Timoleon (Timos) Moraitis

@kalomaze It feels like everything fun was defined in the 90s and now we are just scaling

@SuryaGanguli @noemon_ai Yes, that older work is public github.com/NeuromorphicCo…

Our new paper: "Solving adversarial examples requires solving exponential misalignment", expertly lead by @AleSalvatore00 w/ @stanislavfort arxiv.org/abs/2603.03507 Key idea: We all want to align AI systems to human values and intentions. We connect adversarial examples to AI alignment by showing they are a prototypical but exponentially severe form of misalignment at the level of perception. The fact that adversarial examples remain unsolved for over a decade thus serves as a cautionary tale for AI alignment, and provides new impetus for revisiting them. We shed light on why adversarial examples exist and why they are so hard to remove by asking a basic question: what is the dimensionality of neural network concepts in image space? For ResNets, and CLIP models, we show that neural network concepts (the space of images the network confidently labels as a concept) fill up almost the ENTIRE space of images (~135,000 dimensions out of ~150,000 for ImageNet & ~3000 out of 3072 for CIFAR10). In contrast natural image concepts are only ~20 dimensional. This indicates exponential misalignment between brain and machine perception (neural networks perceive exponentially many images as belonging to a concept that humans never would). This also explains why adversarial examples exist: if a concept fills up almost all of image space, ANY image will be close to that concept manifold. We further do experiments across > 20 networks showing that adversarial robustness inversely relates to concept dimensionality, though the most robust networks do not completely align machine and human perception. Overall the curse of dimensionality raises its ugly head as an impediment to both adversarial examples and alignment: if can be difficult to get AI systems to behave in accordance with human intentions, values, or perceptions over an exponentially large space of inputs. See @AleSalvatore00's excellent thread for more details: x.com/AleSalvatore00…

Models trained with certain biologically-inspired local learning rules rather than backpropagation can be MUCH more robust to adversarial attacks, as we have shown in the past. Our view at @noemon_ai is that the locality of Hebbian plasticity in certain architectures leads to more compositional features, which in turn helps address the binding problem and adversarial robustness. In this new framework of "Perceptual Manifold" (PM) by @AleSalvatore00, @stanislavfort and @SuryaGanguli, our models would have measurably smaller PM than the backprop-trained ones.