The Futurism site is mostly hokey, but this is kinda interesting:
formatting link
"To spit out a five-second clip, the researchers found that it takes the equivalent of running a microwave for over an hour."
Imagine a sort of DDOS attack on an AI site that deliberately burns power.
Maybe it could be viral, as in making various AI sites bomb one another.
There might be one magical question that spins off gigawatt-hours of energy.
The only defense would be to require secure payment for energy used, and even that could be hacked.
Didn't find your answer? Ask the community — no account required.
M
Martin Rid
I'm sure they watched, A hitchhiker guide to the galaxy.
Cheers
E
Edward Rawde
That made me wonder what current AI would make of a paragraph which might be nonsense but can you be really sure about that?
So I enjoyed myself by typing the following paragraph.
When ending the roads should be on time for everything has meaning and we can just throw out the bad parts. So I would like to know why there is no end to this and what we can do to strengthen the reasoning. I will start with more detail when I have arranged the content as necessary for reasoning.
And I pasted it into Grok.
Now I know why AI is going to need Gigawatts, or is that Jigawatts?
I'd have been way more impressed with a more Sloman-like one line response to this, such as "You have no clue what you are talking about."
E
Edward Rawde
If you want to really confuse Grok, try this:
Un petit d'un petit s'etonne aux halles Un petit d'un petit a degre te falle
If you want to be even more ridiculous try this:
Center Alley worse jester pore ladle gull how lift wetter stop-murder an toe heft-cisterns. Daze worming war furry wicket an shellfish parsons, spatially dole stop-murder, hoe dint lack Center Alley an, infect, word orphan traitor pore gull mar lichen ammonol dinner hormone bang.
Grok does get some of it right.
Hard to say what keeps it interested but it clearly spends longer "thinking" about any problem which produces no search engine results at all. Unlike the two examples above.
L
Liz Tuddenham
[...]
Wonderful! Have you ever considered becoming a politician?
E
Edward Rawde
Actually no. I prefer facts. Coming up with nonsense takes too much effort. So maybe I'm a reverse politician.
Back when I had to study the following, I also remember the maths teacher showing us how to prove the quadratic formula.
formatting link
can still prove the quadratic formula (as can most people here) but I can only remember the first few words of that poem.
E
Edward Rawde
Yeah I've seen that one too.
Looks like Grok had online help with that to me.
E
Edward Rawde
Now I can't stop my mind making me hear (in my mind) the president of a large country saying this. Maybe someone out there has an AI system which can make a video clip such that I don't need to hear it in my mind.
E
Edward Rawde
It's clearly true that the kind of chat based AI currently found online does not come anywhere close to thinking for itself. It does not even learn from its mistakes.
T
Tom Del Rosso
Was "I will start with more detail when I have arranged the content as necessary for reasoning." the response, or part of what you asked it?
E
Edward Rawde
Just part of the nonsense I fed into it.
We all know Garbage in Garbage out. Current LLMs seem to be a case of Garbage in and hope something useful comes out.
T
Tom Del Rosso
Well then aren't you going to say what its response was?
E
Edward Rawde
The response given by an LLM AI Chatbot, when fed the following:
[Start text] When ending the roads should be on time for everything has meaning and we can just throw out the bad parts. So I would like to know why there is no end to this and what we can do to strengthen the reasoning. I will start with more detail when I have arranged the content as necessary for reasoning. [End text]
appears to be different each time.
I do not record all such interactions so I can't tell you exactly what its response was. From memory it took maybe 30 seconds to "ponder" my request and come up with a response.
So if you want to see one possible response please paste the above text into
formatting link
I tried again a few minutes ago and it "thought" for 22 seconds before producing word salad which was debatably as good as the input.
The things I find interesting about the response to this text are:
It spends some time searching the web for my text but has to concede that a web search isn't going to help.
It comes up with biblical references (Ecclesiastes 3) in response to nonsense.
So I decided to clear my mind and exercise my word salad generator once more:
[Begin word salad] If the words of many are expected to find time and reasons then why can't we deduce the complexity of forever by using our skills to manipulate the characteristics of the associated events? Is it because we are not centred on the structure of the problem? Or could it be that there are infinitely possible solutions? [End word salad]
That seems to really short circuit it. You can almost see the smoke rising.
Interesting that the word "mathematics" turns up in the answer I got. I kinda thought that "infinitely possible solutions" might provoke that.
E
Edward Rawde
....
After trying that again 12 hours later I suspect that it keeps a cache of recent web searches so it doesn't have to admit to using a web search to assist with the reply.
E
Edward Rawde
This gets a very fast and short response. I didn't have the heart to reply with "Actually I just wanted to see your response to a load of nonsense."
On the timeless raze by the rocks of laze There's a firmless haze with a dappled blaze And the fearless maze is ablaze with taze Like the sexless phase of an endless gaze.
Join the Discussion
Have something to add? Share your thoughts — no account required.
Didn't find your answer?
Ask the community — no account required
Report Content
You are reporting this content to the moderators. They will look at it
ASAP.