AI is inexplicable, unpredictable and uncontrollable. A new book warns about developments in artificial intelligence
Dr. Roman V. Yampolskiy is a tenured associate professor at the University of Louisville and has published the results of his research in the book entitled “ AI Unexplainable, Unpredictable, Uncontrollable ” “Artificial Intelligence, Unexplainable, Unpredictable, Uncontrollable ”. In this book the expert, carrying out an in-depth analysis of the scientific literature on AI, concluded that there is no evidence to support the idea that AI can be controlled safely.
Even the measures that could be taken to prevent accidents and improve safety would be insufficient.
His work on AI safety has been partially funded by Elon Musk and the Future of Life Institute (FLI), founded by MIT professor Max Tegmark and Skype co-founder Jaan Tallinn. FLI is the same institute that called for a six-month moratorium on the development of AI tools more powerful than OpenAI's GPT-4.
The institute's open letter was signed by more than 33,000 engineers, including Musk, Apple co-founder Steve Wozniak, the "godfather of artificial intelligence" Yoshua Bengio and other experts, as Interesting Engineering reports.
Concern about managing and regulating AI risks
As new AI tools have entered the market over the past year, Dr. Yampolskiy suggests that advanced AI systems will always present risks, despite their potential benefits, due to their inherent unpredictability and autonomy.
“Why do many researchers think the AI control problem is solvable? As far as we know, there is no proof of this, no demonstration. Before embarking on the search for controlled AI, it is important to demonstrate that the problem is solvable,” Dr. Yampolskiy said in a press release.
“This, combined with statistics showing that the development of AI superintelligence is an almost guaranteed occurrence, demonstrates that we should support a significant AI safety effort,” he added.
The development of superintelligence is inevitable
Since AI, including superintelligence, is capable of learning, adapting and acting semi-autonomously, it becomes increasingly difficult to ensure its safety, especially as its capabilities increase.
It can be said that superintelligent AI will have a mind of its own. So how can we control it? Are Isaac Asimov's laws of robotics, which establish ethical behavior between humans and robots, still valid in today's world?
Dr Yampolskiy, whose main area of interest is AI safety, explained: “We are facing an almost guaranteed event that could cause an existential catastrophe.”
“No wonder many consider it the most important problem humanity has ever faced. The result could be prosperity or extinction, and the fate of the universe hangs in the balance,” he added.
A major challenge is the infinite potential decisions and failures of superintelligent beings, which make prediction and mitigation difficult.
Furthermore, the lack of explanations in AI decisions raises concerns, as understanding these decisions is crucial to reducing incidents and ensuring bias-free outcomes, particularly in critical sectors such as healthcare and finance, where the AI has penetrated.
As AI increases autonomy, human control decreases, leading to safety issues. Dr. Yampolskiy proposes a balance between capacity and control, recognizing that superintelligence may inherently be uncontrollable. So the level of intelligence would have to be limited to allow for some level of control, however.
He also discusses the importance of aligning AI with human values, just as Asimov did with his famous three laws of robotics, suggesting ways to minimize risks, such as making AI modifiable, transparent and classifiable as controllable or not controllable.
Despite these challenges, Dr. Yampolskiy encourages continued research and investment in AI safety, noting that while achieving 100% safe AI may be difficult, one must still strive to achieve a good level of safety.
Thanks to our Telegram channel you can stay updated on the publication of new Economic Scenarios articles.
The article AI is inexplicable, unpredictable and uncontrollable. A new book warns about developments in artificial intelligence comes from Scenari Economici .
This is a machine translation of a post published on Scenari Economici at the URL https://scenarieconomici.it/la-ai-e-inspiegabile-imprevedibile-e-incontrollabile-un-nuovo-libro-mette-in-allarme-sugli-sviluppi-dellintelligenza-artificiale/ on Tue, 13 Feb 2024 09:00:38 +0000.