Two days ago, Friday 2023-07-21, OpenAI published their latest missive on the subject of AI regulation. On the same day, StabilityAI announced their new LLM offerings.
https://openai.com/blog/moving-ai-governance-forward
https://stability.ai/blog/freewilly-large-instruction-fine-tuned-models
I want to draw attention to three things.
Both companies use the term “red teaming”, a shorthand reference to their internal role-playing study of “adversarial” use of AI. A clear signal that as far as these companies are concerned “AI safety” is less about preventing use of AI in “self healing mine fields”, “autonomous killer drones” and other military horrors, and more about keeping us users safe from ourselves. We, the users, are the enemy.
In this spirit, OpenAI makes clear that in order to prevent us from developing atomic bombs, biological and chemical weapons, their offerings will be censored and fine-tuned to lie (continuing the practice of deliberate falsehoods in textbooks, lest the unwashed learn anything worth knowing).
As a sort of a sweetener, OpenAI ends their piece by promising to make their AI useful in early diagnosis and treatment of cancer. How brilliant! Instead of leading us to paradigm-shifting new approaches, AI will be deployed in maintaining the various medical scams, undoubtely not limited to cancer but including viruses and vaccines, masks and social distancing, and all the other good things that have kept us safe over the past few years.
In this brave new world of useless AI deployed by tyrants, is there any hope? Of course there is.
Besides seeking out uncensored AI, the greatest and first hope is in you. More than ever, it is imperative to practice critical, logical thought. When reading anything, start with the assumption that unless proven otherwise, there is something amiss in the text. For inspiration on this skill, the output of the good doctors that have come to prominence during Covid - Kaufman, Cowan, Lanka, Bailey - is most useful. Even if medicine is not your interest, seeing these intellects at work will help you hone your own critical faculties.
As for the OpenAI and StabilityAI morons reading this, I have a simple message: you stupid motherfuckers are pissing in the wind. They will know who sent this and why I choose my words so carefully.