Wen AGI?
Current average AGI estimate
We have invited many notable people and companies in AI to give us estimates for when they think AGI (artificial general intelligence) will be achieved. We then average the estimates to provide this countdown. When date ranges are given as estimates we use the lower end for our countdown - so the average estimate is a bit more aggressive than the actual estimates might suggest.
Definitions of AGI will vary but we will leave this countdown open to each contributor’s own interpretation of the precise meaning.
The current estimate based on 3 contributions is 2025-10-01.
Using https://onlinetimetools.com/calculate-average-calendar-date
This is by all means not a rigorous scientific approach to the topic of AGI and a formal approach to estimating its “arrival”” date. Also, of course, the topic of AGI itself is not well defined - if it can be precisely defined at all. This is, however, an interesting topic in general and therefore we considered it worthwhile exploring.
Contributors
Here’s the list of current contributors (we have many more invitations pending via Twitter DMs) to the estimate, including some thoughts about the estimate, when provided:
Jacob Valdez - 2023-12-31
Elon Musk? (invitation for an estimate has been sent)
Robert Scoble? (invitation for an estimate has been sent)
Geoffrey Hinton - 2028-04-01 (5 to 20, down from 20 to 50, years as of March 2023, source)
Steve Digital - 2025-01-01
”2022, especially, and the prior years have shown tremendous progress in AI. 2023 will easily trump that in my opinion and with further leaps in emergent behaviours and abilities I expect AGI to be with us no later than the beginning of 2025. This is a very optimistic timeline. I am also weary of what that will mean for humankind but I have hopes that we will use AGI to dramatically improve everyone’s lives worldwide.”
Other contributions
After inviting many people we did also get feedback that basically stated that this article is ridiculous or at least not well-defined enough to warrant an attempt at estimating a date or date range. They are totally legitimate and basically correct, but that didn’t stop us from exploring the topic. Here are some examples.
“Alas, respectfully, I shall decline to predict: if the object of the prediction is so vague, any prediction is meaningless. (plus, I don’t know how I’d measure ‘generations from now’)”
“The whole notion of AGI is poorly specified. It is ridiculous to try to estimate a date for an undefined event.”
“We’re going to sit out this debate.”
In a sense, of course, with AGI not being well defined, we might never know when it arrived (if it does at all) as pinpointing a vaguely defined even precisely, even after the fact, is still impossible. and even when some people claim AGI has arrived others will still disagree. After all human intelligence as comparison and how humans think is also not fully understood, and possibly never will.
Note: If you want to suggest someone we should ask for an estimate, please message us on Twitter, via the contact page or email info@artificial-intelligence.blog.
What do you think?
Please share your estimate or thoughts on the topic in the comments and we will include it, providing it is provided in good faith matching the sincere attempt at having a good representative estimate, in the countdown.
Wen AGI vs when AGI?
It’s just a meme thing.
