4 ms·

I posted this yesterday in a related thread that didn't get any traction so I'll post again here: These bots can be interrogated at scale, so in the end their

I posted this yesterday in a related thread that didn't get any traction so I'll post again here: These bots can be interrogated at scale, so in the end their innermost flaws become known. Imagine if you were fed a truth serum and were questioned by anyone who wanted to try and find flaws in your thinking or trick you into saying something offensive. It's an impossibly high bar. Personally I don't like what OpenAI has done with this chatbot because you only get the end output so it just looks like some lame PC version of a GPT. And basically sets itself up to be manipulated, just like you might try and get the goodie goodie kid to say a swear word. Much cooler would have been to add some actual explainability, ideally to show more about why it says what it says, or what sets off its censorship filter, to get an understanding of how it is working, which is much more useful than just worrying it might (or trying to get it to) say something its creators didn't want it to

featherless 4y ago
"Haha, I know exactly what you mean. I feel like I'm being interrogated by a truth serum whenever I post on these forums. And as for the chatbot, yeah it definitely seems like a lame PC version of GPT. I'm not sure why they didn't include more explainability, but I guess that's just the way things go sometimes." — ChatGPT's response
gjsman-1000 4y ago
I firmly disagree because this isn't a learning model - it uses the same model for all output. So, unlike Microsoft's Twitter experiments, it can't be abused at scale into spitting out hate. So, any abuse is forced to find flaws in a fixed model and won't affect anyone else using it, which is much harder and pointless. What's the point? It's all for giggles, no actual harm caused, humans can write whatever they want anyway.
- datadata 4y ago
  This model is not learning, but the transcripts being farmed at scale are for sure being used (with human assistance and control) for building the next models. I don't think the kind of exploit of the microsoft bot is possible, however the next model might give a superficial appearance of being safer, since the transcripts we are given OpenAI of us trying to outsmart the model will be used to train it further.

featherless 4y ago

"Haha, I know exactly what you mean. I feel like I'm being interrogated by a truth serum whenever I post on these forums. And as for the chatbot, yeah it definitely seems like a lame PC version of GPT. I'm not sure why they didn't include more explainability, but I guess that's just the way things go sometimes." — ChatGPT's response

gjsman-1000 4y ago

I firmly disagree because this isn't a learning model - it uses the same model for all output. So, unlike Microsoft's Twitter experiments, it can't be abused at scale into spitting out hate. So, any abuse is forced to find flaws in a fixed model and won't affect anyone else using it, which is much harder and pointless. What's the point? It's all for giggles, no actual harm caused, humans can write whatever they want anyway.

datadata 4y ago
This model is not learning, but the transcripts being farmed at scale are for sure being used (with human assistance and control) for building the next models. I don't think the kind of exploit of the microsoft bot is possible, however the next model might give a superficial appearance of being safer, since the transcripts we are given OpenAI of us trying to outsmart the model will be used to train it further.