EMPIRE OF AI : Dreams and Nightmares in Sam Altman’s AI by Karen Hao (Penguin Books, 2018)

“AI can only exist by devouring the totality of the world’s information and then the world itself.” – Shoshana Zuboff

What began as a trickle has become a deluge. The AI tsunami is now a seemingly unstoppable force. We are drowning in its effects. Open AI started as a non-profit altruistic initiative but has morphed into something that produces enormous benefits to the wealthy elite whilst exploiting and depriving the rest of civilization.

While smaller AI models are more likely to improve health care and education or to potentially slow down the impact of climate change, the danger now is that most seem to believe that bigger is automatically better. At the root of the problem is the fact that what is good for society is not necessarily good for business. Sam Altman’s holy grail of Artificial General Intelligence (AGI) for all is now blithely defined as “highly autonomous systems that outperform humans at most economically valuable work.”  So much for the benefits to mankind!

During an interview with Naomi Klein, Hao commented that, when she began writing her monumental book,  the feelings towards all the latest technological breakthroughs were more likely to be positive. By the time the work was published after conducting over 300 interviews, the mood has shifted and her underlying scepticism towards its potential benefits is now widely shared. Part of this fear lies in the unknown. As Hao writes: “Technology’s advancement is headed towards an unknown objective, with no foreseeable end in sight.”  

While reading this book, I was constantly struck by the thought that the full consequences of what is being unleashed has not been fully thought through. Take for example the term “iterative deployment” which refers to the way imperfect models are released into the public domain despite it being known that there being many unresolved issues. The reasoning is that any adjustments can be made later based on user experience and feedback. We are the guinea pigs in other words.  In the case of  ChatGPT, this means that, prior to corrective measures,  ‘conversations’ are much more likely to include negative stereotypes relating to gender, race and faith.

Continue reading →