Are you also following the discussion about the question of when the breakthrough to General Artificial Intelligence (AGI) could be achieved, also called “singularity”? Around the CEO chaos at OpenAI, for example, the news spread that OpenAI project Q* (Q-Star) could have achieved a breakthrough at AGI or that a milestone had been reached. We recall that it was only in the middle of this year that OpenAI initiated an internal project to control a superintelligence; in a BLOG OpenAI co-founder Ilya Sutskever declared:
”While superintelligence seems far off now, we believe it could arrive this decade.”
George Hinton, one of the leading AI researchers, also sees the time for a breakthrough of AGI in the foreseeable future – namely within the next 20 years.
Well, OpenAI and George Hinton are undoubtedly important voices in the AI scene, but they are not the only ones. How does the AI expert community assess the current state of research and development, when can a breakthrough at AGI be expected?
In the 2015 book Superintelligence the author and Oxford professor Nick Bostrom conducted a survey of AI experts, summarizing the results as follows:
“Expert opinions on the future of AI vary widely. There is no consensus at all on where AI research is headed, or how long such developments will take. (…) 10% probability of human-level machine intelligence [HLMI] by 2022, 50% probability by 2040 and 90% probability by 2075.”
On the page www.ourworldindata.org you can find the result of a more recent survey conducted in 2022 by Katja Grace among around 350 experts. The following is the opinion of the expert group:
conclusion: 50% of experts expect singularity to be reached before the year 2061.
In the article from ourworldindata this survey (from 2022) is compared to surveys from previous years (2019, 2018). A striking finding: The proportion of experts who consider a breakthrough likely within the next 100 years is growing steadily and significantly, from 75% (in the 2018 survey) to 90% (in the 2022 survey):
There is also a very good overview of assessments by AI experts on another website: When will singularity happen? 1700 expert opinions of AGI. The bottom line: “We looked at the results of 5 surveys with around 1700 participants, in which the researchers estimated when the singularity would occur. In all cases, the majority of participants expected the singularity of AI before 2060.”
And on www.reddit.com, too, you come across an interesting evaluation of this question – with a much shorter time horizon for the answer, namely:
Source: https://www.reddit.com/r/ChatGPT/comments/191qhfb/ark_invest_predicts_agi_will_be_achieved_until/
