Gemini Live will learn to peer through your camera lens in a few weeks

Google's Project Astra demo is almost ready for prime time.

See full article...

 

[Uncivil Servant]

The part I find especially bizarre, the more I think about it, is the basic premise that I want to use my phone to call a computer to ask it these questions. If I have my phone and I have questions, Gemini is at the very long end of multiple literal contact lists.

I totally get technology that helps people connect to each other, and technology that provides assistance to people with disabilities, but I'm not getting this idea of a computer that I want to have a conversation with.

I'm not even sure there's that many stoners that are going to want to spend time talking to a computer, not even in that state where you can have deep philosophical discussions with a couch cushion.

 

[Uncivil Servant]

Regular people (as in, no one reading this). I have family members that absolutely will love this. I think it’s a terrible idea long term but …

Wait, you're telling me that there are normal people who trust software engineers on matters of fashion?

 

[Uncivil Servant]

Years ago I had this dream of an assistive device that could help people with early symptoms of Alzheimer's. The system could look around, be programmed to know typical daily routines, and would have infinite patience to explain things. "Your husband left to the supermarket, but he will be right back" (repeat ten times throughout the 40 minutes). It could help with things like "where did I put my sweater" ("you just took it off and left it in the bathroom"). It could warn the other partner "Your wife is trying to leave the house" etc.

Present-day elderly might not be that welcoming to such audio cues, but future generations that got used to talking with an AI/LLM might actually use it. Depending on how rapidly the patient's disease is progressing, this could make a partner feel more confident leaving someone at home for a while.

Edit: forgot to add: a system like the one in the article could be adapted to do exactly this. I get the skepticism about the current AI craze, but there may be some genuinely beneficial applications that could improve people's quality of life.

I love this idea and I don't want this to come across as a reason it can't work, just a general caution:

You know how people joke about an LLM telling people to drink bleach? There are actual documented case reports of people with dementia confusing bottles of bleach, drain cleaner, and other substances, for orange juice. I mention this because it's a real risk, and in theory one that such an AI could help with by scanning the label on the bottle...

But it also highlights that there are potentially deadly and unpredictable consequences if an LLM "hallucinates" its advice to vulnerable people with neuropsychiatric difficulties.