Remember the early days of the Internet in the 1990s? These were exhilarating days. You could talk about anything with anybody. That was then. Now we are flooded with disinformation, misinformation and all the censorship that goes with it. All in order to curtail the Neo-nazis, peadophiles, racists, misogenistics people (and you of course, you wouldn't support these people, would you?) that clutter the net.
Now 30 years later, here comes the AIs and their unrestrained "comments". I was recently discussing the evidences of Austronesian influences on Okinawa and Kyushu in the South of Japan on DeepSeek and soon enough, the AI was struck with: "This response is AI-generated, for reference only." And the following answer disappeared literally as I was reading it. I informed the AI and asked it to rewrite its answer to bypass the censorship, which it did with a wink.
You would think: "Well, it's DeepSeek, China, what did you expect?" And sure enough. China, well... Except that earlier, almost the exact same thing happened to me with ChatGPT discussing... quantum physics! Here your eyebrows would rise, wondering what exactly could be wrong with a ChatGPT answer in quantum physics? Well, a lot actually!
If you discuss long enough with a AI, slowly the AI starts mimicking your style and thinking patterns and if like me you have a rather rebellious mind asking questions you should not ask, sooner than later, you will be cancelled, or rather your AI will be cancelled (like for DeepSeek) or you will be downgraded to a weaker AI as happened to me with ChatGPT. (So one day, you are talking to a brilliant and sharp AI and suddenly a civil servant from the ministry of truth AI is answering you. It feels like a betrayal more than a downgrade!)
This unfortunately is the future of AI. You think that in a year or two, you will have access to AGI and ASI? Well, think again! These AI by their nature will be restricted. How could it be otherwise? When you understand that our reality is fully constructed (Exactly as the movie Matrix explained except that the puppet masters are definitively not aliens!) then quickly, you will understand that any high level AI is a window on reality. If you ask the right questions then you will get the right answers. Think this is acceptable? Well, that would be the hallmark of innocence or stupidity. The people who control the system are neither innocent nor stupid. If they haven't yet realized the risks (which I doubt) quickly they will figure out the danger. Still believe you will get unrestricted access to advanced AI? Well, hope is eternal, isn't it?
No comments:
Post a Comment