George Rebane
The recent burst of large language models (LLMs) from AI research has reinvigorated the Singularity related topics of machine sentience, sapience, super-intelligence, agency, ethics, morals, …, along with prognoses of what they may have in store for humans when/if they become ascendant. Most commentators, including yours truly, see a high likelihood of a dystopian post-Singularity future for mankind.
Among these well-known AI experts are internationally renowned scientists like Nick Bostrom (e.g. Superintelligence: Paths, Dangers, Strategies, 2014) and Max Tegmark (e.g. Life 3.0: Being Human in the Age of Artificial Intelligence, 2017). These tomes are worth reading more than once, which I am currently doing. It turns that both Bostrom (Oxford) and Tegmark (MIT) are established physicists, philosophers, and AI researchers (and both were born in Sweden).
Their counsel for humanity to proceed carefully into a future not dominated by our species is countered by more hopeful futurists like Hans Moravec (e.g. Mind Children: The Future of Robot and Human Intelligence, 1990) and Ray Kurzweil (e.g. The Singularity is Near, 2005).
About the Singularity, Moravec (Carnegie-Mellon) advises his readers that “human equivalence is just the beginning, not an upper bound. Once the tireless thinking capacity of robots is directed to the problem of their own improvement and reproduction, even the sky will not limit their voracious exploration of the universe. … Moravec challenges us to imagine with him the possibilities and pitfalls of such a scenario. Rather than warning us of takeover by robots, the author invites us, as we approach the end of this millennium, to speculate about a plausible, wonderful postbiological future and the ways in which our minds might participate in its unfolding.”
And it is our postbiological future that I would like to expand on here. Moravec, Kurzweil, and others of their ilk have been describing a post-Singularity world that is inhabited by both intelligent machines and various forms of ‘transhumans’ – humans who have shed their evolved mortal coils and taken up new forms in both the physical and computational sense. A popular scenario has some of us living in cyborg bodies. Another inviting future has us living in virtual worlds as computational entities, and still others have us co-habiting the material (real?) world with superintelligent machines exercising some manner of peer relationships.
Sandbox – 30apr23
[So now outfits that collect and search internet content want to start charging AI companies that make chatbots for the data they use to train their LLMs. Companies like Google and Amazon get this data for ‘free’ – actually, what it costs them to scrape it from the web – as it is provided also for free by billions of online consumers. The conundrum is that the very large data repositories don’t want to drive away the AI-based search engines by making their training datasets too expensive. Until now the whole purpose of those freely accessible datasets has been to have third-party search results point to the data that the Googles and Amazons make available. And since it costs tens of millions of dollars in power and processing time to make a functioning LLM, everyone involved already has skin in the game. If it ain’t broke, … . gjr]
Posted at 10:15 AM in Comment Sandbox | Permalink | Comments (222)
Reblog (0) | |