Summary of Key Points
This article, based on the personal experience of the author "Nai Ba Lao Xu," reveals the stark contrast between the extreme capabilities of current top AI models (such as GPT-6 Astra) and their high usage costs. The author intended to use AI for 3D modeling but encountered significant limitations due to the high amount of "quota" consumed by AI in tasks such as writing assistance, material processing, and translation refinement, which led to the failure of the project. The main message of the article is that while AI can greatly improve efficiency (for example, completing what would otherwise take hours of research in just one hour), its substantial "hidden costs" (subscription fees plus quota limitations) make it difficult for ordinary users (Plus subscribers) to afford, and even for more advanced users (Pro/subscribed at the highest level), it can be quite expensive when performing complex tasks. The author urges everyone to consider the real economic implications of using AI, rather than being merely dazzled by its impressive capabilities.
---
Detailed Analysis
1. AI as a "Super Butler" but also a "Money-Eating Beast": The Disparity between Efficiency and Cost
Many people still think of AI as merely chatbots or code assistants, but in this article, AI has evolved into a "digital employee" capable of operating computers, taking screenshots, converting videos, and even controlling 3D software.
Simple Explanation:
Think of AI as a versatile but expensive butler. If you used to ask this butler to buy milk from the supermarket, he might just stand at the door and ask you what brand you wanted. Now, this butler (Astra) not only knows the brand but can also drive to the supermarket, compare prices, make the purchase, put the milk in the fridge, and even sweep the floor while he's there. However, the problem lies in the "salary." The more capable this butler is, the more "energy" (i.e., computational resources) it consumes. The author found that just asking AI to help with research and material organization used up a quota that would normally last for five hours. It's like hiring a Michelin-starred chef to make instant noodles, only for him to use up all the gas in the kitchen in pursuit of perfect taste, leaving you without money for the next month.
Core Conflict:
AI's intelligence is built on massive computational power. The more autonomous it becomes and the more complex problems it can solve, the more quickly your budget (subscription quota) runs out.
2. The Dilemma of Plus Subscribers: $20 per Month buys "Anxiety," Not "Freedom"
The article criticizes the $20/month (about 135 yuan) Plus subscription plan, which is the most affordable option for most users. However, the author points out that this plan has recently reintroduced a "5-hour quota" system, significantly reducing the user experience.
Simple Explanation:
It's like paying $200 for a monthly gym membership. Before, you could use the gym whenever you wanted. Now, the gym says, "You can use it, but you must rest every five hours, or you can only work intensively during those five hours and then wait until the next period." Worse still, the quota is not based on time but on the amount of work done. Simple tasks like checking the weather or correcting a typo might only use 10% of the quota, while complex tasks like searching online, taking screenshots, or organizing information can use up 50% to 100% of the quota. The author describes this as feeling like you're constantly counting your pennies. You're hesitant to ask AI to do too much for fear of running out of quota, but not using it at all defeats the purpose of using AI to increase efficiency. For most non-professionals, $20 doesn't buy a productivity tool; it buys a sense of anxiety about running out of resources.
3. Model Hierarchies: Astra is a "Doctor," Terra is an "Intern," Sol is a "Skilled Worker"
The article discusses three levels of AI models: Astra (the strongest), Sol (intermediate), and Terra (weaker). The author uses the analogy of sweeping the floor to explain their differences:
Simple Explanation:
- Terra (Intern): If you ask it to sweep the floor and it doesn't find a broom, it will report, "No broom available." It lacks the ability to solve problems independently and needs your guidance at every step.
- Sol (Skilled Worker): If you ask it to sweep the floor and it finds no broom, it will look for one on its own, pick it up, and start sweeping. It can handle some vague instructions and has a certain level of autonomy.
- Astra (Doctor/Expert): It not only finds the broom but also thinks about why the floor is dirty, the best angle to sweep from, and whether to mop it after. It has strong reasoning and exploration abilities. The author notes that the smarter the model, the more expensive it is to use because it consumes more resources.
User Pain Point:
Ordinary users often need the capabilities of a Sol-level "skilled worker," but to occasionally use an Astra-level "expert," they have to pay for a more expensive subscription. As a result, they spend most of their time using the Sol-level model and end up paying for the Astra's higher intelligence, even though they're mainly doing tasks that would be handled by a Sol-level model.
4. The True Cost Behind "Showoff": 3D Modeling is Just the Tip of the Iceberg
The article mentions that many people use AI for 3D modeling on social media, which seems impressive. However, by comparing it to the case of a well-known tech blogger named "Kazke," the author reveals the actual costs involved.
Simple Explanation:
Videos showing AI generating 3D models of the Temple of Heaven look impressive, but they only show the final, beautifully arranged result, without revealing the amount of high-quality materials used, the hours the chef spent, or the rent of the restaurant. Kazke, with a $200/month (about 1340 yuan) premium account, spent nearly half a week's quota (about $168) just on creating a bare, unlit, and background-less model of the Temple of Heaven. This suggests that a complete 3D project would cost hundreds or even thousands of dollars. For ordinary users, this is like spending a month's worth of food on one meal.
Deeper Logic:
The "showoff" aspect of AI often masks its high marginal costs. For commercial or professional applications, these costs must be carefully considered. For ordinary users, such displays are more about psychological satisfaction than actual productivity. They give the illusion that AI can do anything, but when it comes to solving complex problems, there's a gap between the model's capabilities and their budget.
5. Future Trends: More Precise and Hierarchical AI Usage
The article suggests that as AI becomes more powerful, its usage costs will also increase, and users will need to manage their AI usage more carefully.
Simple Explanation:
In the future, using AI might not be as unlimited as it is now. It will require more strategic planning, similar to refueling a car. Users will need to give more precise instructions to reduce AI's computational resources. For example, they might tell the AI, "Sweep the living room; don't clean the whole house," or choose the right model for the task's complexity. Subscription plans may become more tiered, with different levels offering varying amounts of resources and intelligence. Users may need to combine multiple subscription plans or accept the reality of paying for the level of intelligence they need.
Advice for Ordinary Users:
Don't blindly pursue the "strongest AI"; instead, find the "most suitable" AI for your needs. While enjoying the efficiency improvements, be aware of the associated costs. If using AI causes anxiety or burden, it might not be suitable for your current tasks. Learning to use AI wisely or adjusting your budget will be essential in the future.