From: mk_thisisit
The achievement of superintelligence is considered a near-future reality rather than an exaggeration or distraction from other artificial intelligence (AI) problems [00:00:00]. Experts anticipate its arrival within the next 5 to 20 years, possibly even in about 10 years according to Demis Hassabis [02:44:00]. It is expected to be one of the most, if not the most, powerful technologies ever created by humanity [00:00:13], [03:35:00].
Belief in its Attainment
Both Professor Geoffrey Hinton, often referred to as the “father of neural networks” [01:26:00], and Demis Hassabis, co-founder of DeepMind [02:56:00], firmly believe that superintelligence will be achieved [02:26:00]. They clarify that this is not a marketing ploy by large tech corporations but a long-held conviction [02:28:00]. The recent rapid pace of AI progress indicates that this will happen sooner than previously thought [02:39:00].
Risks and Societal Implications
The immense power of superintelligence necessitates serious consideration of associated risks [00:00:19], [03:42:00].
“I regret not thinking about the security issues earlier” [02:01:00] — Professor Geoffrey Hinton
A key concern is how to maintain control over such systems [02:47:00]. There is a limited timeframe to conduct necessary research on the interpretation and control of these systems [03:44:00]. Beyond technological challenges, the development of superintelligence presents a fundamental societal question regarding its intended use and equitable availability to all of humanity [03:51:00].
Regulation and Development
While AI is a technology requiring regulation, its rapid development poses challenges for creating up-to-date guidelines [05:07:00]. Experts advise governments and social organizations to establish flexible regulations, possibly based on existing frameworks in areas like healthcare or transport, and then dynamically adapt them to evolving technology [05:16:00].
However, the military application of AI, such as autonomous lethal weapons, often remains outside the scope of civilian regulations, leading to an arms race among nations like the United States, China, Russia, Great Britain, and Israel [04:23:00].
Benefits and Democratization
Despite the risks, AI is seen as a powerful tool with significant potential benefits. DeepMind, for instance, analyzed the implications of creating such intelligence early on, aiming to use these tools for scientific discoveries [03:10:00]. AI can aid in treating diseases, addressing energy challenges, and tackling climate issues [03:17:00].
Many tools developed by DeepMind are made available as open source, making the technology increasingly democratic and widely accessible [06:01:00]. AI is likened to a new type of microscope, offering fresh perspectives on reality [06:22:00]. It is expected to become a key stage, or even the main element, of future scientific research [06:42:00].
[!NOTE] While AI is a powerful tool, deep learning methods require vast, well-prepared datasets [07:17:00]. The success in protein structure prediction, for example, relied on the work of tens of thousands of scientists and billions of dollars invested in experimental research [07:28:00]. The future development of AI in other scientific fields will depend on the creation of similarly rich and detailed datasets [07:56:00].
The 2023 Nobel Prize in Chemistry, shared by David Baker and the co-founders of DeepMind, reflects the “enormous influence of artificial intelligence” [09:21:00]. While technology provides tools, human intuition and passion remain crucial in identifying problems to solve [09:59:00].