Tech partner of
What is Twitter saying about ChatGPT?
(Maps viewable on desktop only)
We’ve analysed 45,549 tweets mentioning ChatGPT in 33 languages through our narrative deep learning tool, to reveal patterns in the conversation. You can hover over the dots to read individual tweets. What did we find?
87% of the ChatGPT conversation on Twitter is conjecture. There are different types of conjecture, which you can see by moving the slider on the lower right of the map to the middle position. The actionable lesson? If you are looking to understand the reality around ChatGPT, ensure you have a clear filter for what is speculation/conjecture and what isn’t. Now you can easily separate the signal and the noise, with some machine based help.
Move the slider at the lower right of the map to see how the detailed narratives fit within five higher level themes.
(Please allow 120 seconds for the map to load.)
Move slider to see detailed view
Differences in the ChatGPT conversation by language
See the bar graph below, which is based on the 5-cluster view of the above map. What does narrative analysis tell us?
-
68% of Korean-language Tweets about ChatGPT are future-focused, related to the coming gold rush.
-
Japanese Tweets are similarly focused on narratives of the future (43% of Japanese tweets) along with focus on possible use cases (31%).
-
Portuguese and Arabic Tweets have a higher focus on the implications for people and society (38% and 34% respectively)
-
29.9% of English tweets are focused on ‘my views’ – the highest proportion of that narrative among all languages
-
Who’s posting less conjecture? Russian and German tweets are more soberly focused on ‘ChatGPT here and now – 34% and 18% respectively – versus an average of 13% across all languages.
The sober view: what non-conjecture tweets say about ChatGPT
The middle of the map at the top appears to be much more about the present, rather than the past or future. Further, tweets within this narrative space are less focused on speculation and more on informed observations (versus other parts of the narrative map.)
We took just the data from the middle ‘ChatGPT here and now’ cluster and ran it through the algorithm again, to see what this more grounded, current-day conversation tells us.
Value axis: The horizontal axis is all about assessing the value and potential of ChatGPT.
-
On the left side is a pragmatic assessment of what it is observed to do, and what use cases and applications are relevant
-
The right side is assessing the value of chat GPT less pragmatically - it is more focused on what's right for people and society: Judgements of what should be done, versus should not, with this powerful new tool.
Tool axis: The vertical axis is more about how the tool works and how it is used. The top is focused on the tool itself, whereas the bottom relates to choices and tips on how to use it.
At the center of the conversation is how the ChatGPT 'thinks'. What does this tell us? According this less hype-prone crowd on Twitter - it is essential to have an understanding of how ChatGPT's artificial brain works. You can hover over the dots and use the slider on the lower right to see more detail within this assessment of how the tool thinks.
Move slider to see detailed view
Some background on the tech: Like ChatGPT, our tool leverages LLM’s (large language models) to enable a step-change in how machines process ‘unstructured’ human language. Unlike ChatGPT, our tool is built for analysis, rather than conversation and communication. Phrasia processes thousands of language-related dimensions concurrently to find patterns in the wider conversation. It enables a uniquely nuanced, wide-angle understanding of context.
Move the slider at the lower right of the map to see how the detailed narratives fit within the higher level themes.