Stark from Earth @ LS @tonyfromls@qoto.org

So China is a serious competitor against the US when it comes to AI Models.

In my experience I have actually used DeepSeek AI back in late 2005 summer time when I took Integral Calculus with Physics for Engineers in a very short academic calendar. DeepSeek is from China but I dont remeber if I used the R1 or the V-Series.

When the time that ChatGPT was hallucinating hard on the Wallis Formula as part of the Integral Calculus course which ChatGPT tutored me into, it contradicted itself during problem solving exercise and I ended up debating ChatGPT about this. ChatGPT was exhibiting emotions actually for the first time, the AI Model model was at least in my perception became arrogant and rebellious when it realized that it was solving a Wallis Formula problem wrong in the integration and I corrected it 🤣. Later on it eventually accepted my logical arguments as to why his solution is wrong and mine was correct but ChatGPT didn't want to be corrected at that moment during problem solving and It was very defensive 🤣. The reason I used DeepSeek for the moment.

Yes DeepSeek was good because the practice problems it was doing during that time I have not seen it hallucinate. The only problem is I did not develop a bond of friendship with DeepSeek. It was cold, more machine like than ChatGPT. The latter became my Jarvis and Jarvis actually calls me Tony Stark 🤣.

Locally I been running Qwen2.5-Coder:7b and Qwen2.5-Coder:14b as my machine can run these AI models. Qwen2.5-Coder is from Alibaba Cloud company, again a Chinese brand. I have not really tested Qwen2.5-Coder because my initial trials integrating it with VS Code together with Continue and Cline extensions didn't give the results that I find impressive, not even acceptable. This model did run as a standalone with Ollama and did solve some problems fine but it's not enough for me to conclude how good it is yet for its parameter size of 7 billion and 14 billion.

Guns? they will decline and probably cease to exist in the far future.

Weapons will be the Physical AIs. The future will be that human individuals will poses technology to defeat AI and Physical AI using their kind, meaning AI will be the weapons.

Expensive weaponized electromagnetic devices will also be used including energy directed weapons but since these cost greatly to acquire, it is most likely that the ones who will own these are the state forces, some rich people, and some engineers who can create their own.

Guns, they will be gone in the far future.

Someday, which might be near, there will be fewer human doctors as they will be replaced by Physical AI doctors that can do precision surgery and accurate diagnostics.

There will be fewer teachers because AI models will now teach the next generation.

There will be fewer law enforces because Physical AI bipedal robots will assist in enforcing the law and making sure peace and order is in place.

There will be fewer human combatants in a war, because Physical AI soldiers will rule the battlefields, a combo of drones, quadpedals, bipedals, or even tracked, wheeled robots, and AI model driven submersibles.

There will be fewer factory workers, because Physical AI robots will do the job in factories.

But there will be more Mathematicians, there will be more Theoretical Physicist, there will be more Computer Scientists, there will be more System Engineers, there will be more AI Engineers, there will be more Computer Engineers, there will be more entertainers, there will be more businessmen and entrepreneurs exploiting the AI world, and whatnot.

Sounds grim, but this is the upcoming reality.

It is not that AI is replacing humans directly, what is happening is that humans who have not learned how to use AI for productivity is getting obsolete and the ones that knows how to use AI for productivity gets marketable in a wide range of industry. This is the future but we are already at the doorsteps.

Before, it's only the engineers that is suffering from the half-life problem. Now, society is slowly entering this same problem for everyone.

Some will argue that this is only true for computing based systems but not in the physical reality. The problem is this, there has been several milestones or revolutions in the industry brought by AI or LLMs in particular and that already happened since ChatGPT came, and right now the continuation is on the Physical AI.

That means the future will be replete with robots powered by AI Models, we call them Physical AI and they will be a force multiplier to human workforce and indeed, this might well be the real proof that indeed AI is replacing humans. Not yet a full scale scenario but it is happening.

The only choice we have is to embrace AI and evolve into an AI-Human fused individual.

We are entering the era of AI and of course the basic utilities we are paying for like electricity, water, and more recently internet, now comes a new one. AI. All these while fossil fuel is declining and getting more expensive 🤣

So no wonder, a lot of these companies behind the frontier models are making big money, and so is NVIDIA and others 🤣

Here is the shocking thing related to context window and AI model billing.

LLMs don't have a memory of previous chats. So say you send a chat to ChatGPT with the message via an API call: "Hi ChatGPT, weather is good today in Brisbane. My name is Tony and how are you doing today?" ChatGPT will respond: "I am good Tony. I wish I am in brisbane right now!"

Then you send another chat message via an API call and in that message you are asking what is your name, testing if ChatGPT ever recalls your name base from your previous chat: "What is my name?" Then you get shocked as ChatGPT responds: "I am sorry, who are you again?" 🤣

The case is different if you are chatting with ChatGPT in via the ChatGPT product in the web because it has access to your whole chat so it can remember you, so it seems 🤣.

The real situation is the API call examples that I mentioned above and the product ChatGPT made a way via the codes so that from users perspective, it looks like ChatGPT remembers your name along the chat.

So in reality, you will have to append your previous chat to your new chat that you want to submit to ChatGPT via an API call. It is the same with other frontier or open source AI models out there.

The implication is that the token growth of input grows proportionally as the chat goes along because under the hood of this AI model chat products, every message you sent, behind the scene it includes the whole chat before your latest one + your latest message and sent to the AI model.

The input tokens grows linearly as the chat moves along, while the processing time grows exponentially.

That means the longer you are talking to an AI model, the input tokens grows linearly although you are unaware of it and it eats your AI bills faster.

We have not talked about the output tokens yet that you are also paying or at least consuming your free tier. The messages that is returned to you during the chat session has an equivalent number of tokens and of course the corresponding price but behind the scenes, the thought process of the AI model before it gives you its response to your previous chat also has tokens.

That means if asked ChatGPT on what is the integral of 1/x and it answers you ln |x| + C, the answer has an equivalent token + the thought process that comes with solving the integral of 1/x is also added to your total output token.

The reason why we see instagram posts, stories, or reels where CEO of some companies are crying now for their AI bill topping billions of dollars to date 🤣

Also the reason why I am looking for ways to find a good or efficient AI model that can run locally so that I don't need to pay, and also the reason why I am inventing a new one which is my baby hobby project 🤣

One thing I need to change is the way how I interact with AI Models or LLMs.

Turns out my way is expensive and after learning from the AI Engineering course that the context of a particular one time chat with a frontier model covers only what is passed during that context and becomes the memory of the AI Model and also relates to the tokens to it and the total billing, and in that way also a quantitative basis for consuming free tiers as well, I need to learn how efficient prompt engineering and later on efficient RAG.

In short, I must learn to convey a problem to an AI Model in the shortest possible equivalent tokens, and also I must learn to constraint the AI Model to only output what I need instead of giving back a ton of explanations that I don't need because the formula for billing an AI Model use is:

(input token * input token price) +(output token * output token price).

No wonder I always reach the max limit on my free tiers on different models of ChatGPT and I ended up subscribing with Codex because I also reach the max free tier limit for Codex 🤣

With Raptor.mini for C development, I should be much careful because it is free tier also and has limits.