A brain-computer interface (BCI) is a technology that translates brain signals into commands to interact with devices. To this aim, portable neuroimaging modalities, such as functional near-infrared spectroscopy (fNIRS), are particularly appropriate. It is worth noticing that optical signals in the near infrared spectral range can provide information also regarding rapid neural optical properties modulations associated with their depolarization, which are known as fast optical signals (FOS). FOS are characterized by a high spatiotemporal resolution, but also by a poor signal-to-noise ratio, restricting their applicability in BCIs. In order to enhance the feasibility of employing FOS in BCI, approaches based on artificial intelligence (AI) could be beneficial given its capability to identify structures in the data. In this study, an AI approach was employed for FOS identification. Specifically, FOS were obtained using a continuous-wave near-infrared system from the visual cortex in response to a visual stimulus. A photon count, using Direct Current (DC) light intensity at 830 nm, was integrated with an AI methodology for the rapid assessment of visual cortex brain activity. Particularly, a Long Short-Term Memory (LSTM) model was used at the subject level to the optical data, using its capacity to describe temporal dependencies across several acquisition channels for the categorization of visual cortex activity. An average accuracy of 62.8% accuracy was achieved in distinguishing cortical activity from rest across participants. This approach is a first endeavor to detect visual cortex activity using FOS, facilitating the development of FOS-based BCI systems.