Musk's Ilya private email was decrypted by Claude, OpenAI coding information was made public, and Google was hurt

OpenAI and Musk were at loggerheads, but inadvertently revealed Claude 3’s new skills.

Because netizens are not only focused on the matter itself, but also what is written in the coded part of the email has become a hot topic.

So, an AI blogger published the results he had deciphered using Claude 3, and the post was read more than 630,000 times.

In this "riddle game", Claude 3 first revealed what netizens discussed the most, which is this sentence:

Unfortunately, the future of mankind lies in the hands of [? 】Hand

# By analyzing the length of the coding part, Claude quickly gave the first version of the answer - Google.

△The framed part is the content deciphered by Claude

But sharp-eyed netizens soon discovered that the length of the word "Google" seems to be inappropriate when placed here. Mismatch.

However, some people say that it is not necessarily the complete word "Google", but may also be referred to by an abbreviation such as "goog".

But no matter whether it turns out to be Google or not, this speculation has stirred up waves. More and more netizens are beginning to wonder about this person who Musk calls "the helmsman of the future of mankind." ”…

Netizens relayed on solving the puzzle

First of all, the blogger himself was the first to express in the comments that someone pointed out that the person who was coded here should be Demis (the founder of DeepMind).

At least it is closer than "Google" in terms of length, and the identity seems reasonable.

#But after considering the character’s experience, netizens quickly proposed a new candidate—Google co-founder Larry.

The reason given is that Musk is very worried about Larry's position and believes that the latter wants to destroy mankind, but I have never heard Musk say anything about Demis.

#While netizens are sorting out these past events, new progress has also come from the AI ​​cracking side.

Baoyu, a well-known AI blogger, discovered a clue from the front-end code of the OpenAI announcement page -

This seemingly continuous large black bar is actually composed of small blocks. , and the length of each block is different.

So he used front-end technology to separate these chunks, and then sent a prompt to the original blogger.

After such improvements, Claude 3 made another recognition, and this time the result became..."idiots".

However, when the following sentence was separated, Cladue couldn't guess it.

Although the identity of the mysterious person becomes more confusing, this explanation also makes some netizens feel that it seems more reasonable.

Of course, no matter whether it is changing the prompt words or combining it with the actual situation, these speculations cannot be confirmed or falsified, unless one day OpenAI or Musk himself stands up to explain.

The blogger who posted the post also came out to remind everyone that just watch it for fun and don’t take this speculation too seriously.

But in addition to the content of the letter, Claude 3’s performance also attracted a wave of attention.

Some people even said that Claude 3’s prediction made him feel that the hallucination phenomenon of large models is not a bug, but a feature!

One More Thing

Regarding the fight with OpenAI, the latest news from Musk is that he released such a picture early this morning.

The original picture is a photo of Ultraman negotiating as a visitor after he was expelled from OpenAI. The visitor tag he was wearing was P-written with the text "Closed AI", and Musk also wrote "fixed it" .

#But someone immediately asked, is xAI open source?

However, the founder of Anthropic, the company behind Claude, also chose to leave to start a business because he was dissatisfied with OpenAI becoming "closed".

At this time, OpenAI was at loggerheads with Musk, and it was even reported that the product launch was postponed due to this matter, which indeed provided Claude with time to catch up.

The above is the detailed content of Musk's Ilya private email was decrypted by Claude, OpenAI coding information was made public, and Google was hurt. For more information, please follow other related articles on the PHP Chinese website!