Original link: https://astrianzheng.cn/gossip-for-artificial-intelligence/
Due to the “turned out” of ChatGPT and the emergence of applications connected to the GPT interface, the discussion on artificial intelligence has exploded recently.
I can’t be exempt, and I have joined related discussions in the past few days. Here are some opinions published on social media recently for the readers.
talk about attitude
“Pew Survey Shows People Are More Likely to Be Worried Than Excited About AI” The Pew Research Center surveyed 11,004 US adults about their awareness of the use of AI in the products around them. The survey was conducted in December, before ChatGPT was the hottest tech topic. The survey found that the more educated a person is, or the higher their household income, the more likely they are to recognize AI. The more a person uses the Internet, the more likely they are to recognize the use of AI. About a quarter of those surveyed had heard of AI, and 15 percent had never heard of it. Regarding the hidden dangers that AI may bring, 38% of people are more worried than excited, and 15% are more excited than worried. The more people understand about AI, the more excited they are likely to be.
——”Singularity” website report, original address
For me, AI neither excites nor frightens me. One of the questions I really think about from AI is, under the premise of “industrialized production” in “self-media”, in which aspect should the uniqueness of people and the uniqueness of works be reflected; and, practical and unique Is sex really all two ends of the scale?
I think one of the questions that the progress of GPT raises for everyone is, what is creation to yourself? AIs have now shown us that instead of simply piecing words together in a coherent fashion (like ” typewriter monkeys ,” brute-forcing a script), they can use existing information Recombine integration to guess the most likely permutation. Even so, it is difficult for us to classify things created by AI as “creation”. But human “creation” is the same as AI, it will be strongly related to your experience, reading and knowledge, so how to show the thinking that AI can’t show in creation? This is a topic that every creator should think about when every AI singularity really comes.
Talk about the essence of artificial intelligence
In fact, about GPT, a very interesting thing I think of is the struggle between the Enigma machine and the Turing machine.
To put it simply, Germany produced a cipher machine called Enigma during World War II. Although this thing is essentially a (mechanical) Caesar cipher encryption algorithm, due to the addition of many layers of encryption mechanisms (including key rolling and multiple layers of encryption), which made it almost impossible for anyone to decipher the German intelligence encrypted by the Enigma machine in the early days.
At first, everyone was looking for linguists to try to crack it. In the end, Turing scolded you bad linguists that you would never be able to handle it. This matter has to be done by mathematicians, which made those linguists very angry.
The end result, of course, is that Turing was right—a mathematically sophisticated Turing machine managed to destroy the mathematical fog that Enigma had built up.
The reason for this association is that the AI algorithm is essentially an algorithm composed of a set of advanced mathematical equations. People seem to feel that a set of advanced mathematical methods can be used to penetrate the true meaning of language and art.
talk about translation
ChatGPT API-based text translation, text polishing, grammatical error correction Bob plugin, let us welcome a new era without the Tower of Babel!
——bob-plugin-openai-translator project introduction, original address
Don’t have the ghost dream of “completely tearing down the Tower of Babel”. Language is not just a tool for information transmission; in the process of use, language will always become one of the contexts of the text. This is why “literary translation” is largely one of the fields that is almost impossible to be replaced by AI, because literary translation is essentially a process of reproduction. When the translation of “Roaring Heights” was not proposed, AI itself had no way (with the same thinking path) to propose such a translation.
Talking about Human and Artificial Intelligence
AI can have insights beyond ordinary people, but only humans can surpass insights.
Talking about logic learning and feeling empathy
“It’s very difficult to discuss the limitations of large language models. Because two years ago, many people were very confident in commenting on the limitations of large language models. Now these language models are completely different… There is a saying that these models are just going to Learning the laws of statistics, so they don’t really understand how the world works. I see it differently, I think learning these laws of statistics is much more important than it seems. … Forecasting is a compressed computation .Prediction is also a statistical phenomenon. To predict, you need to understand the real underlying mechanism of generating those data. To be able to predict better, to be able to better compress calculations, you need to better understand the generation of these data world. As our generative models get better, they have what I call “a shocking degree of understanding” of the world and its many subtleties. But it’s not just about the world, it’s the world seen through the prism of words.
——Ilya Sutskever, original address
“Logic” itself is related to experience and knowledge, which in turn can be observed and summarized statistically (aka big data models). Therefore, this statement is valid in the development of logical thinking. An intelligent agent can indeed train the formation of logical thinking through a large number of summarization and refinement.
But “emotion” itself is not. The mechanism of emotion generation is extremely complex. A baby who has never been exposed to the world will also have negative emotions about pain and discomfort without the need for so-called “training”. In addition, although the design of the neuron AI model draws on the working principle of the brain, it is not completely consistent with the working mechanism of the brain . Therefore, it is not true to say that a large amount of training is used to complete the evolution of emotions.
talk about creation
Talking about Jipiti <br />He is well-informed and knowledgeable, and he can answer fluently
Large-scale language models can be described as the most fashionable models now. “Talking Jipiti” is the latest product of Kaizhi Company, which can chat with you. Regardless of business, education and politics, they all know everything.
Use it frequently, there will be no more difficulties in work and life, and the work will be extremely efficient.
Produced by Kaizhi Company in the United States <br />Free trial, the interface is available for sale
——A meme created by Celestial Phineas, original address
ChatGPT cannot “create” these things by itself – human beings can’t speak well or create well, and there is no way to teach computers to speak well and draw well.
This article is transferred from: https://astrianzheng.cn/gossip-for-artificial-intelligence/
This site is only for collection, and the copyright belongs to the original author.