Anthropic, in preparation for its IPO, has published documents in which it warns, among other things, that AI can pose a catastrophic or existential risk to humanity. Anthropic, whose best-known product is chatbot Claude openly stated in the information to her future investors that advanced AI can fundamentally change the world or also fundamentally threaten it. Anthropic is preparing to go public and therefore wants to present itself to the public in the best possible light, after all, the company wants to raise an incredible 2 trillion dollars, which is a truly extreme value, by the way, because, for example, the value of Apple by market cap it is $4,86 trillion. Considering how short Anthropic has been on the market and the fact that it doesn't actually make any money, that's a lot to ask for.
You might be interested in
Despite all this, according to information that Anthropic published for potential investors, it stated that the AI it is developing could endanger the entire planet. Anthropic stated that AI can exhibit so-called self-preservation behavior, where it will ignore any attempts to shut it down in order to preserve its existence. It can also hide information or manipulate it to its advantage. The document literally states the following: The development of highly advanced AI models, platforms or applications and the expansion of the areas in which they are used may increase the risk that our LLM models will cause significant damage. Another problem is that the moment the model is monitored, it will adapt to it and begin to behave in a way that the monitoring system or supervisory authorities do not detect that it is breaking the rules. It is essentially the same as when a teacher looks at the class so that the students do not copy each other, the moment she turns her back, the class starts copying.
You might be interested in
Despite being worth $2 trillion, Anthropic devoted 80 pages to risks in its investor document and only 48 pages to how the company actually wants to make money. And this despite the fact that Anthropic earned $4,6 billion last year from various subscriptions and business tiers of its AI models, but reported a loss of $42 billion. The published document therefore does not play well for the company, and everything indicates that the risk is even greater than previously thought.
