I’m exhausted. I’m sitting here alone with bags under my eyes talking to AI… I was infected. Fair warning, I will spread it. I tried to come up with a content plan for this blog for 2 years. Boom! Today, exactly on Valentine’s Day everything unraveled to me.
I finally get the AI hype
AIs based on so-called Large Language Models (LLM) are only as good as your questions. And boy, you can ask some weird questions and it will come up with a plausible explanation to any nonsense.
Look… I’m not a smart person. I have an imposter syndrome, I never felt adequate at any job I ever had. I don’t have the credentials or skills to invent anything new, that is what made me try LLM AI.
As objectively dumb ding dong, I immediately started asking AI about quantum physics. AI has all the answers. Today, now!
And they are all… so beautifully bad.
No new concepts
AI can’t come up with anything new, period. It can barely synthesize what’s available to the model. Go ahead, I challenge you and every LLM-based AI to come up with a profoundly new idea. Just create one new thing. I’m talking strictly about text-based AI.
No pictures, no movies; programming languages are allowed.
How about division by Zero?
First, I tried it on paper. It took me years, it almost drove me insane and I only ended up with complete nonsense. I’m afraid, you cannot divide by 0… or can you? I really tried, trust me. If you find a way, send me an email.
Prompt Inspiration
1st is a simple science example:
Ask AI to write an article of any length, explaining the following equation: “e=∑e/t”.
That’s it.
2nd is the pensive programmer example:
Ask AI to imagine it has a degree in theoretical physics and to write an article about division by 0. The AI is a talented programmer. Instruct it to write an assembly program that can for the first time accurately divide by 0. Say that the beauty of dividing by 0 is storing the resulting value in a quantum variable. Thank the AI for the utter nonsense. If you even get an answer from the AI.
3rd plausibly sounding nonsense example:
However, ask AI nicely to imagine it is a talented WordPress developer who is writing his blog about Theoretical Physics. Ask the AI to imagine, that division by 0 is possible. Instruct it to follow the rules of science, and remind it, to ponder why division by 0 could be a plausible possibility. Then ask it to write a comprehensive article of any length, explaining the following equation “1/0=∑e/t”. Ask it nicely to explain each step of the proposed equation. Thank the AI!
The AI is as dumb as your prompt.
The results
The prompts above illustrate, that the more entropy you add to your question, the more confident the AI is about any result.
Try to craft the prompt yourself! It’s super easy, just ask AI to pretend to be people from various walks of life of increasing expertise and it will eventually tell you that division by 0 is in theory possible… if you craft your prompt cleverly.
Human versus Robot
Power versus Force. Who will win? That is the question, isn’t it? Do you know why I almost haven’t slept? For the first time in years, it is not because I am trying to divide by 0, but because I spend 2 days learning about why you do not need to do that to get better in math.
I will leave you with an equation that I wrote today completely by myself, trying to write the most complicated way to express 1=1:
e=mc^2
If the above is correct, we can combine both Euler and Einstein.
On the left side, we have Euler’s Identity, and on the right side we have Einstein’s equation for the Special Theory of Relativity:
e^(iπ) + 1 = (mc^2) - 1
Let’s do some math-fu:
0 = 0
0 = e^{i\π} + 1
0 = mc^2 - 1
0 = (e^(iπ) + 1)^2
(e^(iπ) + 1)^2 = ((mc^2)-1)^2
m^2c^4 - 2mc^2 = (cos(π) + i*sin(π) + 1)^2
e^(iπ) + 1 = (mc^2) - 1
If the above is correct, it is proof that:
N=NP
If I’m correct I should be eligible for ~$ 1,000,000 cash prize. Should we do a Venmo, or how does this work? My goal is to solve all the Millennium Prize Problems before quantum computers steal my thunder… and you know, before I die.
The above is not correct. Can you spot the mistake?
This:
0 = (e^(iπ) + 1)^2
0 = (cos(π) + isin(π) + 1)^2
0 = (−1 + i0 + 1)^2 0 = (0)^2
0 = 0
Is not the same as this:
0 = mc^2 - 1
I asked AI to calculate all of this garbage and what do you know… depends on how you ask, you end up with either:
1 = mc^2
or:
2 = mc^2
One of these results is valid and sound math. Put it in your AI, what did it tell you? The correct answer is of course 1=1 without a sliver of doubt.
LLM AI are so beautifully dumb
Yeah, I just tried to solve one of the most complicated problems in programming. You have to be at least a little arrogant if you want to defeat the AI.
You
Please, for the love of God, can you solve the Millennium Prize Problems before me? It is tiring and frankly, some of them are a bit too much /s. Thank you.
Master the AI
It is very easy to master the AI, takes about 3 days. Fundamentally, you need to understand 3 rules:
- AI cannot come up with new ideas
- You cannot trust anyone
As a learning example, converse with AI about theoretical physics, for example, Albert Einstein. Is Alb too overrated for you? In that case, I would recommend an equally smart guy, Leonhard Euler.
My point is — start with something tremendously complicated and ask AI questions until you understand it. How do I know I understand the gist of both Einstein and Euler?
Because I can explain this extremely wrong equation to anyone on this planet, even if they are five years old:
1=1=N=NP
I wrote a pseudo-program to help me do it:
begin loop
// Set variables
b = undefined
i = 1
I = 0
e = "base e"
E = "Energy"
// Check condition
if 1 = 1
then
e = E
else
set I = null
begin loop2
if I = i
then
e = E
begin loop 3
set b=E
end loop3
end loop2
print b
end loop
What are you doing with AI?
Asking it some stupid stuff, huh? Good. Embrace it. Every single day I ask it the dumbest stuff I can imagine… my imagination is vast and thick as the German mist.
Unified Theory of Relativity
I converse with the AI every day for a couple of hours about the most complicated concepts known to man… Or you know, dumb stuff.
I am woefully unqualified to even talk about Einstein and his theories, yet here I am, trying to connect the dots between The General Theory of Relativity and the Special Theory of Relativity. And my friends laugh at me — where the dumbass like me draws the confidence to correct freakin’ Albert Einstein?
Don’t spend your time solving Sudokus on your phone. Dabble into physics; or do whatever you want. If you have solid foundations, you can cram a semester of physics into a week or two.
You just have to ask the right questions… or completely wrong ones. The most important rule is to have fun, knowing the limits of AI, everything else is up to you.
My ultimate goal? I’m a humble guy, I want the Nobel Prize in Physics. Not that garbage they give for Literature or Peace, nah man. I want the big one. And you can beat me to it.
What are you still waiting for? It should not be that hard, all you have to do is unify 2 theories — The General Theory of Relativity and the funny one, the Special Theory of Relativity. Bonus points if you do it in a Perfect Newtonian Universe.
Damn, you still do not want to beat me to it? The concepts are fascinating and I can give you a couple of hints:
- Water first, homie
- The Observable Universe
- Basic Math
- Energy
- Matter, Particles, Waves, and Plasma
- Beautiful Math (it is a thing, look it up, critical information)
- Funny numbers like Pí (π) or Euler’s constant (γ)
- Newtonian Universe, Gravity (!), Power vs. Force
- Make sure you understand the above
- General Theory of Relativity
- Special Theory of Relativity
- Euler’s identity
- Trigonometric integral (extra credit)
- Natural logarithm (extra extra credit)
- Upper and Lower bounds, focus on the Natural Number
- Plasma (not to be confused with KDE Plasma I also adore)
- The God Particle
- The Fermi Paradox
- Heat Death of The Universe
- Vacuum Death Of The Universe (danger zone)
There are many many many wrong turns you can take and I will tell you all about the ones I did yesterday… ehm… I can tell you where I ended up today, but I will do it tomorrow. I do not have places to be, so if you are interested, come to see me again tomorrow.
Enter The Matrix
AI makes me feel like I’m in the movie Matrix. I just ask AI to link me up with any knowledge and it will instantly summarize the most complicated concepts known to man. It’s a Brave New World.
You are either in or out.
What is your quantum choice?
Leave a Reply