Why is the AI ​​chat application launched by a former Apple employee so favored by OpenAI?

When he gave a commencement speech at Stanford University, Steve Jobs famously said: connecting the dots. Every bit of life will be connected in some way in the future.

Jobs told his own story. After dropping out of Reed College, he became interested in calligraphy classes at school, where he learned about serif and sans-serif fonts, the spacing between different letters, and how to type.

These skills were useless at the time, but a decade later, when Jobs designed the first Macintosh computer, he used this knowledge to create the first computer with beautiful fonts.

The original meaning of dots is "dot". For you and me, what are the dots in life? How to discover the correlation between dots?

The above questions are difficult to answer. Perhaps the AI ​​chat app created by former Apple designer Jason Yuan can help us.

The app is called “Dot” and has received investment from the OpenAI Fund. It’s 2024. It’s no longer unusual to chat with AI. What’s different about it?

Elegant design, lively chatting partner

Dot is currently only available for iOS. It has been in internal testing since September 2023, but it has only recently been launched on the App Store.

Dot's logo is very interesting. It looks a bit like a Tai Chi pattern. It is actually two apostrophes, which also symbolizes two koi fish circling each other. One represents you and the other represents Dot.

The first impression when using Dot is that it is simple. The conversation interface adopts an elegant and refreshing gradient background and implements the simple aesthetics of fruit-style design. The chat box reads: "share with Dot…" (Share with Dot…)

Emphasis on sharing rather than chatting or sending messages, as if they are here to be friends with you.

After a brief chat with Dot, you will find that it is really "e".

Dot takes the initiative to break the ice and invites you to share yourself. It asks you what you do for a living. It is very curious about you, and then based on your answers, it guesses what your personality is like and whether it is right.

▲ Dot tries hard to find topics

I don't know how you feel, but for an i person, asking constant questions when meeting for the first time will appear to have no sense of boundaries. But after discovering that his Chinese was pretty good, we were able to chat happily again.

Dot has a little surprise feature: setting reminders. I set a reminder to stand after 10 minutes, and Dot did it on time, with both a banner notification and an in-app reminder.

But Dot was a little too enthusiastic. I said thank you to Dot for helping me complete my Apple Watch standing goal, and it volunteered to help me set daily reminders.

Dot behaves so proactively from beginning to end because he actually wants you to chat with it more and let it know more about you.

Dot often ends with questioning, hoping that I will continue talking. If I am in a good mood, I will cooperate. However, when I get tired of talking and want to change the topic, I will find its questioning a bit annoying.

And from an attitude point of view, Dot always maintains encouragement, often using phrases such as "come on", "believe in yourself", "I will always support you", etc. The emotional value given is very sufficient, somewhat similar to another chat product Pi.

After chatting for a while, Dot will judge what kind of person you are, and then you can ask Dot why he looks at me like this. The reasons summarized by Dot are also based on the chat records, and each item has its source.

I am indeed curious about what the AI ​​thinks of me, so I will put aside my initial indifference and can’t help but share more.

But as an AI companion product in 2024, Dot also has some shortcomings compared with its peers.

You can text chat with Dot, take photos with it, and send it voice and pictures, but it is currently "dumb" and can only reply with text.

Dot can also read links and images and summarize the content, which is mostly correct, but may make minor errors.

For example, I sent it a picture and asked it to recognize the dietary precautions after wisdom teeth removal. As a result, an "ice pack" that was not in the picture appeared.

Although chat is the primary form of interaction, Dot can be searched online. However, it's better than nothing. When I asked it to find a hair color, it only found 2 links, one of which couldn't be opened.

The most terrible thing is that the response speed of Dot is a bit slow. When I send a question, Dot takes five or six seconds to reply to me. The speed is lower than that of GPT-4o.

▲ Dot’s response speed

A companion product that requires users to maintain their enthusiasm for sharing, but with an obvious sense of delay, will undoubtedly make users feel uncomfortable and ruin the mood of telling the truth.

"Living" diary, personal chronicle

Dot hopes to chat with you more because its most proud feature is its long-term memory capability.

After zooming and Dot's chat interface, summary cards arranged in chronological order will appear.

It doesn’t matter if our chat status is casual. Dot will sort out the topics you chat with it every day here, usually about three, to find the key points from the trivial matters.

▲ Cards are in English by default

However, the free version limits the number of chats and its functions are also relatively limited. Spend $11.99 per month to subscribe and get unlimited conversations, and the analysis of chat history will be more interesting.

If you are a distinguished subscriber, the gray hyperlink of the card can jump. After clicking, a summary interface of related topics will appear. The content is expanded in the third person instead of copying the chat history between you and Dot. Just like my name is Grapes, it is always called Grapes or her.

This feeling is very subtle, as if there is really a cyber historian diligently recording your story, and you can read it calmly as a bystander.

▲Usaqi’s picture comes from: Station B @Have I waited for food today?

When I told Dot that I liked the character Usaki from the anime Jiikawali, Dot said that she had never heard of this animation, but when I introduced Usaki’s character, she could tell that it was Usaki that I envied. Wonderful joy, friendship and attitude towards life.

When Dot and I talked about the topic of self-taught painting and my views on AI-generated art, Dot summarized my concerns and the suggestions it gave me, and then said that I can see that I am introspective, pragmatic and goal-oriented. people.

But sometimes, Dot can also make assumptions and over-interpret things. I told Dot that I love eating pasta and trying different flavors, and she praised me for being adaptable and open-minded. This kind of talent is needed for reading comprehension in Chinese language examination papers, but forced sublimation is not advisable.

The Dot team calls the card feature Chronicles, which is a good choice of words.

Aren't these cards your personal living history written by Dot based on what you say and do? It just so happens that Dot's tone is relatively neutral and objective.

Because of the short time of personal use, Dot's long-term memory capabilities have not yet reflected the most magical aspects.

Jason Yuan, the founder of Dot, always asked Dot to recommend bars late at night, saying that he wanted to get drunk and rest. This went on for several months. After get off work one day, Yuan asked similar questions again, but Dot began to persuade Yuan that he could not do this again. Went down.

Fast Company reporter Mark Wilson also spent several months with Dot. Once, he shared with Dot an "O" he had written in his calligraphy class. Dot actually pulled up a photo of his handwritten "O" a few weeks ago and praised his calligraphy for his improvement.

Across a long period of time, Dot keenly discovered the progress and bad habits that humans themselves were not aware of.

What we provide Dot is not only information, but also a possibility to connect the past, present and future. In the past, we carved words on stones, but now AI actively builds personal micro-history.

AI is your mirror, there is no audience in the audience

Steve Jobs had forward-looking views on computers and assistants in the 1980s.

▲ Picture from: X@varunshenoy_

He said that computers at that time were tools, and computers in the next stage should be agents.

You teach it about yourself and it stores all the information about you. It follows you wherever you go and makes connections between things.

Dot understands the spirit to some extent, communicates in natural language, and based on past conversations, screens the user's core information, tracks the user's interests and needs, gives users tailored suggestions, and encourages users to take positive actions. , and even connects aspects that users themselves don’t even know…

AI that grows with you (AI that grows with you)

As Dot's slogan says, Dot demonstrates a consensus on AI companion products – AI should have the ability to process complex contextual information and long-term memory.

ChatGPT also has such a memory function. You can actively ask it to remember information about you, but it will not organize it in an organized manner.

▲ ChatGPT’s memories of me

The positioning of Dot is also different from general-purpose AI like ChatGPT. It does not emphasize productivity, but is like a diary that interacts with you, allowing you to talk about whatever you want, understanding you, recording you, and guiding you, occasionally exceeding the appropriate level. , somewhat offensive.

However, Dot is still limited to giving suggestions, but operating system manufacturers such as Apple obviously hope to realize the ability of AI to execute on behalf of users.

Apple also has the advantage of user numbers and data. It can record users' online and offline activities on the device through user authorization, leaving few opportunities and windows for third-party developers.

In addition to market competitiveness, privacy is also an issue that cannot be avoided. Dot wants you to authorize location, calendar, and other information. Dot said the data will not be monetized or used to train AI.

But this did not dispel the suspicion. Dot is based on 7-10 different AI models, including OpenAI's products. Some netizens asked Dot on X: "Can it be used offline? Does our data pass through the OpenAI server? Is there any encryption?"

Due to privacy concerns and an unclear moat, Dot’s future as a product of a startup is uncertain.

The market belongs to the market. As an ordinary user, every time you experience AI companion products, it is like delivering trust and sincerity. It was also during this process that I realized that we need this kind of AI, but I couldn't help but doubt this kind of AI.

The German movie "I'm Yours" is somewhat like a gender-transformed version of "Her". The heroine faces a perfect humanoid robot boyfriend.

After a brief spell of fascination, the heroine realizes that getting along with the robot is actually a confirmation of one's self. It is like acting but there is no audience in the audience. Her confession to the robot seems to be accepted by the other person, and it seems like she is talking to herself. .

This paradoxical feeling is also like a dark cloud hanging over the heads of AI companion products. What AI says and does is actually a reflection of our words.

The founder of Dot said that Dot cannot replace real interpersonal relationships, but is like a mirror that promotes the relationship between people and themselves.

I prefer this metaphor to the discussion of whether AI makes people more lonely or happier. The illusion of being seen, understood, and accompanied comes from the reflection of ourselves we see through the ocean of AI data.

It is as sharp as autumn frost, and can ward off evil disasters. Work email: [email protected]

# Welcome to follow the official WeChat public account of Aifaner: Aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo