The biggest question for me about large language model interfaces - ChatGPT, the new Bing, Google's Bard - is this:
How long does it take for regular users (as opposed to experts, or people who just try them once or twice) to convince themselves that these tools frequently makes things up that aren't accurate?
And assuming they figure this out, how does knowing it affect the way they use these tools?
One argument here is that people will blindly trust any chatbot that supports their previous biases
Is that cynicism justified?
What happens when the chatbot speaks against their biases? In particular, what if it both counters their biases AND does so in a way that is demonstrably factually incorrect?
We are already seeing furious complaints from some corners that ChatGPT has a liberal bias - how does that affect how those complainants trust and use these tools?
Hindu nationalists are FURIOUS about ChatGPT right now: https://www.wired.com/story/chatgpt-has-been-sucked-into-indias-culture-wars/
How will that impact their trust of systems like this in the future?
@simon No research, but after an afternoon of 'playing' with Chat GPT, I had worked out it's limitations.
My takeaway, and note of optimism, is that people will be able to 'smell' bot-generated text quite easily. Whether they'll care is another discussion.
@glyph My question remains: if a right-leaning person encounters replies from ChatGPT that directly counters their existing beliefs (and which they can fact check through other sources), do they stop believing that ChatGPT is an infallible source of information?
Even if their conclusion is "It's a conspiracy! The chatbot has been neutered!", does it still provide some level of protection for them in terms of helping them understand that these things are deeply fallible?
The introduction of ChatGPT has garnered widespread attention in both academic and industrial communities. ChatGPT is able to respond effectively to a wide range of human questions, providing fluent and comprehensive answers that significantly surpass previous public chatbots in terms of security and usefulness. On one hand, people are curious about how ChatGPT is able to achieve such strength and how far it is from human experts. On the other hand, people are starting to worry about the potential negative impacts that large language models (LLMs) like ChatGPT could have on society, such as fake news, plagiarism, and social security issues. In this work, we collected tens of thousands of comparison responses from both human experts and ChatGPT, with questions ranging from open-domain, financial, medical, legal, and psychological areas. We call the collected dataset the Human ChatGPT Comparison Corpus (HC3). Based on the HC3 dataset, we study the characteristics of ChatGPT's responses, the differences and gaps from human experts, and future directions for LLMs. We conducted comprehensive human evaluations and linguistic analyses of ChatGPT-generated content compared with that of humans, where many interesting results are revealed. After that, we conduct extensive experiments on how to effectively detect whether a certain text is generated by ChatGPT or humans. We build three different detection systems, explore several key factors that influence their effectiveness, and evaluate them in different scenarios. The dataset, code, and models are all publicly available at https://github.com/Hello-SimpleAI/chatgpt-comparison-detection.
@simon I fully expect chatbots to turbocharge the existing hatred of experts. Instead of "ChatGPT told me something incorrect", most people will just say "hah, ChatGPT proves those experts are wrong."
And if everyone has access to ChatGPT and maybe 1/5 of the population even knows an SME?
Maybe every field will look like medicine does now (even pre-pandemic), where virtually every non-expert believes a half-dozen impossible things they heard once without context.
@django depends on the question!
The bigger problem here is that the models are faster (can generate answers in less than a second) and output everything with an extremely confident writing style
@simon I work in a School Division, and what some are starting to find is that there is currently a limit to how good these are and while the reach of information is very wide, the depth is not.
There are frequent mistakes and downright plagiarism occurring in some of the responses that the Machine Learning system is providing.
This technology will make things easier, but does not eliminate the need to validate the information.
@codedread I have a good understanding of why they lie so much: all they're ever doing is predicting the next word in a sequence of words based on their training data
They have no concept of truth - they just know statistically which words are most likely to follow "The Kennedy assasination was a conspiracy by ..." - based on the TBs of scraper data that was used to build their models
The fact that they get anything right at all is pretty astonishing!
@simon @codedread I really wish they would remove the personification and chat UI. It should not look the same as a box where I send messages to people and people reply to me. User expectations would be better matched to the tools with a better explanation of the prompt and response.
I really get a lot out of using GPT-3 through the playground interface (thanks to your intro, Simon) but I haven't been using the chat interfaces because it feels like the wrong tool.
@Rob_Russell @codedread oh that's really interesting - I hadn't thought about how strongly the chat interface reinforces the science fiction "AI" aspect of it all
The playground interface never seemed to click for a lot of people
Simon Willison
Dan Sugalski
Leigh Garland
Adam Casto
Sean Bleier
Glyph
Ari Koponen
William Gunn
Drew Breunig
mike bayer 
: j@fabrica:~/src; 
Moebeus
mikesten
John Mark 
☑️
Marshall Applewhite Law
Dan Hon
Noah Cook
Django
Julian Elve
Robert Litchfield
Jeff Schiller
Rob Russell