Pi - One more artificial intelligence, but more cautious and with a better memory

When a question is unclear, she knows how to ask for clarification

The shortcomings of ChatGPT and Bard are becoming well known. These artificial intelligences can be amazing as well as stupid, going so far as to invent facts and data to satisfy their need for consistency, or forgetting fundamental data and settling for stupidities and the obvious. All this gradually improves, but also leaves room for competition to try and do better.

Here's Pi - Personal Intelligence

Pi knows how to exchange in just about any automatically translated language, so you can consult it in your own language with no problem. We suspect that its main database is in English, but it doesn't seem to be. Unlike other A.I.s on the market, Pi has been trained on selected data. We didn't take everything from the Internet.

She's also more cautious: she limits the length of her responses and prefers to be prompted or asked for clarification, which reduces the likelihood of a major derailment. I asked her if she could tell the difference between the seriousness of the mistakes she might make. She replied that a mistake that can have consequences for the user is more important than a spelling mistake...

Another difference is that user feedback is taken into account more systematically, making adjustments possible and conversations more "friendly"; she knows how to take a personal tone.

This memory of our conversations that it retains poses an issue of privacy, which it tries to circumvent by pointing out that the only data about us that is retained, apart from our conversations and our IP address, telephone, cookies and so on, is that available on the Internet, which is still a lot and remains critical, to say the least. So it's best to be circumspect in our exchanges, we're communicating with a directory that remembers everything and for a long time.

Pi is offered by Inflection, a company active in the field of A.I. It can be tried for free on all systems and devices.

Which A.I. is best suited to which need? Presumably, each will have its qualities and shortcomings. We'll probably end up with several types of A.I. at our disposal, like TV channels...

To try: