At the end of 2027 I'll try to resolve this myself based on available data - if my resolution isn't satisfactory, it can be appealed to manifold moderators if anyone disagrees. I think vague criteria here are better than specific criteria ,curious if anyone disagrees.
"large language models" are the central example of "AI", here
1,000
3.00Workable voice assistants and speech to text stuff has been around for years but people overwhelmingly prefer to type.
As it stands right now vastly more people type in ChatGPT than use voice mode. Since voice mode is fully competent and could hardly improve more in terms of its speech recognition and its voice realism it's hard to see why the market think it's a 72% chance.
@jacksonpolack What sources will you be using to resolve this? If you take Apple's Siri usage stats on face value (500m installed devices 19% or 95m of which are active users) and add active Alexa users (71m). That is greater than the number of active ChatGPT subscribers (150m). So, how exactly do you define AI in the context of this question? And at what point does Siri, Alexa and ChatGPT meet the criteria of AI? Would you define the NL classifiers used by social media companies as text based AI that all the users of those platforms interact with?
What sources will you be using to resolve this?
Whatever sources are available! I don't even know what gen AI products will be popular in dec 2027, so I can't really know which sources right now. There will probably have to be some judgement calls in comparing different available statistics. If people think I'm biased, then as the description says a mod (eg Joshua, evan, etc) can do the comparison instead, they're more than capable.
So, how exactly do you define AI in the context of this question?
Sure, I'll edit that. In essence, a large language model is AI, talking to a LLM via Siri is also AI, a pre-LLM Siri isn't AI. Text classifiers clearly aren't in the spirit of the question.
The Economist pointed out several things that I found convincing.
https://www.economist.com/briefing/2024/01/25/could-ai-transform-life-in-developing-countries
"Start with education. A typical sub-Saharan pupil spends six years in school but retains only three years’ worth of learning, Wolfgang Lutz of the Wittgenstein Centre in Vienna estimated in 2015. A typical Japanese student spends 14 years in classes and absorbs 16 years’ worth of education. Using a different methodology, the World Bank also finds that education is spectacularly worse in poor countries than in rich ones (see chart 2).
Tonee Ndungu, an entrepreneur in Kenya, thinks ai could help bridge this gap. He has developed two apps that he hopes to launch this year. One, called Somanasi (“Learn with me”), is for children. It allows pupils to ask a talking chatbot questions related to the Kenyan school curriculum. The Economist asked, “How do I work out a percentage from a fraction?” The chatbot offered a step-by-step worked example."
https://www.economist.com/leaders/2024/01/25/the-tantalising-promise-of-ai-for-the-emerging-world
"As our briefing this week sets out, policymakers and entrepreneurs around the world are exploring ways that ai can help. India is combining large language models with speech-recognition software to enable illiterate farmers to ask a bot how to apply for government loans. Pupils in Kenya will soon be asking a chatbot questions about their homework, and the chatbot will be tweaking and improving its lessons in response."
Given that cell-phone access is near-ubiquitous, many people have limited literacy/typing fluency, and the AI of three years from now will be better at helping with daily tasks, it seems like there is a plausible case here.
I think this mostly turns into whether people end up interacting via voice more.
The tech is already here for a better phone integrated voice assistant, just hasn't been made.
I don't think I have an actual sense of how many people use Siri, or even just use voice to search Google.
(But I mostly avoided that because of eh voice recognition. Whisper isn't perfect, especially because it isn't smart and corrects uncertain pauses like you might want, but it is still far better. Also Google by 2027 may end up having AI typically integrated into the search output which would up the number significantly, though it would also up the text ai usage too)
@Eesl seriously, have any of these tried whisper? also on device ai chip powered llms will easily make this happen by 2025.