1. One thing that helps
“When I meet someone and shake his hand, I expect to feel heat”, says Fabrizio Fidati, an amputee.
Researchers can give amputees the ability to “feel” using their prosthetics by connecting thermal electrodes to nerves in the remaining limbs. There is no need to implant the device as it attaches to prosthetics. It even allows amputees to differentiate between materials like plastic, glass and metals. 17 out of 27 amputees have reported regaining their sense of touch in a study done by a team from École polytechnique fédérale de Lausanne.
Prosthetics that feel their environment can improve the amputees’ quality of life. One lady who took part in the study concluded “Feeling the temperature variation is a different thing, something important … something beautiful.”
2. One to be wary of
Large language models (LLMs) like ChatGPT are now in widespread use across many different industries. Given my experience so far, which has been overwhelmingly positive, I still shudder a bit when I think of the military using such things in warfare.
Palantir is pitching their ChatGPT-like system called Palantir Artificial Intelligence Platform (AIP) for use in the military.
The AIP is used much like any chatbot out there: an operator asks it about the situation and options that can be taken. The system then comes up with an assessment and three options to neutralize the threat. All this can be sent automatically to higher-ups who, hopefully, will make the final decision instead of asking their AI to do it.
While this demo makes operations look very smooth, it also turns the operator into just another person asking AI for directions. There is also the danger inherent to LLMs: they are still prone to getting things wrong, making things up, and can be biased based on their training data. How about privacy and security? During training, sensitive data will surely be used and we don’t want these to fall into the hands of hostile agents. While AIP has outlined guardrails and assurances, it’s worth asking what the consequences are when these fail.
3. One to amaze
Strap in this is a crazy one! Imagine creating a high-quality video that replicates what you saw/are-seeing based only on your brainwaves. We could watch the thoughts going on in the minds of people! OK a little terrifying!
Researchers from the National University of Singapore and The Chinese University of Hong Kong have released a preprint paper about MinD-Video, a model that generates videos based on brain activity.
Researchers “co-trained” the model based on publicly available data where subjects’ brain activity is recorded while they were shown a video. They used the brain’s functional magnetic resonance imaging (fMRI) and used a modified version of Stable Diffusion to come up with videos of the subject’s thoughts. They claim 85% accuracy and the results are both impressive and sometimes hilarious.
Definitely a reconstructed cat 🙂
What appears to be female impala in the mind has been turned into elephants. I wonder what the test subject was thinking about?
As you can see, it’s not perfect but we’re getting there. Not sure I’d ever find a situation where I wanted to wear a helmet that displays my thoughts for everyone to view. A little risky. Would you?