ChatGPT knows things that Google doesn't

January 25, 2024 · 5 mins · 1074 words

This morning I remembered a quote I heard when I was a teenager. I don’t know who told me the phrase, or the context, but it was something like this

I’m not a believer, but I prefer my barber to be a Christian, even more when he’s using his razor on my neck.
Some French philosopher

I don’t remember exactly which philosopher was quoted saying this phrase, but it was attributed to someone from the French Enlightenment, someone like Robespierre or Voltaire.

I don’t know why, but this phrase stayed recorded in some recess of my subconscious. After some years, when I was around 20, this phrase floated to the conscious layers of my mind and I tried to search for it on Google to verify its veracity. And I completely failed. After some hours of googling, I was unable to find any reference, so I thought I had probably made up that phrase, and for the next decade I lived thinking that I invented that phrase.

This morning, for some unknown reasons, the phrase popped into my mind again, and I decided to give it another try. Still no luck with Google. But the good thing is now we have more tools than a decade ago! ChatGPT is very good at working at the semantic level - as opposed to Google that works at the keyword level - so I thought it would probably be a good idea to ask ChatGPT about the origin of this phrase. And surprisingly, it knew about it!

My first reaction was of relief because I had finally found the origin of the phrase. However, it’s a known fact that ChatGPT tends to hallucinate answers. So who is right in this history? Does ChatGPT know things that Google doesn’t? If you try the same query with Google you get plenty of useless results.

Here are the possible options that explain this scenario

  1. ChatGPT and I are right because ChatGPT has data Google doesn’t. It’s strange because Voltaire’s writings aren’t an obscure piece of data, so Google should have access to them.
  2. ChatGPT and I are right. The phrase is attributed to Voltaire in a text in some other language, and ChatGPT is internally translating the texts. However, I tried the same query in my mother tongues Catalan and Spanish because it makes sense that if I heard this phrase it would be in one of these languages. I’ve also tried in French, the original language of Voltaire. But still no results.
  3. ChatGPT and I are hallucinating the quote. Which is weird, because it means somehow ChatGPT and I are hallucinating the same things… Maybe given all my experiences, my brain has come to the conclusion that Voltaire said something like that. Is this AGI? Am I ChatGPT? Is this a simulation?

I’m afraid I can’t find an answer to this problem, and I still don’t know if Voltaire ever said something like that. But I’m happy to have taken a small step forward in the solution. I’m using this post to track my progress so that when I ask myself the question again in 10 years I can read it and convince myself that none of this was a hallucination.

P.S. If you happen to know the origin of the phrase, please contact me and free my mind from this decades-long torment.


EDIT 1 (2024-01-27)

Thanks to Grimblewald I know the actual quote was

I want my lawyer, my tailor, my servants, even my wife to believe in God, because it means that I shall be cheated and robbed and cuckolded less often. … If God did not exist, it would be necessary to invent him

Maybe I should rename the post to “Grimblewald knows things that ChatGPT and Google don’t”.


EDIT 2 (2024-01-27)

After this post got some traction in HN I got an email from Andrés where he explained me a similar things that happened to him. In particular, Andrés was working with FPGA from China (Icesugar 1.5 specifically), and apparently there’s no documentation, just a list of specifications in Chinese. If you Google “FPGA Icesugar 1.5 How can I reset the board from inside a circuit?” you don’t get too much information. But ChatGPT knows the correct answer. This is the conversation Andrés had with ChatGPT

#Andrés question
How can I reset the board from inside a circuit?

#ChatGPT answer
To reset the IceSugar FPGA board from inside a circuit, you can use one of the general-purpose input/output (GPIO) pins on the board to control the reset circuitry. The IceSugar board has several GPIO pins that can be used for this purpose, including GPIO0 and GPIO1.

Here are the steps to reset the IceSugar board from inside a circuit:

  1. Connect a GPIO pin to the reset pin on the board. On the IceSugar board, the reset pin is labeled “RESET” and is located near the bottom of the board, next to the USB connector.
  2. In your circuit, set the GPIO pin to a low logic level (0V) for a brief period of time (e.g., a few milliseconds) to initiate the reset. The exact duration of the reset pulse may vary depending on the specific reset circuitry used on the board.
  3. Release the GPIO pin and wait for the board to restart.
    It is important to note that the exact method for controlling the GPIO pins on the IceSugar board will depend on the specific programming environment and tools you are using. For example, if you are programming the FPGA using the IceStorm toolchain and Verilog or VHDL, you can use the “SB_IO” module to configure and control the GPIO pins. If you are using a different programming environment or language, you may need to consult the documentation or community resources for that toolchain to determine the appropriate method for controlling the GPIO pins.

According to Andrés, the answer works and is valid. In his own words, “it is a valid result that I don’t know where it comes from. If you press me for an answer, I would say that it was synthesized from the air”.

If you have more examples where ChatGPT “knew” something that Google completely didn’t know feel free to reach me and I’ll add it to this post.