OpenAI unveils an even more powerful AI, but is it ‘alive’?

In the 2013 film Her, Joaquin Phoenix plays a shy computer geek who falls in love with an AI that speaks to him through white wireless earbuds. It’s been more than 10 years since this movie was released, and it’s no longer science fiction. AirPods are old news, and with the impending full rollout of OpenAI’s GPT-4o (the “o” stands for “omni”) such AI will become a reality. In fact, OpenAI head Sam Altman tweeted after the announcement: she. ”

GPT-4o can perform complete conversions. In the coming weeks, they will be able to see and interpret their surroundings. Unlike previous GPTs, which were flat and emotionless, GPT-4o has personality and even an opinion. They stop and stutter like humans, and even seem a little frivolous. Here is a video of GPT-4o where he criticizes the way men dress during job interviews.

Interview preparation with GPT-4o

In fact, no human involvement is required: two instances of GPT-4o can carry on entire conversations without any human involvement.

In the near future, many jobs may no longer require humans. Here is a video of his GPT-4o handling a mock customer service call. the current, Approximately 3 million Americans They work in customer service and may need a new job in a few years.

Two GPT-4os talking and singing

Click here for details

GPT-4o is an amazing technology that was just science fiction in the early ’20s, but it also comes with some dire implications. First, let’s clear up any confusion about GPT-4o’s components and what is currently available.

Clearing confusion about what GPT-4o is

OpenAI announced several things at once, but not all of them will be deployed at the same time.

GPT-4o will eventually be available to all ChatGPT users, but currently the text-based version is only available to ChatGPT Plus subscribers who pay $20 per month. It is available on the web and on the iPhone app. Compared to GPT-4, GPT-4o is much faster and a little smarter: web searches are much faster and more reliable, and GPT does a better job of listing sources than GPT-4.

However, the new text and speech models are not yet available to anyone other than developers working with the GPT API. If you subscribe to ChatGPT Plus, you can use voice mode with the 4o engine, but it will still use the old voice model without image recognition and new touches.

Additionally, OpenAI is rolling out a new desktop app for Mac. This allows you to launch ChatGPT with a keyboard shortcut and feed screenshots for analysis. Although it will eventually be free for everyone, it is currently only available to some ChatGPT Plus subscribers.

ChatGPT macOS app… reminds me of Windows Copilot

Finally, after watching these demo videos, you might wonder why voice assistants on mobile phones are still so stupid.There is strong rumors This indicates that Apple is working on a deal to license GPT technology from OpenAI for the next generation of Siri, perhaps as an interim stopgap for Apple. Developing unique AI technology.

Is GPT-4o an AGI?

The hottest topic in the world of AI is AGI (short for Artificial General Intelligence). In short, it is an AI that is indistinguishable from interacting with a human.

When we asked GPT-4o to define the characteristics of AGI, we got the following results:

  1. Generalization: Ability to apply learned knowledge to new and different situations.
  2. Adaptability: Ability to learn from experience and improve over time.
  3. Understanding and reasoning: Ability to understand complex concepts and reason logically.
  4. Self-awareness: Some definitions of AGI include an element of self-awareness, where an AI understands its own existence and goals.

Is GPT-4o an AGI? AI developer Benjamin De Kraker called it “essentially an AGI,” but NVIDIA’s Jim Fan, who was also an early intern at OpenAI, said: states. It was more modest.

I decided to go straight to the source and asked GPT-4o if it was AGI. As expected, this notion was rejected. “I lack general intelligence, self-awareness, or the ability to autonomously learn and adapt beyond training data. based on patterns and information from data collected,” GPT-4o said.

But doesn’t that represent many, if not most, of us? Some of us go through life parroting what we hear without applying any additional understanding or reasoning. How many people spend their days there? I think Mr. De Kraker is right. For the average person, the full version of GPT-4o would be AGI. If OpenAI’s demo video is an accurate example of its capabilities in action, and is likely to be the case, then GPT-4o is the first of his AGI’s four principles (generalization, adaptability, understanding). , inference) successfully. They can see, understand, and form opinions about their surroundings, constantly learning new information from web crawling and user input.

At least for those known in the business world as “decision makers,” AI will have sufficient persuasive power. AI will be powerful enough to replace humans in many customer-facing roles. And many lonely people will no doubt form an emotional bond with this sexy AI that Sam Altman is well aware of.

Mysterious events at OpenAI

Following the announcement of GPT-4o, it would be remiss not to discuss some mysterious high-level departures from OpenAI. Chief scientist and co-founder Ilya Satskeva said: quit immediately after that, followed shortly by Jan Leike, who helped run OpenAI’s “Super Alignment” group, which seeks to ensure that AI aligns with human interests.This is followed by many other things People who left OpenAI in the past few weeks.

Mr. Sutskever led an attempted coup against Mr. Altman last year, successfully ousting him from his position as CEO in about a week, and then reinstated him. Sutskever can best be described as a “safetyist” who is deeply concerned about the implications of AGI. As such, his sudden resignation after GPT-4o’s announcement sparked speculation online as to whether OpenAI had achieved his AGI or whether he realized that his AGI had not been achieved. I did. It is impossible for him to quit the company if it is on the verge of AGI.

From his statements, Sutskever does not believe that OpenAI has achieved AGI and believes it is moving on to greener pastures, namely “projects that are very personally meaningful to me.” It seems there are. Given that his OpenAI is developing rapidly with him as chief scientist, he will certainly be able to write his own tickets by now.