The AI apocolypse is going to be:
- an internet consumed by bots and AI slop (such as a Facebook where it's all AI bots chatting with each other while we watch)
- an economy in ruins because all the bosses want to hire AIs, not people
- social and environmental destruction from AI and big data companies influencing our government (the real deep state)
- epidemic AI psychoses in a society dumbed down by always-available, not-always-correct advice
- and more

If AI "escapes the box" it's because some human designed it to be so. Humans will do this, just like they maliciously deploy viruses on the internet.

Some viruses are made just cause destruction, not even to tangibly financially benefit their makers.

If AI actually kills people, it's because someone put in a box with access to guns or more, like killer robots.

.... and, of course, there are pros for all of these cons. And remedies we can apply.

I did not take the article to be dismissing those concerns. I took it to be dismissing the idea that there is a "self" inside a chatbot that has some sort of will-to-power that is malevolent and can be beyond control. A ghost, or a demon if you will, inside the machine. If you understand what actually is inside a chatbot and how it works, that extreme anthropomorphism and panicky reaction can be dispelled.

I do not share your optimism.

The article doesn't dismiss my concerns. It ignores them to make its meager point.

.... on a specific point. A peak under the hood.

And I do, my self, tend to be optimistic, hopefully in a realist way. We shall see....

.... this news item about AI and local journalism: https://www.democraticunderground.com/1016424278

... AI and its affect on society - which you seemed to be - you may want to keep up on such developments, positive or negative.

Sorry for having bothered you with something I thought you might want to keep track of.

Sarcasm because I certainly do not need you to educate me about the evils of AI.

.... I won't bother you further with matters of such complexity.

.... on reading discussion threads.

You really lost the thread. Enough of you.

... some education about about the evils of AI. So, enough!

I was the one who pointed out that the article was insubstantial. And you think you need to prove it to me.

No last word for you. Just ignore.

...disguising the reality that there's not much there when you open the box.

If you can convince a certain sort of business owner or bureaucrat that AI can replace employees they'll jump at the bait. These same types would be happy to buy and sell actual slaves if it was legal. They imagine AI as something they can own, a slave.

At the moment the greatest threat posed by AI is all the money that's going to evaporate once its limitations become apparent.

short version: TSF admin booted Anthropic AI because they insisted on certain red lines in their TOS. OpenAI jumped at the chance for all those gov contracts. You might want to ask what Anthropic's red lines were. Answers in the videos below. Summary: mass surveillance of Americans (think of all those gmail accounts that google is happily scraping, plus plenty of others); AI running military attacks without human oversight. (Think of the pace of AI change ... who we have running things in this country ... and the mistakes you have seen from AI if you've ever tried asking it a few things ... feeling comfy?)

(I should make this into an OP, if it hasn't already been covered on DU)

Note that the headline of the first is completely incorrect ... it was sort of true for about 5 minutes ... but it is shorter and the video itself seems to be an okay summary of the situation prior to OpenAI stepping into the breach to give TSF's administration what it wanted.

I don't watch news and opinion videos, ever.

This current round of AI grift and hype ( there have been several previously ) will no doubt end in catastrophe but it doesn't have to be military.

Some have suggested, even here on DU, that AI would be great for managing energy grids, especially as it relates to the integration of renewable energy into these grids. That has a huge potential for catastrophe, especially if instabilities in the grid escalate to the point of destroying infrastructure that takes years to replace.

Imagine all the damage some vibe coding idiot could do to our financial sector...

Could I trick Claude into writing some virus that takes down every Windows 11 machine on the internet?

Etc.

she seems to conflate Large Language Models with all of AI research. She then "takes down" LLMs, and by association AI proper, by saying they will never be anything but stochastic parrots. It only takes a cursory search to learn that there is more going on in AI research than LLMs and chatbots.

Sorry, but I'll take the opinions of non-anonymous experts with transparent credentials over some rando.

.... in my work environment. I don't work on it directly, but, since this is a research institution, we conduct seminars and training, and I've been following through on that. Which is one reason I've been interested in this topic on line. From my training and limited hands on experience her article is accurate.

You are right that LLMs are not all of AI. They are the AI that the public sees and interacts with and shapes their opinions of it. People in general are not actually involved in, say, the investigation of protein folding, on their own or read much about it, or understand its importance. But they do interact with chat bots and ask Google questions on the web. So that is where they get their impressions, and their fears.

And there seems to be a lot of folks who are quite willing to exploit those fears. At work, one AI researcher I know calls one outlet, The Futurist , "moral panic as a service." Understanding the actual inner workings of the AI that most people interact with can be very helpful in keeping our feet on the ground and combating overreaction, moral panic, and FUD. I did not see this article as a "take down." I saw it as an explanation and an attempt to spread understanding.

My AI experience and training, such as it is, indicates to me that all successful AI begins with, and continues with, the training of some sort of statistical model on a large collection of data. A large statistical model is at the core of all successful AI that I am aware of.

Are you aware of any different approach?

In the first few paras. Evil Annie...proofread your screed before you 'publish' it.

... on the Kos post.

Why do you characterize her post as a screed?

..... are part of the problem with AI chatbots and their public perception. They are trained on all of this scifi dreck, and then they mechanistically spit it out when prompted.

However entertaining, war games, terminator, space odyssey, etc. are fiction, not reality, and are driving misunderstanding and unnecessary fear.