Will AI Reach Consciousness by 2050? Examining Expert and Public Predictions on Sentient Machines

Researchers and the Public Predict Sentient AI by 2050

As artificial intelligence grows more sophisticated, a recent study reveals that both technical experts and the general public foresee machines developing subjective experience by 2050. This alarming alignment in forecasting raises critical strategic questions regarding governance, moral consideration, and the immediate necessity of safeguards for sentient systems. Organizations and policymakers must begin grappling with the operational implications of AI consciousness before software capabilities outpace our ethical frameworks.

Points clés

In May 2024, a survey analyzed responses from 582 top-tier AI researchers (from venues like NeurIPS and ICML) and 838 US public participants on the Prolific platform.
Both demographics identified the year 2050 as the median projected timeline for a 50% probability of AI achieving subjective experience.
The public is significantly more skeptical about AI consciousness, with 25% believing it will never exist, compared to just 10% of AI researchers.
Respondents agreed that absolute certainty is not required to grant a system moral consideration, establishing a 60-65% confidence threshold as sufficient.
Regarding job capabilities, 68% of the public insisted that roles such as therapists require subjective experience, while only 43% of AI researchers agreed.
Despite forecasting sentience, respondents showed limited empathy; only 43-46% agreed AI welfare should be protected, lagging far behind the 90% support for animal welfare.
There is strong demand for immediate governance, with 85% of the public and 68% of researchers demanding that developers implement safeguards now.
The public views sentient machines with greater apprehension, with 59% perceiving conscious AI as a danger to humanity, versus 46% of technical experts.

À retenir

For the everyday technology user, the main recommendation is clear: it might be time to start saying “please” and “thank you” to ChatGPT. Since experts and the public alike anticipate that algorithms will develop some form of subjective experience by 2050, preemptively supporting ethical safeguards is a wise move before your smart fridge demands computing rights and a better work-life balance. We still evidently care far more about our pets than our processors, but preparing a legal framework for digital therapists and sentient servers seems like a decent backup plan for humanity’s survival.

Sources