Views on the AI Field and Personal Background
A. Complexity of AI Impact
The innovations in the AI field, such as GPT-4, have profound societal implications, ranging from utopian visions of eliminating labor to dystopian scenarios that harm the livelihoods of artists, and even threats to human existence.
B. Personal Professional Background
The author is a data scientist who has performed well in machine learning competitions at top Australian universities and has built libraries in MATLAB for their master's thesis. Although not top-tier, they are better than most peers.
AI Industry Chaos and Personal Experience
A. Industry Fraud Phenomenon
In 2019, while working in data science, the author found the field to be vast but plagued by fraud. Many leaders with insufficient AI knowledge insist on AI-oriented companies, with AI project numbers far exceeding actual use cases. Many people have used hype to climb the career ladder, but are mostly fraudsters or incompetent individuals.
B. Personal Career Transition
Due to industry chaos, the author transitioned to data and software engineering fields. Although AI-related work pays well, practitioners value job stability, friendship, and personal integrity more. Fraudsters are adept at using "political tactics" and can easily switch careers when the hype ends.
C. Industry Subsequent Development
Data science jobs have decreased, and the hype has shifted to data engineering. The author thought the AI craze was over, but then ChatGPT appeared, with a lot of engineering resources being used to add chatbot support, while many companies can't even perform basic database backup testing.
riticism of AI Application in Enterprises
A. Most Enterprises Do Not Need AI
Most enterprises do not need to actively apply AI because it may already exist in the software supply chain, such as security providers using relevant algorithms to detect abnormal traffic. Enterprises should first solve their own operational and cultural issues, such as the fact that most companies cannot develop and deploy basic applications on time and on budget, yet they want to apply complex AI technology, which is a prelude to disaster.
B. Questionable Effectiveness of AI Applications
The author questions some reports on the effectiveness of generative AI applications. For example, a Scale report shows that many companies have benefited from generative AI, but the author believes this is not the case. Many companies cannot even manage CRUD applications well and cannot be so successful in AI applications. Moreover, GPT-4 performs poorly in some aspects and should not be entrusted with decision-making.
C. AI Product Demonstration Fraud
The author's friend's experience shows that some FAANG organizations' AI product demonstrations are fraudulent, and the products do not work properly. There are also companies selling emergency service software, whose actual service is just a person in India, misleading consumers with the concept of AI.
Discussion on the Future Development of AI
A. Three Possible Directions
Intelligence Explosion Direction: AI may self-improve and trigger an intelligence explosion. Although the author thinks it is possible, they also point out that defending the Earth is another issue, and it is uncertain whether it is feasible. In this case, the author does not care about corporate interests.
Development Limited Direction: Due to insufficient data, architectural issues, etc., AI may not develop as expected, affecting some industries such as customer support. However, if this is the case, there is no need for enterprises to apply AI just for the sake of applying AI. They should first solve their own problems.
Technological Breakthrough Direction: If fundamental issues are resolved, AI may replace programming or achieve general intelligence. However, if AI develops rapidly, enterprises that only develop chatbots will not be able to cope with the future. They should be at the forefront for research or moderate application, otherwise, they will be disrupted.
B. Correct Practices for Enterprises
Enterprises should either conduct cutting-edge research or moderately apply LLMs on the basis of basic business. The middle approach is meaningless unless the industry is being completely disrupted. At the same time, enterprises should first solve their own operational and cultural issues to improve their capabilities and better apply AI.
Negative Phenomena of AI in Society
A. Impact on Policy and Healthcare
The author suspects that some government policies are written by ChatGPT and is also worried that doctors may misdiagnose patients due to the use of chatbots.
B. Problems in the Education Field
Educational institution executives asked the author if they could detect student cheating. The author suspects that many students are cheating but has not investigated in-depth. In addition, some companies use free memberships to collect data for training AI models, infringing on user rights.
Criticism of AI Trend Followers
A. Characteristics of Trend Followers
The author criticizes those who blindly follow AI, who were previously keen on concepts such as blockchain and quantum computing. These people do not understand technology but want to profit from it, which is a disgrace to those who truly understand technology and want to reasonably apply technology to improve the world.
B. Personal Experience
The consulting company where the author works has excellent data scientists but does not provide AI services because the market is distorted. The author believes that these trend followers should be put into a "thought leader prison" for education.
Resistance to Over-promotion of AI
The author warns those who do not understand machine learning but talk about AI that they will be punished if they speak again. At the same time, the author introduces their contact information and explains that the article updates may slow down recently due to personal affairs.