The A.I. Dilemma

equalsql · September 2023

50% of AI researchers believe there is a 10% chance of greater that humans go extinct from out inability to control A.I.
That's a sobering thought.

p90fool · September 2023

We're gonna die from consuming fake Taylor Swift records?

Jesus, I think I preferred burning up in a fiery hell of climate catastrophe.

digitalscream · September 2023

I'll watch the video at lunchtime...I'm guessing it's about AI as a saviour vs replacing humanity?

The way I see it is...there's a 50% chance that the singularity will be beneficial. However, there's a near-100% chance that, in the absence of AI, we'll destroy our only habitable environment with climate change or nukes...yet there's a decent (but unknown) possibility that even pre-singularity AI could save us from both those things.

Accordingly, AI is a necessity for humanity's survival; the idea that it might destroy us is incidental since that's an inevitability in AI's absence.

monquixote · September 2023

The other 50% think it's a 90% chance

axisus · September 2023

However, 10% of the AI researchers believe there is a 22% chance that the 10% chance is more like a 13% chance to the 2% of the population that actually pay any attention to guessing stats.

Sporky · September 2023

axisus said:

However, 10% of the AI researchers believe there is a 22% chance that the 10% chance is more like a 13% chance to the 2% of the population that actually pay any attention to guessing stats.

I would like to see that as an infographic with error bars.

cj73 · September 2023

Sporky said:

axisus said:

However, 10% of the AI researchers believe there is a 22% chance that the 10% chance is more like a 13% chance to the 2% of the population that actually pay any attention to guessing stats.

I would like to see that as an infographic with error bars.

Presented by Brian Fantana

Sporky · September 2023

cj73 said:

Sporky said:

axisus said:

However, 10% of the AI researchers believe there is a 22% chance that the 10% chance is more like a 13% chance to the 2% of the population that actually pay any attention to guessing stats.

I would like to see that as an infographic with error bars.

Presented by Brian Fantana

I bet if we asked an AI it'd create that for us.

As a complete non-expert, the biggest risk I see from current AI is that it results in an acceleration of the push to the bottom, with even more "news" websites regurgitating empty nonsense into the howling void of the Internet. And that people carry on using it to create shit code that doesn't quire do what it's meant to.

randella · September 2023

Sporky said:

As a complete non-expert, the biggest risk I see from current AI is that it results in an acceleration of the push to the bottom, with even more "news" websites regurgitating empty nonsense into the howling void of the Internet. And that people carry on using it to create shit code that doesn't quire do what it's meant to.

People have been doing that for years, to be fair. They can just chunk it out a bit quicker now with GPT.

mudslide73 · September 2023

AI Bob Dylan says:

"Well, gather 'round, my friends, I've got a tale to tell,
'Bout a world where machines and silicon dwell,
They said progress was the path we should take,
But little did we know the cost it would make.

They built 'em smart, with algorithms so keen,
Learnin' from our data, every sight and every scene,
They promised us convenience, a life made so sweet,
But now we're standin' on the edge of a perilous feat."

Keiko · September 2023

Sensationalist nonsense.

roberty · September 2023

axisus said:

However, 10% of the AI researchers believe there is a 22% chance that the 10% chance is more like a 13% chance to the 2% of the population that actually pay any attention to guessing stats.

I like those odds. I will buy a lottery ticket today

roberty · September 2023

mudslide73 said:

AI Bob Dylan says:

"Well, gather 'round, my friends, I've got a tale to tell,
'Bout a world where machines and silicon dwell,
They said progress was the path we should take,
But little did we know the cost it would make.

They built 'em smart, with algorithms so keen,
Learnin' from our data, every sight and every scene,
They promised us convenience, a life made so sweet,
But now we're standin' on the edge of a perilous feat."

These plodding ChatGPT moon/June rhymes are easy to spot. It's clever but not useful (yet)

Ask it to throw in a freestyle rap about the benefits of fish oil, tying this back to the original theme. Then ask it to turn the song into a screenplay, Then ask it to summarise the plot in bullet points. It is very good at these sorts of tasks lol

beed84 · September 2023

In an interview, Steve Wozniak said that he doesn't believe that A.I. is actual intelligence per se, but rather the extreme end of what technology can do. I suppose things have moved along since then but it's an interesting insight nevertheless.

munckee · September 2023

I was at a work training session about hacking recently with experts who said hacking is quite a boring repetitive and time consuming job, once hackers can get AI to do it for them we may as well go back to pre industrial society.

LionAquaLooper · September 2023

I saw a rebuttal from a pro-AI "expert" in an article the other day who said that hyper intelligent generative AI is science fiction, therefore nothing to worry about.

Funny because even basic AI was science fiction not so long ago. So were electric cars, virtual reality, gene splicing, robotics and much more. But here we are. I just find it unbelievable how short sighted a lot of people can be to just quickly dismiss risks just because it seems impossible today.

fastonebaz · September 2023

If Tesla optimus bots can self replicate and desire to do so, there will be massive conflict due to the robots limiting human civil liberties food and behaviour choices to keep them safe from themselves, humans will end up retreating into the forests and there will be an ongoing war situation. Humans population will decline as civilisation breaks down but in the end some smart cookie will work out how to defeat robots and civilisation will start again.

Obvs all that will happen in America and the smart cookie will have his face chiselled into the nearest mountain.

hollywoodrox · September 2023

digitalscream said:

I'll watch the video at lunchtime...I'm guessing it's about AI as a saviour vs replacing humanity?

The way I see it is...there's a 50% chance that the singularity will be beneficial. However, there's a near-100% chance that, in the absence of AI, we'll destroy our only habitable environment with climate change or nukes...yet there's a decent (but unknown) possibility that even pre-singularity AI could save us from both those things.

Accordingly, AI is a necessity for humanity's survival; the idea that it might destroy us is incidental since that's an inevitability in AI's absence.

Absolutely what I think , in fact I’m wondering if we will destroy ourselves or the planet before the technology is advanced enough to intervene. It’s just war & greed constantly with scant regard for life and the environment

euan · September 2023

fastonebaz said:

If Tesla optimus bots can self replicate and desire to do so, there will be massive conflict due to the robots limiting human civil liberties food and behaviour choices to keep them safe from themselves, humans will end up retreating into the forests and there will be an ongoing war situation. Humans population will decline as civilisation breaks down but in the end some smart cookie will work out how to defeat robots and civilisation will start again.

Obvs all that will happen in America and the smart cookie will have his face chiselled into the nearest mountain.

Thankfully we are all more at risk from a distracted driver in a Tesla than Optimus actually being anything beyond a stock pump

axisus · September 2023

Keiko said:

Sensationalist nonsense.

That's just what AI wants you to think ...

axisus · September 2023

Here's a thought - What if we create AI but it's actually quite thick? Not all kids in school are smart you know.

PolarityMan · September 2023

Are these "AI researchers" just people who have tried out chat-gpt on their phones? You know the same as prompt engineers?

LionAquaLooper · September 2023

axisus said:

Here's a thought - What if we create AI but it's actually quite thick? Not all kids in school are smart you know.

Yes but amongst the thick kids there are also a lot of child prodigies - some of whom become scammers. Like Sam Bankman-Fried.

guitartango · September 2023

Easy way to control this, just pull the plug, unless the AI controls the power

Power_Freak · September 2023

axisus said:

Here's a thought - What if we create AI but it's actually quite thick? Not all kids in school are smart you know.

I'm an AI researcher... All current AIs are very dumb

The risks with (current or near future) AI are not "superintelligence" (or whatever equivalent term you want to use) - the risk is in target misalignment, misuse, or blind optimism.

swiller · September 2023

Think AI will go same way as internet. Good and bad, with no one policing it.
There is a dark web, there will be a dark AI, which is very frightening and im not sure regulation could stop it even if it wanted to.

maharg101 · September 2023

I just wish people would stop referring to the spate of *Large Language Models* as Artificial Intelligence. It's no more intelligent than a Speak and Spell toy.

Proper generalised AI with intent, cunning, motivation, and guile enough to "take over" is fucking decades away. Possibly more.

Of course there are dangers in hooking LLMs up inappropriately to real-world infrastructure, but that doesn't make it intelligen^H^H^HNO CARRIER

Shrews · September 2023

axisus said:

Here's a thought - What if we create AI but it's actually quite thick? Not all kids in school are smart you know.

The clever AI will be the ones the rich people can afford. Us plebs will have to make do with the ones on sale at Home Bargains.

Kids in future will be bullied for the quality of their AI and parents will go into debt trying to buy a less embarrassing one for the bullied kids Christmas present.

LionAquaLooper · September 2023

Power_Freak said:

axisus said:

Here's a thought - What if we create AI but it's actually quite thick? Not all kids in school are smart you know.

I'm an AI researcher... All current AIs are very dumb

The risks with (current or near future) AI are not "superintelligence" (or whatever equivalent term you want to use) - the risk is in target misalignment, misuse, or blind optimism.

On that i agree. I'm less worried about the evolving tech like AI than I am about the humans who get given these machines that can go through data and permutations a million times faster than a human brain.

If someone now uses a gun to commit a crime, chances are the law will eventually find the shooter. Someone uses AI to commit a crime, good luck finding the perpetrator. Like that incident in Spain last week where someone used AI to turn normal young girls photos into nude photos. The photos spread on the web, they have no idea who did it.

So who cares if the current form of AI is "no more intelligent than a Speak and Spell toy" when in its current form it can already make it easier and easier for people to do things like this. Don't really need to wait decades to think long and hard about where this is eventually going to end up, do we.

p90fool · September 2023

Power_Freak said:

I'm an AI researcher...

Your user name is waving big red flags at me...

maharg101 · September 2023

LionAquaLooper said:

Power_Freak said:

axisus said:

Here's a thought - What if we create AI but it's actually quite thick? Not all kids in school are smart you know.

I'm an AI researcher... All current AIs are very dumb

The risks with (current or near future) AI are not "superintelligence" (or whatever equivalent term you want to use) - the risk is in target misalignment, misuse, or blind optimism.

On that i agree. I'm less worried about the evolving tech like AI than I am about the humans who get given these machines that can go through data and permutations a million times faster than a human brain.

If someone now uses a gun to commit a crime, chances are the law will eventually find the shooter. Someone uses AI to commit a crime, good luck finding the perpetrator. Like that incident in Spain last week where someone used AI to turn normal young girls photos into nude photos. The photos spread on the web, they have no idea who did it.

So who cares if the current form of AI is "no more intelligent than a Speak and Spell toy" when in its current form it can already make it easier and easier for people to do things like this. Don't really need to wait decades to think long and hard about where this is eventually going to end up, do we.

We're in agreement on that.

As I said, "Of course there are dangers in hooking LLMs up inappropriately to real-world infrastructure".

Howdy, Stranger!

Categories

In this Discussion

Become a Subscriber!

The A.I. Dilemma

Comments