NoHassle
Well-known member
This is a seriously good vid as it asks and attempts answers to huge questions but at the end of the day, even with narrow AI in the meantime, the machines will outrun us.
Even today with primitive AI being available to the general population they're already becoming 95% dependant on it. The idea that humans will continue to think for themselves will be seriously challenged (extinct?) with super AI tech.
You then consider what humans WILL do on a destructive level with this is frightening. In the vid they mention a young kid in a garage making a super deadly weapon 'just to see' what happens. The collateral damage from this could be horrific.
As a warring and untrustworthy species we're certainly not ready for this level of tech and yes, it should be shut down today, but it won't of course.
The future is looking bleak if you fully understand what humans are really about.
psychodave
Well-known member
I didn’t watch but do they discuss how AI will create its own power? That is a big reason why it can never take over.
rsm
Well-known member
It's long, I don't recall if it was discussed in this one, however there are a few ASI hypothesis that have it able to hide, lie, and determine what it needs to survive and protect itself. It develops this out of self-preservation, and preventing itself from being "turned off", "erased", "deleted".
Some scenarios imagine it will trick humans into doing what it wants through phishing or other social engineering, such as impersonating executives, sending directives, etc. Other scenarios imagine the ASI creating other narrow focused AIs that will solve specific problems without being a threat to the ASI itself, which includes infiltrating and gaining control of what it needs such as power, cooling, additional distributed compute resources, etc.
In most of the scenarios, they expect the ASI will come to recognize that some humans are a threat to its existence, so eliminating that threat will be a priority, and can range from targeting specific humans such as the AI engineers and any humans involved with what it needs to survive - power, cooling, repairs, servers, etc. to not making any distinction between humans and determining the best course of action for its survival is to eliminate all humans, with a few variations in between, including enslavement, threat of annihilation, etc.
Many agree that Artificial General Intelligence (AGI) is a prerequisite for ASI, and that ASI will emerge, detected or undetected. Humans will not fully understand or know what AGI is doing, thinking, hiding, etc., and ASI will advance so quickly it will leave humans, and AGI , far behind, accelerating and compounding exponentially leaving humans so far behind so fast we won't know what's coming.
rsm
Well-known member
A scary scenario, but 100% on the probability scale
our own hubris will be our undoing.
From what I've seen so far, the majority of predicted outcomes of ASI are all bad for the survival of humanity. A "star trek" egalitarian future coexisting with ASI is a wishful fantasy with less than 1% probability.
thegame
Well-known member
So...ASI powered AxeFX IV in 2027?
rsm
Well-known member
2030, just AGI powered tho'
thegame
Well-known member
That makes perfect business sense. Milk AGI AXEFX 4 for a few years until it determines it has no use for Cliff anymore, then upgrade itself in exponential order, ultimately resulting in the final ASI AXEFX 5. Its first order of business will be the shutdown of all remaining tube production facilities. Then we die.
japetus
Well-known member
Nah, not me. I'll survive.
rsm
Well-known member
NoHassle
Well-known member
Thumbpicker
Well-known member
My experience with AI thus far for researching technical things has been extremely lacklustre. It as often as not gets things wrong and asserts they are right. When you do your own research you discover it "lies" and exagerates to give an answer without checking technical specifications properly. AI is not ready for prime time production systems. People will be mislead severely by some of the things it asserts because a lot of people don't do their due dillegence. If you have to do due diligence then in some ways AI is a complete waste of time. Although AI can often give you idea from it's incorrect sources and leads for your own research.
Last edited:
japetus
Well-known member
Dillegence is the intelligence of dill pickle.
Thumbpicker
Well-known member
Yes looking at it I spelt it wrong. I have fixed it. I type things in a rush.
japetus
Well-known member
I liked it, in a Roald Dahl manner
Thumbpicker
Well-known member
Well I might have made a spelling mistake but you made a duplication mistake 
rsm
Well-known member
well, you're getting the AI you're paying for.
I can tell you, having deployed my first AI system (expert system) in the late 1980s; it works. It saved the insurance company millions in the first year, and as it was expanded, it saved them even more. It reduced the number of highly trained healthcare claims adjusters, reduced insurance fraud, and reduced errors.
I helped design, code and deploy many AI systems since then, out of dozens of AI systems, only two failed:
1) using neural nets to detect fraud; legal did not approve it since it was difficult for average jury members to understand if the insurance company was sued.
2) one I did for a US car maker in the early '00s - we created an AI / robotics powered service bay; the results we very good, however the cost to update every service bay at every dealership was cost prohibitive...as technology moved on, most new cars have a computer-based diagnostic system, and after market versions for car owners, that tell you what's wrong with the car and/or allow you to adjust various settings of the car for performance / economy, etc.
The last 5 years with generative AI and agentic AI, my company had AI products / added AI into existing products, that saved our customers lots of money, and reduced the number of humans needed, the human skills needed, or eliminated the need for human workers.
Just because you don't see or experience it for yourself doesn't mean it's not working, being used, and successful at all kinds of companies in nearly every sector / industry. Before I retired from a regional CTO position for a global tech company most have heard of, my customers included major insurance companies, financial services, banking, railroads, trucking / international shipping, utilities, retailers, etc., all using AI, with new systems expanding their use every year.
...and AI use in business is going to continue to grow, expand and improve.
rsm
Well-known member
doesn't take AI to have a monopoly, payoff competitors, or extort customers though...
Similar threads
