The final weights in themselves don't tell you anything about what went into the training process, e.g censorship.

The question is more whether this game exists as open source somewhere in the training data (probably does).

You can't possibly think those models are only trained on open source data?

Makes me think of that part in Philip K. Dick's Do Androids Dream (..) -- where Deckard reflects on the androids' indifference to their imminent deaths, saying that this was due to them lacking the aversion to death acquired trough evolution.

100%. Companies are paperclip optimizers, with money as the objective. For example, Uber used ride data to circumvent investigations by regulators. There is absolutely no reason to assume that AI companies would not use their data in any way possible to reach their objectives.

Hopefully they password-protect the datasets:

https://arxiv.org/abs/2305.10160

What a usability nightmare this site is: 3-4 popups before I could even read the title. No thank you. And this is with an adblocker turned on.

Don't these sites realize how many users they're losing?

That adblocker does not sound very effective

No popups when using uBlock Origin and/or uMatrix

   scheme=https://
   host=simpleflying.com
   ip=34.233.113.241
   path=/united-airlines-767-returns-newark-bluetooth-name-alert/
   {
   echo url=$scheme$host$path
   echo output=/dev/stdout
   } \
   |curl --resolve $host:443:$ip -K/dev/stdin \
   |sed 's/<img src=[^>]*>//;/user-comment/,$d' \
   |grep -o "<p>.*</p>" > 1.htm
   firefox ./1.htm
   #links -dump 1.htm

The real "nightmare" is the browser that will automatically run all that garbage returned in the response body without any input from the user

It requires an "adblocker" to stop its default behaviour

Alternatively, one needs to disable Javascript, restrict the browser's access to DNS, etc.

When an advertising company releases a "browser" that intentionally allows website operators to cram pages fuil of advertising and tracking is that a coincidence

Is that the only way a browser can be designed

No

How many users realise this

A small number

For example, I'm using a browser that cannot automatically request resources, run Javascript, CSS, etc. where HTTP headers, including cookies, are trivial for the user to create, edit, save and delete. I do not need an "adblocker"

"Don't these sites realise how many users they're losing?"

The number is so small why would they care

My eyes, my eyes! (closed the tab after 4 seconds)

ChatGPT is the only bot that reliably cites sources (through Web search mode).

The other bots either make up links or simply don't provide any information that is distinguishable from the LLM predictive output.

Ironically Gemini is also very bad at this, while it should have been the best at Web search.

Gemini also does something very patchy, which is to provide "links" which are in fact GET queries into classic Google search. I'm guessing they did it this way because the links generated/hallucinated by the LLM were too unreliable.

Interesting angle, didn't think of this. How do you think/find that current tools are optimized for being addictive?

I think there's a few things, but its a little subjective and its more about the style the ai uses when doing these than the actual specific behavior:

- Nuggesting improvements to the code after finishing the task you gave it, very irritating when the improvements were obvious and the ai didn't implement them on its own

- Not trying very hard when implementing something, leading to bugs, which leads to more tokens used (this behavior can be incentivized and learned with RL)

Since its a known fact if a user continues a session after the LLM says something, its not hard to train against this. The least efficient way to do this would be to GPRO directly against the user base and try to get as many people talking to the AI, and with OAI having a billion monthly active users the least efficient method would work really well for them.