To calculate a mean value, we need to consider the range of predictions provided. Since the predicted dates range from “now” to at least 2100, the first approach should be to estimate the mean by taking the midpoint of this range. Note that this mean value is not weighted.
To refine this, we have to consider potential factors that could influence the timing. We added a (-/+ years) estimation for each factor, with (-) bringing the singularity day closer, and (+) when it makes the date farther away.
The following information and aspects, at least, should be taken into account:
- Current AI advancements: Assessing the current state of artificial intelligence research and development, including breakthroughs, limitations, and progress toward achieving artificial general intelligence (AGI). Tracking advancements in areas such as machine learning, natural language processing, and robotics can provide insights into the direction and pace of AI development. (-10)
- Computational power and hardware advancements: Monitoring the progression of computing power, such as Moore’s Law, and the development of specialized hardware architectures optimized for AI tasks. Improvements in computational capabilities can potentially accelerate AI advancements and contribute to the emergence of AGI. (-10)
- Funding and research investments: Observing the level of funding and investment in AI research and development from various sectors, including governments, private companies, and academic institutions. Funding can fuel research breakthroughs, attract talent, and drive innovation in the field. (-10)
- Technological convergence: Considering the interplay between different technological domains and how advancements in areas like nanotechnology, biotechnology, and information technology can synergistically contribute to the development of AGI. Exploring the potential for interdisciplinary breakthroughs and their impact on AI progress. (-10)
- Ethical and regulatory considerations: Recognizing the significance of ethical and regulatory frameworks in shaping the trajectory of AI development. The establishment of guidelines, policies, and safeguards can impact the speed and direction of AI advancements by addressing concerns related to safety, transparency, accountability, and fairness. (+10)
- The continuous neutering of models (the process of making the models less capable or reducing certain aspects of their functionality to prevent them from generating inappropriate, harmful, or sensitive content. This could involve modifying the model’s training data, adjusting its parameters, or implementing filters to ensure that its outputs align with certain guidelines or standards. Neutering is often done to ensure the responsible and ethical use of AI models in various applications, such as chatbots and content generation platforms.) (+5)
- (NEW) New advances in AI hardware are making the singularity more likely. AI systems will be able to learn and process information much faster, which could lead to a breakthrough in AI capabilities. These advancements include quantum computing and neuromorphic computing, but more specifically the rise of affordable models like NVIDIA H100 and, more recently, GH200 models. (-3)
Again, please remember that predicting the singularity is extremely difficult and uncertain, that is why we use rounded numbers for the calculations.
Main References
- In June 2023, Ray Kurzweil published The Singularity is Nearer, a fresh perspective on advances in the singularity—assessing the progress of many of his predictions in his previous book and examining the novel advancements that, in the near future, will bring a revolution in knowledge and an expansion of human potential.
- In April 2023, AIMultiple summarized 1700 experts´ opinions on AGI and singularity. It also introduces a reference to the over-optimistic predictions of the experts in the past, and why some believe that we will not reach AGI.
- In 2022, ESPAI rerun the2016 Expert Survey on Progress in AIthat researchers at AI Impacts previously collaborated on with others. Both surveyed authors who recently published in NeurIPS and ICML, major machine learning conferences.
- In 2019, Emerj, an AI Research company, conducted a survey among researchers in the AI field and asked them about when AGI and the singularity are estimated to take place.
- In 2013, Vincent Muller et. al. conducted a 550-participant survey about when high-level machine intelligence is estimated to occur. They published the paper “Future Progress in artificial intelligence: A Survey of Expert Opinion” with the results and analysis.
- In 2011, S.eth D. Baum et al. published in Technological Forecasting & Social Change, 2011, an analysis of expert assessments on their AGI-09 conference, related to previous references.
- In 2009, Baum et al.surveyed 21 attendees of theAGI-09conference, onAGItimelines with and without extra funding. They also asked about other details of AGI development such as social impacts, and promising approaches.
- In 2005, Ray Kurzweil published the well-known non-fiction book The Singularity Is Near: When Humans Transcend Biologyabout artificial intelligence and the future of humanity.
- Metaculus, an online forecasting platform, and aggregation engine, maintains an up-to-date collaborative survey estimation on different areas, including AGI and related concepts.