsight what the model was doing based on the results. At the same time, the program revised all the structural elements (alpha helix) isoleucine that was changed was found, for reasons that do not pose a risk.

It’s a good reminder that, at this point, software programs are tools: they allow us to do things that would otherwise be impossible, but they don’t help us understand much. We are still left to reason through events using the neural networks inside our skulls.

This should not be the case; we can place more emphasis on disclosing how the software works when we develop it to inform decision-making. But for now, I think the emphasis has been (rightly) on finding something that works.

A surprising success, but practical?

All in all, this is an amazing job. These proteins have to interact, interact with ribosomal RNAs, transfer RNAs, messenger RNAs, growth proteins that the ribosome produces – including all the good proteins on a large scale. Each of them has had billions of years to evolve the ability to work with each other. The fact that we can change the system so much in a few years is amazing.