Counterattack starts with controlling probability

Zhang Feng smiled, his tone a bit amused: "Young man, you are indeed a speculator. You are quite direct in doing things."

Li Hao's heart tightened, but he maintained a smile and responded, "Teacher Zhang, you're right. I was indeed a bit presumptuous. However, it's a rare opportunity for me to meet an expert like you."

Zhang Feng wasn't being modest. He nodded and said, "You said before that you were working on quantitative models, selecting factors for speculation? Then let me test your basics! In modeling, what are overfitting and underfitting?"

Li Hao was stunned for a moment, then calmly replied, "Overfitting means the model performs well on the training data but has poor adaptability to new data. Underfitting means it performs poorly on the training data and fails to learn the patterns in the data."

Zhang Feng smiled after hearing this. "Not bad, you do know something. Okay, let's add each other on WeChat and chat again when we have the chance."

At this time, the car happened to be waiting for a long time at a traffic light.

Li Hao was delighted and quickly took out his mobile phone. The two of them successfully exchanged contact information.

This time, the wait brought him good luck.

When the car arrived at Qiandu Building, Zhang Feng smiled and said, "You are quite smart, young man. Feel free to come to me if you have any questions in the future."

Li Hao responded with a smile: "Then I won't be polite, Teacher Zhang!"

……

At this moment, Li Hao was very much looking forward to how Lao Pi would develop next.

[Not bad, Xiao Li, you finally completed the task neatly]

"Why? I'm also a graduate from a top domestic university many years ago. Are these performances surprising to you?"

Lao Pi paused, as if he was "thinking" about how to respond.

[Although you hold a master's degree from a top university, based on my assessment of your entire process, up until five minutes ago, your failure rate was still 38%. It's not an exaggeration to say it's unexpected.]

Li Hao couldn't help but laugh inwardly. "Okay, everything you say is right. You're an expert at assessing probabilities! So, after completing the mission, what improvements have you made to your abilities?"

[You hit the nail on the head! Currently, there are two improvements: you're now able to use the app multiple times per day, temporarily limited to 10 times per day, and...]

Lao Pi simulated the sound of taking a deep breath, which aroused Li Hao's curiosity.

[Currently, you can perform real-time probability assessments based on the event you are in. If the success rate is already high, as you hoped, there is no need to waste an opportunity to use it.]

"Sounds great, Jake! Let's see how powerful you are after you upgrade!"

Li Hao is already eager to try.

……

He immediately took out his phone and searched for some keywords online: "Qiandu Building", "AI", "Conference"...

What caught Li Hao's eye was the Qiandu Building in front of him, the "Third National Artificial Intelligence Model Symposium", and the instructions for attending the conference.

He noticed that the guests who spoke and other invited participants were basically well-known experts and scholars in the industry. Among them, Li Hao also discovered Zhang Feng's name and title - Senior Researcher at the Amazon Artificial Intelligence Center.

At the same time, Li Hao discovered that the vast majority of participants in this conference were invited, and there were very few places for voluntary applications, and the seats were full with no vacancies.

There are still 10 minutes left before the meeting starts.

"We're already at the door, it would be a shame not to go and listen to some cutting-edge technology! Lao Pi, help me check, if I want to get an electronic invitation before this seminar starts, what is the current probability of success?"

[3%. Unless there's a miracle, it's basically impossible. First, you're not part of the circle. Second, even if someone doesn't come at the last minute and a spot becomes available, you have no way to find out.]

"Then let's increase the success rate and make a miracle happen."

[No problem, use it today - 1, 9 remaining. I'll increase the success rate to 99% right now.]

“Won’t it be raised to 100%?”

Xiao Li, you have to understand that nothing in this world is certain to happen. The only certainty is the uncertainty of life.

"Do you think you are Luo Xiang?"

……

Just when Li Hao asked Lao Pi to use his ability, he suddenly found that a person who was walking quickly into the building suddenly received a phone call, and his expression became very unnatural, as if he had encountered some emergency.

Then he operated something on his mobile phone and turned and left.

Li Hao knew that this should be his miracle moment. He quickly applied to attend the meeting on the interface he had just searched, and did not stop until the remaining number of places returned from 1 to 0.

"Done!" Seeing the high-tech conference certificate, Li Hao breathed a sigh of relief and strode into the Qiandu Building.

In a spacious conference room on the second floor of the building, several invited AI experts and other participants were having a heated discussion about OpAI's latest o1 model.

Li Hao found an empty seat in the back row and sat down, looking forward.

The host smiled and tapped the microphone. "Hello everyone, and thank you for coming. OpAI has just released a new model, o1, which has sparked much discussion across multiple fields. Now, let's hear what the experts have to say."

Professor Wang spoke first, pushing up his glasses. "Our evaluation found that O1 performs better than GPT-4o in reasoning-intensive tasks like programming and science competitions. It's particularly adept at analyzing a problem first and then gradually advancing it, forming a complete 'chain of thought.'"

Zhang Feng took over the topic and nodded. "That's right, especially in programming, especially in low-code development and cybersecurity, where it can break down complex tasks into multiple subtasks. However, the application of AI in cybersecurity also brings some risks, such as the possibility of being used for cyberattacks."

The host asked: "Can o1 bring us into a new era of AI agents?"

Dr. Li continued, "AI Agents are indeed a challenging problem, but O1's reinforcement learning and thought chain technology give us hope for a breakthrough. Although this also requires more computing power, especially in the inference phase, future optimization is still expected."

Professor Liu added: “However, while MindChain technology excels in mathematical and symbolic reasoning, it offers little improvement in other tasks and may even affect the overall performance of the model. When building general artificial intelligence, we must find a balance between reasoning capabilities and task instructions.”

The host concluded: "It seems that although O1 has advantages, its shortcomings in natural language processing still exist. The future of AI Agent requires more efforts."

Professor Wang smiled and nodded, "Indeed. The next key point is how to truly put these technological advantages into practical application."

As the experts' speeches drew to a close, the moderator smiled and glanced at the attendees. Then he said, "Now, let's move on to the question-and-answer session. If you have any questions, you can ask the experts."

After a period of hushed discussion, several participants raised their hands.

The host smiled and called a middle-aged man sitting in the front row: "Sir, please go first."

The man cleared his throat, looking a little nervous. "I'd like to ask, is the O1 model smarter than GPT-4o? Can it handle all types of problems? For example, can O1 achieve the kind of general artificial intelligence we'll see in the future?"

Professor Wang smiled slightly, clearly accustomed to this kind of question, and spoke gently, "O1 does outperform GPT-4o in certain specific areas, but it is still a long way from true general artificial intelligence. General artificial intelligence (AGI) needs to demonstrate broad adaptability across multiple complex domains, while O1 is currently more focused on specific scenarios, especially reasoning-intensive tasks."

"Next." The host called another young man wearing glasses.

The man stood up quickly and said excitedly, "I want to ask, since O1 is so powerful, will it replace programming jobs? Will there be no programmer jobs in the future?"

Zhang Feng frowned slightly, then smiled. "AI can improve programming efficiency, especially in low-code development, where it has great potential. However, it's still too early to say that it will completely replace programmers. AI is more like a tool, helping humans complete repetitive, mechanical tasks. However, innovation and the development of complex systems still require human creativity."

In the conference room, the question time was already halfway through, but the questions asked by the first few questioners seemed a bit superficial, making the atmosphere a little dull.

Li Hao sat in the back row, observing quietly.

He knew that this opportunity was rare and he had to ask a profound question to really attract the attention of the experts.

When the host scanned the audience again, Li Hao raised his hand.

The host saw him, smiled slightly, and nodded, "Sir, please ask your question."