Might just be Google. Their AI launch was notoriously bad. Tainted with errors and propaganda. They actually had to apologize for it.

Maybe this is just more of the same.

Oh great, AI gaslighting.

OP, get ready for this to happen, but with your phone hypnotizing you to commit a crime...

It's does save me a lot of time. Going through contracts is some of my least the favorite parts of my job.

Having AI flag areas is huge because my prompts are copy and pasted from a document that I keep updated.

Most flagged items are not deal.breakers just areas that have to be changed or reworded. Often times clients also copy and paste boiler plate paragraphs in their contracts and some things don't apply.

As I read through the contract I will add or delete from what was flagged by AI.

Since I have already been doing this for a while AI knows my conditions as I keep the model updated. Then generates my response to items flagged if I choose to keep them.

The initial flagging just acts like a highlighter pen to double check things but with the ability to generate a response or a fix I choose.

When I use AI I am equally amazed and terrified. It won't be long until a lot of jobs will be eliminated.

I just have the Gemini that came on my Android, I don't pay for any extra features

Asked it to set a 10 minute timer. It responded "I've set an alarm for 10 minutes from now. It will go off at 4:41 PM." At 4:41pm the alarm went off.

You should be able to see the chat history. Post it so we can see what it did.

What "AI" (not really AI) does, is respond to a prompt with a complicated prediction of what the response to that prompt would be. Thats all it does. Inherent in that prediction, it also has to infer or specifically be told what it's own role is.

To think about this, lets say you ask it to "Please give me a weather prediction". What it will respond back to you could be a response to "(How would a National Weather Service meteorologist respond if asked) Please give me a weather prediction" or it could be a response to "(How would the typical dumb idiot respond if asked) Please give me a weather prediction"

So you see those are two very different things it would be asked to predict and reply back to you. Almost all of these AI LLM's will default to the first one and pretend to be an expert first and foremost. So it will almost always reply back to you as if it were a meteorologist when it really isnt, and if it has no knowledge to use in its response, it will make it up with something that sounds like a meteorologist would say as part of its prediction. You can easily ask it the to do the second thing, as well by including the detail about who it is supposed to role play as to get a more customized response, but it wont do that by default because then AI would just be like asking any random idiot something and it would always respond that it "doesn't know", or worse it could purposely try to answer wrong because it thinks that the role (a dumb person) requires it.

So really when it doesnt have tools or thinking logic at its fingertips (some do to some extent) they are almost like psychopaths or scam artists unless you are really very thoughtful with your prompts. They tell you back what they think you want to hear. In your case it basically responded back to you to start with as if it was your Google Assistant, but it has no abilty to reason out that you were probably not aware you were asking a question to it by mistake.

And the elites say this tech will replace most jobs within a few years

Do androids dream of electric sheep? Prepare for the future.

The answer is yes.... yes they do

Not familiar with Gemini

I use Grok

Gemini is trained on Google, which is inherently messed up and corrupt. Gemini will never be a good LLM because Google has spent the last 20 years being a biased search engine.

I don't know why ANYONE would use Gemini when Grok and ChatGPT are superior, though imperfect.

I have had issues with ChatGPT and relatively simple math problems. It spits out a different answer each time when I ask what number is a percentage of another number.