Zuckerman the philosopher
The urge for tech owners to write manifestos isn’t new, and Meta founder and CEO Mark Zuckerberg added to this burgeoning literature last week with his 6,500 word defense of AI. In it, he speaks as a tech optimist, telling people that AI is ok, that it will not take away people’s jobs, that it should be open sourced, that innovation shouldn’t have guardrails and that AI will enrich human life and not cheapen it. He is baffled by the increasing backlash against AI use leading to layoffs, against data centers, against rogue AI’s, including Meta’s, hacking businesses. “I do not understand why anyone who believes that AI will eliminate most jobs and much of humanity’s relevance would rush to build that future,” he writes.
Mark Andreessen, the co-founder of Netscape, wrote something similar in his 2023 manifesto, “We believe growth is progress—leading to vitality, expansion of life, increasing knowledge, higher well being.” In a 2024 manifesto, OpenAI’s Sam Altman wrote that “We need to act wisely but with conviction.” What Zuckerberg’s manifesto makes clear is that things are not looking too rosy right now for AI companies, their unending investments and zero profits. There is a growing number of influential people calling for AI companies to be regulated, and the recent rogue AI episodes—Meta, OpenAI, Anthropic were all culpable—and the backlash against Meta Glasses may have unnerved Zuckerberg a bit.
A case of watermarks
When AI regulations bite, companies have to change tack. That was first outcome of the European Union’s Artificial Intelligence Act, which came into force on 2 August. Under Article 50 of the Act, AI system providers must disclose that the images, text, audio or video being generated with agents must be disclosed as such. And that chatbots must at all times disclose to human users that they are dealing with AI. As a result of this, Anthropic has announced that content generated by its AI models will be watermarked. “Because the watermark is part of the text, it will travel with the text when it’s copied and pasted elsewhere, and may persist through some editing. Watermarking will be applied at the model level, which means it will be present no matter which Claude product or surface the text comes from.”
Now, Claude-generated content will also carry hidden watermarks, but detectable in the code. Similarly, Google announced on Friday that users can remove visible watermarks from Gemini-generated content, but invisible SynthID watermarks and C2PA metadata will remain. This way, Google reckons, users can pass off AI-generated content as their own, but the truth of its generation will remain embedded in the code.
