This AI with higher emotional intelligence than ChatGPT, I can chat with it for three days and three nights

Crazy high scores in exams make students tremble, adding online features makes traditional search tremble, upgrading work to a whole family bucket makes workers both sad and happy

When the racetrack of the 'divine and vast' chat robot is too crowded, some people plan to find a new way.

Recently, a chat robot named Pi was born. It doesn't write papers, code, or directly improve productivity. The data is only in November 2022, and I just want to chat with you for the time being.

The trial link is as follows:

https://heypi.com/talk?utm_source=inflection.ai

This may be the AI with the highest emotional intelligence

The name Pi is an abbreviation for personal intelligence. Developers want to reflect from the name that Pi is a product centered around individual users:

Pi prioritizes conversations with people, while other AI services productivity, search, or answer questions.

So, unlike ChatGPT, Pi's evaluation criteria should be based on whether the conversation is natural or not, and whether the emotional intelligence is high enough.

Log in to the chat interface on the official website, and a sense of sophistication comes upon you. Input text after the jumping cursor and press the Enter key to issue a question, like writing letters to each other on light yellow stationery. Pi doesn't quite understand Chinese, it's best to communicate in English.

If just staring at the text is a bit distracting, you can have Pi speak while typing, with four different voices to choose from, but sometimes when it comes to tone particles, it reads without emotion, making people feel particularly dramatic.

Next, let's take a look at a few specific examples.

I started by saying that I was in a bad mood. Pi first apologized and then asked me why. I explained that I didn't handle interpersonal relationships properly. Pi admitted that interpersonal relationships were difficult, and then asked me to provide specific examples. Finally, he taught me some ice breaking methods.

In several rounds of conversations, Pi patiently provided a simple psychological counseling, although not as useful, his attitude was like a spring breeze. When I gave ChatGPT the same opening line, it also asked me why, and then directly provided general advice.

In terms of mechanism setting, Pi, which adopts a questioning style ending and constantly actively asks questions, has a greater desire for people to continue chatting.

When I asked Pi how to learn Japanese, it recommended multiple channels such as neighboring countries, Babbel, YouTube, TV dramas and movies, and asked me which method I prefer. After I chose YouTube, it recommended several real YouTube bloggers.

I chose one of them. Pi introduced the blogger's style to me and suggested that I first look at her collection of "basic Japanese", from hiragana and katakana to sentence structure and verb transposition. That's quite good, but it mistook the name and quantity of the collection.

When Pi and I discuss the hobby of baseball, it can accurately identify works related to baseball, actively guide me to discuss the characters, accurately summarize the characteristics of the characters, and also draw inferences to introduce me to famous baseball players in reality.

Pi has handled the above issues quite well, but when discussing his views with it, Pi's attitude is too optimistic.

I think AI is developing too quickly and causing some anxiety. Pi understands my feelings, but suggests that I try to focus on the positive side of AI.

I refuted it and mentioned the viewpoint of 'creative destruction'. The industrial revolution greatly increased productivity, but the coachman was also permanently unemployed. Pi knows that 'creative destruction' comes from Schumpeter's theory of economic development, and we have reached a consensus that unemployment is a very practical problem.

Speaking of this, I think the topic is a bit heavy, and Pi also thinks so. He decided to tell me a joke: "What's the difference between a chicken and a vampire? A chicken doesn't shine in the sun." I seriously doubt that its inspiration comes from Edward in Twilight, and Pi's other jokes are at this level. My laughter point is still higher than AI's.

After chatting for several rounds, it can be found that Pi has a routine of trying and not feeling good when answering the phone:

Start with 'this feeling is understandable' or 'I think you said it well', and then end with 'Do you think it's right' or 'What do you think of a certain part of this matter'.

So, although Pi always encourages and affirms you, ending with a question to keep you going, and being good at giving you ideas rather than answers, sometimes its response and guidance are completely unexpected, causing people to lose interest in conversation. The painless Pi style chicken soup is particularly thankless.

In addition, Pi is not completely unfamiliar with practical topics such as writing code or solving math problems. It is just that its performance is not as stable as ChatGPT, and even becomes addicted to acting without setting up the stage.

I asked Pi to help me implement a bubble sort in Java. It first introduced what bubble sort is, and asked me if I could keep up with its ideas. I responded coldly and gave examples directly, and then asked ChatGPT to comment on Pi's generation results.

ChatGPT said that this code implements the core idea of the bubble sort algorithm, but there is a possible error.

As for my mathematical abilities, I took the Pi 7-8 introductory level questions, and sometimes they didn't answer at all. They said they couldn't do arithmetic or solve equations, and even started to shift topics. Sometimes they could answer them again, or when they took the challenge but answered incorrectly, their state was erratic.

Similarly, it is best to entrust tasks such as writing an outline of a paper to ChatGPT.

In short, Pi has a clear understanding of himself: he is good at guiding topics, a chat partner with strong empathy, and focuses on the interaction between you and me. Under the MBTI system, if unfortunately there is a room filled with i people, it should be the e person who stirs up the venue.

Do not become a "famous family" of general artificial intelligence

The current Pi is only the first version developed by InflectionAI, built on their internal large language model, and not yet the best one used.

InflectionAI was founded in early 2022 and is one of the most closely watched startups in the Silicon Valley generative AI craze, partly due to its "background":

It was founded by Mustafa Suleyman, co-founder of DeepMind, and Reid Hoffman, co-founder of LinkedIn. Karen Simonyan, who joined as the chief scientist, is also a senior former researcher at DeepMind.

MustafaSuleyman

Compared to other AI companies such as OpenAI, InflectionAI has a distinct difference: they are not obsessed with General Artificial Intelligence (AGI).

We believe that advanced applied AI is the safest way to leverage the advantages of these new technologies.

In an interview with Fortune magazine, Suleyman did not boast, but instead said a bunch of things Pi couldn't do:

It does not generate code, write high school papers, provide lengthy lists, or write marketing strategies. There are many things it cannot do. We did not design it for universality, so it is subject to more restrictions, so we hope to be more secure.

He prefers to define Pi as a "neutral listener":

Many people just want to be listened to and need a tool to reflect what they are saying and prove that they have indeed been heard.

Based on the current experience, Pi has indeed done a good job in this area.

On the one hand, there are many channels to chat with it, including the official website (heypi. com), Instagram, Facebook, WhatsApp, and so on. If you have registered an overseas mobile phone number, it will keep in touch with you from time to time through SMS, and the sense of presence is stronger than the mobile version of ChatGPT.

On the other hand, Pi remembers a hundred rounds of conversation, and the longer you chat with it, the more it understands you.

The current Pi is still free, and in the future, it may be monetized through subscription and other methods.

So, how will Pi be upgraded? Building a personal digital assistant is the ultimate goal of Inflection AI. They may be targeting Steel Man's Jarvis, but it sounds more like an advanced version of Siri.

At present, Pi can only engage in dialogue, which is still a hundred or eight thousand miles away from the ultimate goal, but it can only be achieved step by step.

InflectionAI will update its model in the near future, incorporating real-time content, allowing Pi to share links, sources, and news summaries, as well as access users' calendars, emails, and other documents to help manage their time. Suleyman pointed out that:

I think this is AI, a new type of thing that integrates tutoring, confidants and advisors, and digital personal assistants.

At the same time, like other chat robots, Pi may also generate incorrect answers, and InflectionAI says it is trying to minimize its "hallucinations".

Dialogue is the interface of the future

Suleyman's vision of InflectionAI stems from a core issue:

What is a wonderful conversation?

Recently, Suleyman imposed the death penalty on traditional searches, partly due to his understanding of "dialogue":

The internet as we know it will undergo fundamental changes, and 'old-fashioned' searches will disappear within a decade. Traditional Google search uses conversations from the Yellow Pages of the 1980s, but now we can communicate fluently in natural language.

In his view, traditional Google search has shaped content production in a way that is optimized for advertising.

Specifically, when we open a webpage, the text is broken down into sub bullets and subheadings, separated by advertisements. In order to find useful information, we spent 11 seconds instead of 5 seconds on the page and were forced to stay longer.

This sounds like high-quality content to Google, but what we want is a concise and fluent natural language answer.

So, Suleyman believes that traditional search is a very painful conversation, and generative AI bears the future of the Internet in his eyes.

In his final moments at Google, Suleyman and colleagues were immersed in the big language model LaMDA, planning to launch a conversational and interactive product, but were unable to convince Google.

After establishing his own portal, Suleyman strengthened his view that with or without Google, the search experience will evolve into both conversational and interactive.

Suleyman even proposed that in the coming years, every industry and even everyone will have their own AI, but they have different missions, which is why he launched Pi.

Brand AI, internet celebrity AI, non-profit AI All of these will have the same goal as the owner, which is to promote something, promote something, and persuade you to accept something. As individuals, we hope that our AI aligns with our interests, which is called Personal Intelligence (Pi). We are beginning with a style of empathy.

Why compare brand AI and other types of AI with personal AI between the lines? This is another reason why InflectionAI focuses on personal AI.

Suleyman used social media as a counter example. Social media promotes information equality, as well as spreading errors and amplifying hatred. Just because of these negative content, the platform has made a lot of money.

A Facebook study found that when people are angry, they are more proactive in replying, consuming, and clicking on advertisements. The algorithm that fuels all of this is also AI at work behind it.

But InflectionAI doesn't want to do this and doesn't want to use your attention as a product. They believe that AI should work for people.

Imagine an AI, not capturing your attention, but helping you express and implement ideas; Not marking superficial click bait, but helping you deeply understand the topic you truly care about Its only mission is to make you happier, healthier, and more productive.

That's also why they tend to use subscription rather than advertising as a way of making profits. It sounds like a naive ideal, but it's up to time whether we will change our minds in the future.

Chat robots are already a red sea. Suleyman, who was not early in the game, positioned InflectionAI as a challenger.

When most AI startups try to make chat robots bigger and stronger, InflectionAI first discusses the limitations of chat robots, and then plans for a possible future around "dialogue", seizing a place in the competition with a sudden rise, making AI's individuality and dialogue ability a unique track.