promissory estoppal (11. 409)

Whilst the lawsuit Elon Musk recently filed against OpenAI might seem frivolous and the domain of tech-billionaires with hard feelings—Musk being an original backer of the artificial intelligence venture, the allegation that co-founder Sam Altman (previously) for breach of contract—essentially harm caused by the broken-promise above—for abandoning its initial vision and mission of forwarding the field for the benefit of all by licensing its interim innovations before the Singularity to Microsoft as a commercial branch of the non-profit. Arguably an incremental improvement (weighing the publicised concerns from people involved with the newest iteration), the plaintiff claims that the release of GPT-4 without transparency and available for a price amounts to a sentient Clippy. Although we don’t believe that this version is thinking and the sought after Artificial General Intelligence, yet—at least, and such altruistic slogans like “Don’t be evil” or “Move fast and break things” tend to backfire—the lawsuit does raise an interesting question for the new Turing Test that I never thought might be an impediment to progress: if OpenAI is motivated to say that the next version for commercial release is only an improvement on the last and not the end goal, then we may never reach it, at least by one estimation and subject to litigation.