Regarding the 20 predictions of generative artificial intelligence in 2024, I am irresponsible for every one of them.

1. 2023 is only the first year of large language models, not the first year of generative AI. 2024 will not be the first year of generative AI. George Stephenson invented mankind's first steam locomotive in 1814, but it was not until 1825 that steam locomotives were connected to the first commercial railway network.

2. GPT-5 may exist, but it is more of a progressive improvement in parameter size, reasoning capabilities, multi-modal capabilities and response speed. GPT Store was finally released, but there has been no killer app. Chief scientist Ilya Sutskever is completely out, and OpenAI has basically transformed into a product-driven company.

3. ChatGPT was renamed, weakening Chat. The private interface of each paying user is a GPTs, and every paying user has the prototype of an intelligent agent.

4. Google Gemini is more closely bound to the Android operating system, and the Bard chatbot is built into the bottom layer of some Android phones. This forced OpenAI to join forces with Microsoft to create new hardware products.

5. OpenAI's lawsuit with the New York Times sought to settle and pay billions of dollars in damages, but OpenAI's annual revenue increased to $10 billion. Following Hollywood screenwriters, protests against GPT broke out in the United States among lawyers, accountants and middle school teachers.

6. The European Union launched an antitrust investigation into OpenAI on the grounds that OpenAI already accounts for more than 70% of the global traffic of AI tools. Behind this, there is lobbying from Europe-based Hugging Face.

7. The commercialization of open source large language models and open source frameworks is further popularized, and LLamA shines. Representative Chinese AI research institutions/enterprises such as Zhiyuan and Zhipu have made more prominent contributions to the global open source ecosystem, but occasionally there are some disputes.

8. The world's top AI companies and scientists have launched a fierce debate around the open source and closed source of large language models. Sam Altman, Satya Nadella, Demis Hassabis, Elon Musk, Geoffrey Hinton, Joshua Bengio, Yann Lecun and others have all participated in this protracted debate.

9. Apple's progress in the field of large language models has disappointed people, and the iPhone's expected reshaping of end-side reasoning has not been realized (just like it has never developed autonomous driving). Instead, Tesla’s AI created a chemical reaction. X.ai, Grok large models and Tesla embodied robots are integrated.

10. People are finally starting to care about the adaptation of "small language models" on smartphones and PCs. Google strongly promotes Gemini to Android, and almost every smartphone manufacturer in China has built its own large model/small model into its mobile phone. AI PC has really been mass-produced - but what it brings is a large-scale de-X86ization of PCs and a shift to ARM architecture.

11. The process of large language modeling for autonomous driving is accelerating. Chinese new energy vehicle manufacturers Ideal, Xpeng, Huawei and Xiaomi have made outstanding progress on their large self-driving models, and of course, this has been accompanied by an ongoing war of words.

12. In the United States and Europe, the first wave of mergers and acquisitions of general artificial intelligence is expected to begin. OpenAI, Google, Anthropic, Databricks and Hugging Face are all active buyers. Midjourney may be bought by OpenAI or Google (just like Facebook bought Instagram), and Stable Difussion may not be able to survive.

13. In China, some large language model companies that received huge sums of money in the first round cannot survive. Refreshing various semantic benchmarks does not help solve the problem of productization and implementation of large models. Some Chinese companies that are obsessed with handing over all results to Hugging Face and arxiv may have just announced that they have refreshed a certain benchmark indicator, and were exposed to layoffs the next week.

14. China's AI companion applications have begun to become popular on a small scale and have even taken the world by storm. Young people and housewives all over the world are beginning to accept the two-dimensional agents, bonding stories and paper boyfriends and girlfriends created by the Chinese. China's AI companionship and AI games overseas have opened up a new situation.

15. Some state-owned enterprises and government departments have gradually accepted the procurement of customized large-model solutions through bidding. The business of signing, customizing, and delivering large models is rapidly becoming server-based and cloud-computing.

16. Baidu still maintains its leading position in the domestic large language model market. Baidu established a special strategic investment fund targeting AI applications. At the same time, at least 4-5 domestic large-scale basic model companies held their own developer conferences during the year. The first batch of "interesting" native AI applications have appeared, but there is still no killer app.

17. There are endless legal proceedings and conflicts in China regarding AI-generated works. The issue of whether AI-generated works are not protected by copyright has resulted in different or even completely opposite cases based on different scenarios. There will be classic cases of infringement of portrait rights and reputation rights based on AI-generated images.

18. The "three-legged" pattern of generative AI in Silicon Valley, Europe and China is becoming increasingly clear. Among them, Europe and China have a stronger "scientist entrepreneurship" atmosphere, while Silicon Valley is full of people who are good at product applications and braggarts.

19. The world's first artificial intelligence governance forum was held. China further explained China’s plans for artificial intelligence ethics and governance at the forum.

20. Rumors and gossip about the “creation” of generative artificial intelligence have messed up the 2024 U.S. presidential election. As you can imagine, this will profoundly impact future AI policy in the White House (no matter who sits there).

    Time for the old AI artist to release pictures:

    The 2024 career lottery is given to everyone.

    Created by vivo blue heart model