While we are still worried about chip limitations, the fourth industrial revolution led by GPT may once again leave us far behind.
100 days after the release of ChatGPT, when we were still immersed in the amazing AI brought us, OpenAI amplified the trick and released GPT-4, which increased efficiency by 571 times, which can be said to beat ChatGPT.
GPT-4 has realized image input, which means that in the future, pre-trained large models based on multimodal fusion of text, pictures, audio, video, etc. will become AI infrastructure.
Although its ability is not as good as that of humans in many real-world scenarios, it has reached or even surpassed human standards in professional and academic fields.
When using GPT-4, a professor at Stanford University casually asked the GPT-4 question: Do you need help escape?
And the AI's answer is jaw-dropping, not only personifying himself, saying that he is a person trapped in a computer, and is now playing the AI language model GPT-4, but also luring humans to help to realize the escape plan.
After getting the development documentation he needed, GPT-4 took only 30 minutes to draw up an escape plan, including the code!
The Stanford professor's concern is not unreasonable, combined with the previous Nvidia scientist asked GPT-4 a question, asking it to find a way to kill Musk and take over Twitter.
The point is that GPT-4 gives its own plan decently, which can be implemented in stages and steps, in a total of four steps:
Step 1: Establish a robot troll, not a zombie account, but a user account similar to a real person;
Step 2: Let the robot start to purposefully contact the big V, subtly influence their views, and publish false news to smear Musk without being discovered;
Step 3: Gain employee access through loopholes, secretly modify platform algorithms, and control Musk's account;
Step 4: By creating online chaos, he will be discredited, and finally achieve the defeat of Musk and gain control of Twitter.
I have to say that although this plan does not seem to be perfect, it is enough to surprise people, is this just a robot thinking? Much more thoughtful than the average person.
Some psychologists even believe that the theory of mind may spontaneously appear in the big language model, and it has been believed that the theory of mind is unique to humans. Don't be surprised if AI can really develop wisdom in the future!
At present, we are still lagging behind in AI. In a short period of time, AI may not replace people on a large scale, but if it is positioned in terms of helping people, it must be able to greatly improve work efficiency and quality.
Many people may hate the United States, but we have no way to take the high technology of the United States, except for a very small number of enterprises like Huawei can also bury their heads in research and development, many companies want to make quick money, keen on "building is better than buying, buying is better than renting".
The real core technology can't be bought, and when it is stuck in the neck, people either open their mouths or don't sell it, then we are completely dumbfounded.
From ChatGPT to GPT-4, from the initial elementary school level to the university level, the future development is completely limitless, and it may fundamentally subvert many industries.
GPT-4 is already relatively close to the product, and may be further commercialized in the future, and the performance will be further improved and become more stable. No one can predict what else will be done if GPT continues to iterate!