AI is now being weaponized!
How it started?
Today Pentagon ordered Anthropic to restrict their most powerful models, Mythos and Fable, to "foreign nationals," that is, non-Americans, so now AI is also being weaponized. That means even governments can't access these models for military purposes. There was a rule after the Cold War with the Soviet Union that if any tech could help other countries in military development, then the US can restrict it.
Now the question is why Mythos, so the answer is simple and straightforward: it is a really, really powerful tool and the Claude team renamed it as Fable for public launch, which is less powerful than Mythos yet far more powerful than their previous model Sonnet, which is used for coding and really complex tasks. If the real version gets launched, then people can jailbreak the model, which is the hard part, but if they manage it, then they can find vulnerabilities in any website and hack it real quick because of its power.
When it was first tested, it found some deep vulnerabilities that existed for years and top hackers couldn't find them, but it did. Even RBI asked Anthropic to use it and make sure there are no loopholes in Indian banking gateways.
Impact on the World
Since Anthropic is a private company, no one knows how much the stock prices and valuation will go down, but major income comes from private companies and governments, so the expected loss is 5–15%. Current valuation is approximately $985B and the expected loss is around $100B, which is huge because by valuation it is by far the biggest AI stand-alone company.
Now this is not the major impact; the real impact will be on the new world order of AI and its developments. Countries like China and the US have advanced AI, and if governments did this to every company, it will impact countries like India a lot because we never developed our full autonomous AI. We have Sarvam AI, but it is only good in speech and voice things, which is not even close to what ChatGPT, Claude, and other AI companies are doing.
Semiconductors and critical minerals were already weaponized; now it's time for AI and it will happen for sure, and it may slow down AI developments, which is not good from a technology advancement perspective.
Some Good News
Before AI was developed too much, there was a concern that it would take jobs from humans, but now companies are re-hiring. Why? Because AI is good and fast but not sustainable. Uber exhausted its 95% AI budget in 4 months, there is an unnamed company which forgot to use limits on AI usage and employees used AI that was worth $500M, which is insane.
So, will the world change back to normal? Well, not really. We already got huge exposure to AI, so there is no way things will go normal now. AI will be used, but only in complex tasks, and the people who know how to use AI efficiently and know the skill, their jobs are secure. And when I mean "use AI efficiently," then I mean prompt engineering. It is also a skill because we can get things done better if we instruct it properly so that fewer tokens are consumed, yet our project is good enough!
Also, prices of RAM were going up like crazy. Now there is a chance that it may get under control again because of Microsoft's Project Silica, which is a 2mm glass plate that can store 4.8TB of data very fast and efficiently, and we already saw the prices coming down. It will never come to the price it was before, yet they will be down, so that is really good news, I believe!
That's it for today!!! Arnav Chauhan, signing off.