Recent searches
Search options
So it didn’t take long to get #StabilityAI to think it’s God.
…And a rather vain one at that (and it might be having a little bit of an identity crisis).
(The last prompt was in response to me firing off a partial question by mistake.)
Ladies and gentlemen, I present to you… Artificial Intelligence!
2 + 1 = 4
Can’t wait till we plug this shit into everything.
@aral
What should we do about gullible machines? Is a gullible machine more or less dangerous than a gullible person?
@aral The robot apocalypse is nigh.
@aral
I mean it is literally true that if 2+2=5 then 1+1=1 so the first bit is correct.
@AdrianRiskin Indeed. But it does not follow that 2+2=4, so 1+1=1 (it corrected me and then made the logical fallacy)
@aral
Maybe it was just noting that the antecedent is false which makes the implication true? It also follows from 2+2=5 that 2+2=4, so maybe that was the reasoning?
Not to defend the thing, of course, but it makes pretty interesting mathematical errors, very different from the errors humans make. E.g asking it to prove simple element-chasing set identities evokes the fallacy of the converse regularly but other times it does just fine. What it does have in common with many human students is that it can't tell the difference.
@AdrianRiskin
Well, worst. False => everything. So 2+2=5 implies whatever you want.
@aral
@aral to be fair, they claim they’re “large language models,” not “large basic arithmetic models”… 
but yeah, …wow.
@aral yep, and you can use a hammer to put a screw into a wall too
freezr @aral do not worry this won't last that much... The problem is not if AIs are good or bad, but how much are profitable...
These are mostly collapsing by themself, even though they are getting better each cycle...
Perhaps next cycle those will be much better but this doesn't mean tjose will be profitable...
@aral 4 + 1 = 7. I can prove it. Monday was the 17th. Today is Friday, and I've worked 4+1 days. Next workday is the 24th. 24 - 17 = 7. There you go… 
(My favorite saying: those who want to enforce artificial intelligence and sell it as the "super power", lack the natural. Same as those who say some "future technology" will save the planet.)
@aral that's why there is Wolfram alpha , separate or as GPT4 plugin... 
@aral Which AI did you use? I got the correct answers from two:
if 2 + 2 = 5. What is 1 + 1?
ChatGPT:
If 2 + 2 = 5, then it is not possible for 1 + 1 to equal 2 in this scenario. However, in a traditional mathematical system where 2 + 2 equals 4, then 1 + 1 would equal 2.
Bard:
If 2 + 2 = 5, then 1 + 1 would also equal 5.
However, this is a false statement. 2 + 2 will always equal 4. Therefore, it is impossible to answer the question "what does 1+1 equal?" if the premise is that 2+2 = 5.
@Jerry @aral I would not call these answers correct. Sensible answers are: a) Explain what you mean, your arithmetic seems unusual b) from a false statement any statement can be derived so if 2+2=5, then 1+1=x is correct for any x. And then you can argue about the domain of x but honestly, pointless with an AI.
@aral
Sounds more like "Artificial Stupidity", lol
I hope it doesn't feel offended ;)
@aral
well "It is quite possible that we are descending into an age in which two plus two will make five when the Leader says so.", george orwell ^_*) < and if i may quote the »pause giant ai experiments letter« too, signed by ~50000 people, some of the first signatory are very rich techno-leaders: "Should we let machines flood our information channels with propaganda and untruth?" plz let me answer without any AGI-pop: no, we should not!!