News Researchers built a molecular film that stores 16,384 states — the team used it to create an analog computer that works like a brain

Classical computing has 2 possible states and quantum computing has 16 possible states if you have four qubits — but this one has 16,384 possible states.
Gargling the marketing blurb a bit too hard here. A binary system with 4 bits will also have 16 possible states. A binary computer with 14 bits will have 16,384 states.
Then you have trinary computers, where a bit can have 3 states. You have NAND cells, which can have (today) 32 possible states per cell (PKC). You have encoding states like QAM where a single symbol can have hundreds of states (e.g. 32768-QAM has, as expected, 32,768 possible states).

As per the original paper, all the "16,384 possible states!" sillyness is just because their ADC sampling precision was 14 bits.
 
Gargling the marketing blurb a bit too hard here. A binary system with 4 bits will also have 16 possible states. A binary computer with 14 bits will have 16,384 states.
Then you have trinary computers, where a bit can have 3 states. You have NAND cells, which can have (today) 32 possible states per cell (PKC). You have encoding states like QAM where a single symbol can have hundreds of states (e.g. 32768-QAM has, as expected, 32,768 possible states).

As per the original paper, all the "16,384 possible states!" sillyness is just because their ADC sampling precision was 14 bits.
And to suggest that a 14bit storage/processing system is somehow competitive with QC badly misunderstands why Quantum Computing is so powerful.
 
  • Like
Reactions: bit_user
You have NAND cells, which can have (today) 32 possible states per cell (PKC).
My thoughts, exactly. NAND is an apt point of comparison, where up to 5 bits per cell have been demonstrated to work at production-grade reliability levels. IIRC, we've even seen robust demonstrations of 7 bits per cell, on industrial-focused implementations using older, low-density process nodes (note: this test chip also used a very old process node, as is typical of research chips).

Thanks for that. I followed the link in the article (which went to another news article about the paper), but didn't take the additional step of looking at the nature.com link, since I knew it'd be paywalled. However, I just now followed your link and I'm glad I did because the abstract does a far better job of characterizing their accomplishment than either Toms' or the TechXPlore news articles about it.

From the paper's abstract:

"Here we introduce an analog molecular memristor based on a Ru-complex of an azo-aromatic ligand with 14-bit resolution. Precise kinetic control over a transition between two thermodynamically stable molecular electronic states facilitates 16,520 distinct analog conductance levels, which can be linearly and symmetrically updated or written individually in one time step, substantially simplifying the weight update procedure over existing neuromorphic platforms. The circuit elements are unidirectional, facilitating a selector-less 64 × 64 crossbar-based dot-product engine that enables vector–matrix multiplication, including Fourier transform, in a single time step. We achieved more than 73 dB signal-to-noise-ratio, four orders of magnitude improvement over the state-of-the-art methods, while consuming 460× less energy than digital computers."

So, essentially, their goal wasn't to build a denser memory, but rather to build a better analog computing system suitable for accelerating neuromorphic computations (which is a key detail, since it requires only approximate accuracy).

Analog computing is far from new. The efficiency benefits are well-known. Credit to the researchers for pushing the state of the art, in this area.