cryptocurrency

Simulations in quantum computing put researchers deep in the mind to shame

deep mind

Quantum simulation that occurs on a large scale is beyond human comprehension

In what appears to be a feud between two sets of opinion, quantum simulations may have opened the doors to a debate about the extent to which AI can generate accurate results, given the volume of computations that require millions of predictions beyond human understanding. Recently, Alphabet’s AI subsidiary, Deepmind, released a paper titled “Pushing the Frontiers of Density Functions by Solving the Partial Electron Problem”, concluding with the potential for simulations at the quantum scale. Eight months later, a group of Russian and South Korean researchers expressed concerns that the paper’s conclusion might be wrong, putting the entire DeepMind theory to shame.

Physical matter consists of a microcosm of particles numbering in the millions, and it has a number of interactions, which means that studying the interactions at the quantum level is next to impossible. The complexity increases as the particles continue to be added and thus the difficulty of predicting the energy level at which the electron ends up. The researchers believe that the new model could help find ways to manipulate the building blocks of matter through artificial intelligence. In the current paper, the researchers claim to radically improve functionality – which earlier were only like guidelines – by developing neural networks to predict the quantum behavior of electrons and other atomic particles. A statement from the Deep Mind blog says, “By expressing the function as a neural network and incorporating these subtle properties into the training data, we learn functions that are free from important systematic errors—resulting in a better description of a broad class of chemical reactions.”

But why do the conclusions not hold? Apparently, the Russians are not happy with the training procedure, which clearly lacks transparency, because the Russian scientists could not understand how the neural network could come to this conclusion. They suspect the reliability of the neural network function of finding the probability of any electron at each site as in the DFT (density functional theory) proposed in the 1960s, for which scientists won a Nobel Prize. Researchers have been stuck since explaining the partial charge system, and the partial spin system being insufficient to reach such a revolutionary conclusion. They accuse the DeepMind researchers of teaching the neural network to memorize answers to specific questions that are generally asked in the measurement process. The research boasted improvements to the previous DM21m, a neural network model for mapping electron density to chemical reaction energy. However, the Russians could not find a good reason why the improved DM21 outperformed. In the comment paper, they stated, “In our opinion, the improvements in DM21 performance on the BBB test data set relative to the DM21m may be due to a more realistic reason: an unintended overlap between the training and test data sets.”

In a quick and ready response, posted on the same day as the comment, DeepMind clarified the validity of the DM21 improvised neural network, noting technical details and saying: “We disagree with their analysis and believe the points raised are either incorrect or not relevant to the main conclusions of the paper and the overall quality assessment of DM21″. While the Russian response awaits, we leave an interesting turn – can AI be trusted in its predictions – especially when major companies with commercial interests are involved in its development? Does AI have to remain under the umbrella of secrecy in order to garner industry support? Although open source appears to be a discouraging factor for individual interests, it may be a blessing in disguise in the long run, giving way to higher usability and strong interoperability for AI/machine learning models.

A post-simulation appeared in quantum computing that puts researchers in the depths of the mind to shame first.

#Simulations #quantum #computing #put #researchers #deep #mind #shame

Related Articles

Leave a Reply

Your email address will not be published.

Back to top button