The Real Dangers of AI No One is Talking About

The Real Dangers of AI No One is Talking About

Photo by cottonbro studio: https://www.pexels.com/photo/a-woman-looking-afar-5473955/

Introduction:

AI is incredibly good at scanning through large amounts of data. In fact, that is how it is trained. My argument in this paper is that AI is fully advantaged to violate all privacy of people to teach itself more and enrich its creators. Things once thought impossible will become reality, as the powerful of this world will now be able to track everything you say and everything you do. They will transcribe your phone calls, collect your texts and emails. They will eavesdrop on personal conversations. And people will celebrate this as we get stronger AI models. They will be able to surveil like never before thought possible. I do not think this is considered science fiction anymore and will become a credible risk that we must all face. We are already facing privacy violations from these models, from them being trained on books without an author's permission to online art being stolen and replicated. Is it that hard to imagine they will continue this trend and take more of your data? I’ve also provided in this paper my thoughts on how the average person can push back on this inevitable privacy invasion. Hopefully this doesn't read as a conspiracy theorist spiralling into incoherence.

Discussion:

 “If Anyone Builds it, Everybody Dies” written by Eliezer Yudkowsky explores the idea of a super intelligent AI taking over the world in a sci-fi style of story. But is this actually the most likely outcome of how AI would “take over” humanity? I argue in this paper that this book misses a critical point about how AI is used. Starting off, AI is a tool controlled by those who have created it and those who have the money to fund it. Everyone that fits that definition loves one thing, control. From tech billionaires to politicians and the NSA the people who are behind AI development benefit from your information. It helps them sell you things, gain your votes and manipulate your beliefs. These people have never had the resources that AI now grants them, which will make them all the more efficient at gaining your data. AI is already able to take entire chat logs and summarize them, look for key themes, and provide actions based on the conversations. And we make this easy for them, by putting our every thought into google, chat gpt or in messages to friends. They see the media we interact with, all tallied up as an additional data point. This gives people with access to this data full power to research the masses and carefully influence hearts and minds.

 What are some trends that can be made worse by this revelation? One idea is the growing political divide that faces the west. If any party leader wanted to ensure they got your vote (and had you as active as possible to help them gain power), they could use your online data to effect algorithms. They find the subjects that pull your emotional strings, and put that back in your face. This reinforces the echo chambers already present in these spaces. They can then aggregate data from the masses, using it to shape campaigns and messaging in its entirety. For example, if 55% of people oppose an idea, then you can target those groups with specific messaging to gain control over them. Already we have seen growing concern over how social media damages our mental health. What happens once their insights into you get sharper? And they find ways to extract more of your time and money into their platform? We will become less happy, and will look to disconnect. Our broader views of society will become tarnished. Your outlook of the world will dwindle, and you will be ready to manipulate.

 What can be done? You can spend less time on social media, engage less with content, follow less people etc. The less you put online the less they can take from you. And you spend less time being manipulated and more time in the real world with what matters. This also means being present with your feelings. “Why did this post make me feel angry, sad or fearful?” is a useful question in today’s world. To understand if there are specific accounts, apps or topics that make you feel strongly. Deconstruct these feelings and step away from the content. Seek opposing view points. Treat politics as a conversation among others where you aim to see both sides. This will lead you to feeling more balanced when the narratives in front of you make you upset.

My new idea of how we can further obfuscate our online persona to avoid AI detection is encryption. Prior to making this post, I used python to make a simple script that does a very basic two stage encryption on text. I used AI to help make it (funny enough) and now have a way to encrypt text locally so that someone with access to my chat logs could not use it to train an AI model on my conversation. You just need to provide the key via one line of communication, and share the message via another line. Currently this seems completely unnecessary but I argue we won’t know when we need to implement this until it is too late. People will find this idea appealing as more of their data becomes available to the wider world. Hopefully, some indie social media developers could make this more widespread and available. Where followers receive a key from people they follow, removing the chance of bots scrapping your data. This would also protect your creative works from mass AI replication. For example, currently this blog has about 600 total views (very small, just a hobby). Based on some key metrics, it appears around 20% of that traffic is AI bots scraping my work.

Conclusion:

If you want to read my full conclusion, you can try to use my encryption/decryption script. All you will need is pycharm and the below code. The authorization code is 9631804 (this is the example value that pops up as well, if you use this tool yourself you can make it any numeric value). Thanks to everyone who gave this a read! Also feel free to share/edit the code with friends, it's a fun way to experiment with coding. I am sure there are people out there much more talented that could make this more accessible to others as well.

13 F7K XL13ROW X5 13PP 5J XL5W10 AL5 V73H X5 XL10 10RH 13RH X15O XL10 X7Q10 X5 H10GVCTX XL7W J7R13P Q10WW13K10. XL7W 10RGVCTX35R L13W F70R X10WX10H 13K97RWX XL10 P13X10WX 13RH KV73X10WX Q5H10PW 13RH L13W R5X F70R FV5O10R. XL10V10 13V10 Q12GL F10XX10V A13CW X5 10RGVCTX C22V H13X13 7J C22 13V10 V73PPC G5RG10VR10H 13F22X W5Q65R10 P7WX10R7RK 7R X5 C22V G5QQ12R7G13X35RW. XL7W A13W Q5V10 N12WX X5 J70P 22X XL10 XL22KLX 10BT10V7Q10RX 13RH TV13GX7G13PPC W70 XL10 GL13PP10RK10W A7XL 10RGVCTX7RK 5RP7R10 H13X13. Q5WX T65TP10 A7PP R5X V12R XL10 G5H10 C22 L13Z10 W5 GL70VW X5 C22 J5V J7K12V7RK 7X 22X 13RH X13O7RK XL10 X7Q10. N12WX X5 W7KR 5JJ, QC G5RG10VR 7W XL13X QC T10VW5R13P X10BX Q10WW13K10W 13RH TL5R10 G5RZ10VW13X35RW A7PP F10 12W10H A7XL22X QC 10BTV10WW G5RW10RX X5 XV97R XL10W10 Q5H10PW. XL10C 13PV73HC WL5A 13 P13GO 5J G13V10 J5V T10VW5R13P TV7Z13GC V7KLXW, GV73X7Z10 G5TCV7KLXW 13RH 13PP XL13X. 7 H5 R5X XL7RO 7X 7W 13 WXV10XGL X5 XL7RO XL10C A7PP 12W10 Z13K90 X10VQW 5J W10VZ7G10 X5 XV7GO C22 7RX5 13KV667RK X5 XL10Q WGV13T7RK C22V Q10WW13K10W 5V 10Q97PW. 7W 7X W5 L13VH X5 7Q13K7R10 13 A5VPH AL10V10 XL10C G5RX7R90 X5 Z35P13X10 22V GV73X7Z10 A5VO 12RH10V XL10 K87W10 5J F73X7RK GL7R13 X5 XL10 97 W12T10V K10R42W 13VQW V13G10 5V AL13X10Z10V XL10C 13V10 X13PO7RK 13F22X? 7’H W13C 7X'W P7O10PC XL10C 13V10 13PV73HC H17RK XL7W, 13RH 7X 7W W5Q10XL7RK A10 WL22PH TV5X10GX 22VW10PZ10W JV5Q.

import tkinter as tk

# --- Encryption function ---

def encrypt(text, key):

   # Step 1: Assign vowels

   vowels = "AEIOU"

   vowel_map = {vowels[i]: int(key[i]) for i in range(5)}

   # Replace vowels with numbers

   transformed = ""

   for ch in text.upper():

       if ch in vowel_map:

           transformed += str(vowel_map[ch])

       else:

           transformed += ch

   # Step 2: Caesar shift

   shift = int(key[5:]) if len(key) > 5 else 0

   encrypted = ""

   i = 0

   while i < len(transformed):

       if transformed[i].isalpha():

           base = ord('A')

           encrypted += chr((ord(transformed[i]) - base + shift) % 26 + base)

           i += 1

       elif transformed[i].isdigit():

           # Collect full number (single-digit per vowel, but handle properly)

           num_str = ""

           while i < len(transformed) and transformed[i].isdigit():

               num_str += transformed[i]

               i += 1

           num_val = int(num_str)

           encrypted += str(num_val + shift)

       else:

           encrypted += transformed[i]

           i += 1

   return encrypted

# --- Decryption function ---

def decrypt(text, key):

   shift = int(key[5:]) if len(key) > 5 else 0

   # Step 1: Reverse Caesar shift

   shifted = ""

   i = 0

   while i < len(text):

       if text[i].isalpha():

           base = ord('A')

           shifted += chr((ord(text[i]) - base - shift) % 26 + base)

           i += 1

       elif text[i].isdigit():

           num_str = ""

           while i < len(text) and text[i].isdigit():

               num_str += text[i]

               i += 1

           num_val = int(num_str)

           shifted += str(num_val - shift)

       else:

           shifted += text[i]

           i += 1

   # Step 2: Map numbers back to vowels

   vowels = "AEIOU"

   vowel_map = {str(int(key[i])): vowels[i] for i in range(5)}

   decrypted = ""

   i = 0

   while i < len(shifted):

       matched = False

       for num, vowel in vowel_map.items():

           if shifted.startswith(num, i):

               decrypted += vowel

               i += len(num)

               matched = True

               break

       if not matched:

           decrypted += shifted[i]

           i += 1

   return decrypted

# --- Tkinter GUI ---

def encrypt_text():

   text = entry_text.get()

   key = entry_key.get()

   result = encrypt(text, key)

   output_label.config(text=f"Encrypted: {result}")

   print("Encrypted:", result) # <-- prints to console

def decrypt_text():

   text = entry_text.get()

   key = entry_key.get()

   result = decrypt(text, key)

   output_label.config(text=f"Decrypted: {result}")

   print("Decrypted:", result) # <-- prints to console

root = tk.Tk()

root.title("Custom Encryption App")

tk.Label(root, text="Enter text:").pack()

entry_text = tk.Entry(root, width=40)

entry_text.pack()

tk.Label(root, text="Enter key (e.g. 9631804):").pack()

entry_key = tk.Entry(root, width=40)

entry_key.pack()

tk.Button(root, text="Encrypt", command=encrypt_text).pack()

tk.Button(root, text="Decrypt", command=decrypt_text).pack()

output_label = tk.Label(root, text="", wraplength=400, justify="left")

output_label.pack()

root.mainloop()