AI: Major Emerging Existential Threat To Humanity

[ryan]

Like code preventing certain topics and fail safes that prevent it from taking over the world

[Etern4l]

They will probably try, but

1. These algorithms are not programmed in the traditional way, so implementing those failsafes is difficult

2. What if there is a problem with one of the failsafes?

3. What if one of the sailsafes is disabled by a malicious, manipulated, or dumb actor?

3. What if an AI figures out how to bypass one of the failsafes?

[ryan]

True. Super interesting topic. Hopefully others chime in for their two cents

[Etern4l]

Not an existential threat to the whole of humanity, but possibly to artists:

 

Sony World Photography Award 2023: Winner refuses award after revealing AI creation

 

Death by a thousand cuts.

[serpro69]

19 hours ago, Etern4l said:

Not an existential threat to the whole of humanity, but possibly to artists:

 

Sony World Photography Award 2023: Winner refuses award after revealing AI creation

 

Death by a thousand cuts.

 

Love the quote "AI is not photography". That's the feeling I always have then I see people saying "AI can create art" or smth like that. It's not art for f***s sake... 

[Etern4l]

4 hours ago, serpro69 said:

 

Love the quote "AI is not photography". That's the feeling I always have then I see people saying "AI can create art" or smth like that. It's not art for f***s sake... 

 

The problem is that it can be perceived as art, it's basically indistinguishable from it. If this guy didn't say anything, he would have just won the award. Luckily, he was human enough to take a stand. 

 

If someone needs a bunch of graphics for their website, they will just use midjourney, dall-e etc and be done with it in a matter of minutes. This will kill most artists, graphics designers etc off. 

 

Edit: There is an Elon interview with Tucker Carlson. Generally a harsh critique of Google trying to build a super-AI. Then kind of struggles to justify the creation of his own AI company. Also pretty poor responses regarding the risks of AI (although definitely deserves credit for pointing some out). Talks about the need for regulation, but then his idea of regulation is weak sauce. Tucker Carlson downlplaying the economic risks of AI big time "This thing is cool, can write college papers haha". Idiot. Not worth linking to overall. 

[serpro69]

14 hours ago, Etern4l said:

The problem is that it can be perceived as art, it's basically indistinguishable from it. If this guy didn't say anything, he would have just won the award. Luckily, he was human enough to take a stand. 

 

If someone needs a bunch of graphics for their website, they will just use midjourney, dall-e etc and be done with it in a matter of minutes. This will kill most artists, graphics designers etc off. 

 

Yeah, no, I totally agree with this. I'm just a little annoyed when people call AI-generated stuff "art" is all :D

 

Can't help but feel sorry for artists/designers/writers. They will be among the first ones to be affected by this bs. And the irony is that these models were trained on creator's own works. What a mess...

[Etern4l]

45 minutes ago, serpro69 said:

 

Yeah, no, I totally agree with this. I'm just a little annoyed when people call AI-generated stuff "art" is all :D

 

Can't help but feel sorry for artists/designers/writers. They will be among the first ones to be affected by this bs. And the irony is that these models were trained on creator's own works. What a mess...

 

Start weeping for programmers who are next in line. Here too the models have been trained on their own work. 

Then lawyers. Sure, nobody really likes them, but there are undoubtedly some good jobs in the legal profession that will be gone. Yes, you will still need one to take things to courts, but dealing with a lot of smaller issues and advice will be automated. 

 

Then folks like accountants, analysts and scientists. Basically any job which is not significantly physical (those will be a bit safer for a while at least, with the exception of farming and driving). 

 

Let's not forget about actors, and film industry professionals, if AI gets to a point where it can generate movies, which are just a collection of individual frames AI can generate extremely well. 

 

There is a deeply biased report by OpenAI where they claim "only" about 50% of the jobs will be affected, sad lol. 

 

The kind of jobs I can see being unaffected for now are personal services, such as hairdressing etc.

 

The US university system will no doubt collapse. A year at Cornell already costs $90,000...  There won't be enough jobs for people to be able to justify spending the time, and paying the exorbitant fees. 

 

Capitalism will follow shortly, OpenAI say as much themselves. We can't all be hairdressers, the oldest profession in the world will be overcrowded as well and dime a dozen (being a connaiseur billionaire like the former President will get even better). We can't all be builders, and anyway construction tends to die first during economic downturns. Wages in the "safe" low-skill professions will contract further (already an automation-related trend), due to the pressure from all the newly unemployed trying to make a living. Joy to the world! 

[ryan]

50 percent seems like a mighty high estimate considering most jobs won't be affected. Think doctors nurses tradesmen like electricians plumbers sheetmetal oil and gas workers. Framers drywall ect so many jobs are unaffected like 18 percent seems more accurate. Or 300 million jobs

[Etern4l]

Sorry, what time frame? This year only? Agreed, 2%. That's basically unemployment in the US going up by 50% from where it is today, in some of the highest paying jobs so there will be a snowball effect soon after. Even if those people start flipping burgers for a minimum wage, that's effectively unemployment. 

 

Farming is one of the most heavily targetted industries, in 5 years time you are in for a surprise there. Construction? Hold on to your toolbox buddy... 

 

 

[ryan]

A quick Google search shows Sachs estimating 18 percent. That's all.  I think even 18 percent is a bit on the hyperbole side of things

[Etern4l]

And once again, what time frame? Have a link? Sachs who? Crooked Goldman?  Could Google results be biased BTW? 

 

Anyway, +18% unemployment would be tantamount to economic cataclysm, made worse by the fact that initially a lot of the jobs lost will be high paying (huge ripple effect on the local economy). There is no comparable situation in the past, all the estimates are biased best case scenarios. Unemployment doesn't tell you how poor/wealthy the employed are, and there will be huge downward change there as well, pretty much across the board. 

 

It must be easy to think "oh, I'm a tradesman etc so I don't care", that's just a lack of fundamental economic awareness. Economy is interconnected. Once you start losing clients and getting undercut by all the new tradesmen entering the field out of neceessity you will care, and at that point it will be too late. Unlike the Great Depression, the changes will be permanent and getting ever worse. 

 

Probably the most interesting, the least weak so to speak, bit of the whole Musk-Carlson interview:

 

 

Selling himself as a "speciest", while still watering things down there quite a bit, for example "AI cannot be distributed/will be in the data cetnre" - at the moment it is, but sure it can and likely will be distributed in the future, also steering well clear of the topic of economic consequences. Focusing on the Twitter-related aspects, and not at all mentioning his own humanoid robot and the AI efforts at Tesla... 

 

Carlson very happy to be just a chummy cheerleader a opposed to being a proper critically-thinking journalist. 

 

 

Elsewhere in the interview Musk basically admitted he is a "specieist", therefore planning to create AI slaves... What could go wrong...

[Reciever]

Given the sensitive nature of the topic at hand please keep it within scope. Obviously many potential rabbit holes can come up due to this topic so I ask that all involved make a concerted effort to move the discussion constructively as possible. 

 

Thanks everyone! 

[Etern4l]

GPT-4: What is it and how does it work?

It can accept visual prompts to generate text and, interestingly, scores a lot higher in the uniform bar exam than GPT-3.5 in ChatGPT did (it scored in the top 10% of test makers, while GPT-3.5 scored in the bottom 10%). It's also more creative and has a larger focus on safety and preventing misinformation spread.

 

 

There are other performance improvements as per the company's research, including improvements in Leetcode, AP level class, and SAT results.

 

So, let's unpack:

* More creative, although that's hard to quantify - still, an ongoing full frontal attack on what was previously hoped would be a fairly unique feature of HI (Human Intelligence), besides the Holy Grail of AI: self-awareness

* Even more advanced content filtering according to OpenAI's idea of what's information and what's misinformation.

* Vastly improved performance on the bar exam and LSAT scores, humans need not apply - a huge step towards making lawyers, and other legal professionals redundant

* Significantly improved SAT Math scores, but struggling with introductory-level college stuff, fortunately for mathematicians/engineers (not for very long would be my guess)

* A large bump in USABO / medical knowledge exams (medical and biotech professionals/students: start getting ready!)

* Codeforces rating upgraded but still really poor - seems like top programmers are kind of further down the line from the AI chopping block (but the sociopaths at OpenAI are doing what they can to change that)

* Doing well with college level Art History and Biology

 

All this progress in less than a year from the introduction of GPT 3.5.... 

 

For the last century or so, people have been given a social mobility route through their intelligence, hard work and education. Soon AI will not only cut this off for most, but also destroy the livelihoods of those who have historically managed to leverage that opportunity. 

[ryan]

Where do I get chat gpt 4 on android . Little update I found chat gpt 4 for android but you have to pay. Wth why would I pay

 

Also how are people making ai artwork and videos. Where do I go to play with ai

[serpro69]

On 4/22/2023 at 3:12 PM, ryan said:

Also how are people making ai artwork and videos. Where do I go to play with ai

 

I have only tried a few silly prompts with DALEE, which you can play with for free by registering an openai account (same account as you'd use for ChatGPT, for example, if you already have that). But you will have limited number of prompts per day, unless something has changed, and anything beyond basic stuff is paid I think.

The more advanced stuff like Midjourney is paid AFAIK, so you'll have to pay to be able to play.

[Etern4l]

Max Tegmark: The Case for Halting AI Development | Lex Fridman Podcast #371

Only managed to watch in part so far. Generally still very biased, understandable given that AI research is how he makes his living. For example seems to trivialise the economic impact of AI, and claims that AI alignment "seems solvable but very difficult". Well, we don't know if it is solvable in theory, and even if it is - whether we manage to do it before it's too late.  Came up with a quotable statement though:

"We're rushing towards a cliff, but the closer we get, the more scenic the views are."

 

Overall, a bit weak, but did raise a good few AI red flags, which is obviously much better than the usual intellectual nothing (Yay Hooray AI).

 

BTW He is one of the people who defended the engineer sacked by Google for claiming its AI seems sentient:

 

MIT professor warns Amazon's virtual assistant will be 'dangerous' if it learns how to manipulate users - as he defends Google engineer who said its AI program has feelings

 

Meanwhile EU Lawmakers Call For Summit To Control ‘Very Powerful’ AI

 

A parliamentary committee is debating the 108-page bill and hoping to reach a common position by April 26, according to two sources familiar with the matter.

[Etern4l]

Here is another, more concise, piece with Tegmark. He is still making a mistake by proposing that regulation and control are the solution. They are not, we provably cannot control AGI, however, he does a good job explaining the terminal risks involved. Enjoy!

 

 

[kojack]

Yep.  Can't wait to go to the ole Pepperoni hug spot.  CRINGE!

 

AI has A LONG LONG WAY to go.

 

Pepperoni Hug Spot

[Etern4l]

On 4/26/2023 at 11:12 AM, kojack said:

Yep.  Can't wait to go to the ole Pepperoni hug spot.  CRINGE!

 

AI has A LONG LONG WAY to go.

 

Pepperoni Hug Spot

 

Nice comforting strawman, hope you have found it helpful. Have you seen this though:

 

Sony World Photography Award 2023: Winner refuses award after revealing AI creation

 

The thing to consider here is the pace of progress. Where was this 10 (what a preposterous idea lol), 5 (nowhere really), 1 (wait, something's going on) years ago, where is it now, and where will it be in 1,, 2, 5 and 10 years. The following illustrates this exponential progress issue and our flawed and biased (by hope) perception of it nicely:

 

 

(A relevant slide showing the chart in the correct scale, and a human startled by AI progress at T+delta deleted by site admin)

 

Normally not a fan of those kind of mixes, but looks to the point and perhaps that kind of content is what's needed to get the point across:

 

[kojack]

I love the "sky is falling" mentality in this place. It's quite amusing and humorous. Always good entertainment...  

 

you might as well sell everything off move to a cabin in the woods.  Your mental health would be extremely better off. 

[Etern4l]

You're missing the point, unfortunately. The sky isn't exactly falling right now, but AI is already causing some serious issues, and it will likely get exponentially worse in the relatively near future. Interestingly, no comment on the Sony World Photography Award 2023 topic...

 

Here is another (super-benign in the grand scheme of things) aspect of the big tech AI abuse:

 

 

We could all just sit in an armchair, grab a bottle or smoke some reefer, and wait, or try to do something about this.

Moving to a cabin in the woods is not really necessary yet, and wouldn't actually accomplish much if we eventually lose control of the planet. It's not like you are going to be able to wait an immortal entity out. I guess the idea would be to stay under the radar, like a monkey in the jungle, or a cockroach, in some part of the planet the AI is less likely to claim and transform for its own purposes. Animals (the lucky free ones) do precisely this already, I guess we'll just join our bio-bros.

[serpro69]

I think the "dramatic tone" here is more to bring awareness rather than to convey "sky is falling" meaning.
Besides, as @Etern4l already pointed out, some of the issues AI is already causing are quite dramatic as is, so should we sugarcoat and ignore it instead of pointing things out as they are?

[Etern4l]

3 hours ago, serpro69 said:

 @Etern4l

 

Reporting for duty :) 

 

BTW I'm skipping the whole host of issues already brought about by narrow AI (ANI). 

There is a Netflix documentary "The Social Dillema" which explains pre-ChatGPT concerns, which is always good since not everyone understands they have been more or less subtly manipulated by ANIs for a while now, and actually this guy Tristan Harris and his Institute for Humane Technologies has some good content on the impact of large language models on social media/online content.

 

This thread is more about the drama which is about to unfold. I appreciate that people not familiar with AI may have a hard time wrapping their heads around this deeply unpleasant material, but the bottom line is that unless we do something about this together, we are toast. 

 

Again, the first steps are obvious, and although clearly not trivial to execute, they beat trying to make it burrowed in the woods:

1) Demicrosoft, degoogle, demeta, deamazon and deenvidia (and probably best to deelon just to be safe, despite his helpful rhetoric) - in that order

2) Talk to other people

3) Talk to politicians

[kojack]

On 4/19/2023 at 3:47 AM, serpro69 said:

 

Yeah, no, I totally agree with this. I'm just a little annoyed when people call AI-generated stuff "art" is all 😄

 

Can't help but feel sorry for artists/designers/writers. They will be among the first ones to be affected by this bs. And the irony is that these models were trained on creator's own works. What a mess...

I just got hired by a very well known tech website to write for them. They are 100 percent against AI for their writing and photos/videos/graphics. I think my job is safe with them.  that's until the Terminators rise up against humanity....🤪

 

1 hour ago, Etern4l said:

 

Reporting for duty 🙂

 

BTW I'm skipping the whole host of issues already brought about by narrow AI (ANI). 

There is a Netflix documentary "The Social Dillema" which explains pre-ChatGPT concerns, which is always good since not everyone understands they have been more or less subtly manipulated by ANIs for a while now, and actually this guy Tristan Harris and his Institute for Humane Technologies has some good content on the impact of large language models on social media/online content.

 

This thread is more about the drama which is about to unfold. I appreciate that people not familiar with AI may have a hard time wrapping their heads around this deeply unpleasant material, but the bottom line is that unless we do something about this together, we are toast. 

 

Again, the first steps are obvious, and although clearly not trivial to execute, they beat trying to make it burrowed in the woods:

1) Demicrosoft, degoogle, demeta, deamazon and deenvidia (and probably best to deelon just to be safe, despite his helpful rhetoric) - in that order

2) Talk to other people

3) Talk to politicians

you left out dementia.