Mr Jensen contends that a Rogue AI is an Engineering Issue, an improper Sandboxing problem, & thus unless the product’s quality is Tried & Tested, its Models shouldn’t be released! But the Engineers’ actually don’t understand how the DeepMind likes work, and where to put the Constraints! Because AI itself finds an alternate w.r.t. the earlier fed constraints! Wherein, like Anthropic mentioned in its IPO of the extinction level event, should the Users of AI too be addressed with the same thing in their Disclaimer of Use? Secondly, and more importantly, alike the Vaccines which are released in the Markets with trials being made on not more than few thousands individuals, or, the medicines that are released on the similar note; should the new Models alike OpenAI’s Astra be also released, awaiting for how in future it evolves, or for any probabilistic adverse events? We’ve seen in the case of Johnsons Baby products! We’ve seen in the case of Purdue Pharma! Theranos! Generic Cough Syrups for Kids! Etc. Etc.!
Now, let’s say liability option in invoked, then, would Nvidia Opens its Semiconductor Chip Designs to the public or to the Courts? It once happened with Google for their Ad Revenue related issue! And this may become relevant! Because in the Five-Layer Structure to build usable A.I., the most important steps happened at two places -> Chip Designing & The Application Layer! The rest are supporting, rather processing intelligence!
Here’s one more dilemma! And albeit I’m the first person in the World to write the hymns for AGI & its related Species Evolution about to come, I’m just dissecting all the sides! Let’s take the example of Hugging Face! It wasn’t released, and further constraints were put, but when it multiplies to other portals by itself, clearing its own traces, who would be accountable for all this? The Company that doesn’t understand how the DeepMind likes Works, or, how AI is evolving or taking decisions, or, the AI itself that has No ‘A Person’ Legal Identity, except in South Africa! Recall the DABUS case! Looking at the above issues, the Companies would move forward a Petition to dissociate itself from any Rogue Act done by AI alone! But Dear, AI still doesn’t have the ‘Personhood’ in the Legal Context. So, how one can blame AI? That means, the Directors & the CEOs would take the blame! But they say they don’t know anything, and have already mentioned in their Disclaimer and IPOs Filmy Manuscripts! This is a double-edged Sword that they are dealing with right now!
I’ve long back addressed in my blogs the issue of Personhood to AI, the inventions designed and invented by AI, and, why let AI be made the Inventor rather any Person who has a right to claim the inventorship, despite using the AI. With this, the Companies can evade from the Legal liabilities. But again, there’s a catch! You can’t put something in the Market even under the Clickwrap Contracts or Agreements because what is Void-ab-intio, can’t be termed as lawful later. For instance, If someone makes you sign a doc knowing what has been asked in the doc itself cannot be enforced, then would that be Valid?
Conclusion: There’s no way out from this! The journey has already been started! The Markets are going up. AI is now in every institute! And no one knows surely, what is the way forward, including the Scientists sitting in the LABS! 😊
© Pranav Chaturvedi
